Wearable smartwatches for atrial fibrillation detection and burden estimation after ablation: comparison with continuous monitoring Authors: Martin Aguilar, Laurent Macle, Ralph Chamieh, Paul Khairy, Marc W Deyell, Richard G Bennett, Jason G Andrade Published: 2025 ABSTRACT AbstractAimsWearable ECG-enabled smartwatches have been validated for atrial fibrillation (AF) screening, but their accuracy for monitoring AF recurrence and quantifying AF burden after catheter ablation is uncertain. To evaluate the simulated performance of three commercial smartwatch algorithms for AF recurrence detection and AF burden estimation compared with implantable cardiac monitors (ICMs).Methods and resultsUsing continuous ICM data from 346 patients in the CIRCA-DOSE trial, we simulated three smartwatch algorithms, assuming daytime wear (8:00 AM–10:00 p.m.). We also simulated commonly-used non-invasive intermittent rhythm monitoring strategies. Primary endpoints were sensitivity for arrhythmia recurrence and correlation with ICM-derived AF burden. Analyses were stratified by daily wear time and patient activity. AF recurrence occurred in 47.1% of patients. Simulated detection sensitivities were 82.2% (Apple Watch AF Burden), 70.6% (Apple Watch IRN), and 64.4% (Fitbit IHRD), compared with 15.8–64.6% for simulated intermittent AECG monitors. Wearables outperformed commonly-used Holter/patch monitoring strategies. AF burden correlation with ICM exceeded r = 0.97 for all algorithms. Among missed recurrences, median AF burden was <0.02%. Longer daily wear improved sensitivity (>90% with 24-hour use), whereas patient activity modestly reduced detection.ConclusionSmartwatch-based AF detection algorithms demonstrate strong correlation with ICM-derived AF burden and clinically good sensitivity for recurrence detection, outperforming conventional non-invasive strategies. These findings support the integration of wearables as a scalable alternative for post-ablation rhythm monitoring.Keywords: Atrial fibrillation, Catheter ablation, AF burden, Wearables, Implantable cardiac monitor This is the first study to benchmark commercially available smartwatch atrial fibrillation (AF) detection algorithms directly against continuous implantable cardiac monitoring in the post-ablation setting. This is the first study to benchmark commercially available smartwatch atrial fibrillation (AF) detection algorithms directly against continuous implantable cardiac monitoring in the post-ablation setting. Smartwatch algorithms (Apple Watch AF Burden, Apple Watch IRN, Fitbit IHRD) demonstrated good sensitivity for AF recurrence detection (64–82%), outperforming conventional Holter and patch monitoring strategies. Smartwatch algorithms (Apple Watch AF Burden, Apple Watch IRN, Fitbit IHRD) demonstrated good sensitivity for AF recurrence detection (64–82%), outperforming conventional Holter and patch monitoring strategies. Wearable-estimated AF burden was highly correlated with implantable cardiac monitor–derived burden (Pearson r >0.97), even under conditions of variable daily use and patient activity. Wearable-estimated AF burden was highly correlated with implantable cardiac monitor–derived burden (Pearson r >0.97), even under conditions of variable daily use and patient activity. Missed arrhythmia recurrences with wearables were of extremely low burden (median <0.02%), suggesting limited clinical relevance. Missed arrhythmia recurrences with wearables were of extremely low burden (median <0.02%), suggesting limited clinical relevance. These results support the integration of consumer smartwatches as a scalable, non-invasive alternative for post-ablation rhythm surveillance and AF burden assessment. These results support the integration of consumer smartwatches as a scalable, non-invasive alternative for post-ablation rhythm surveillance and AF burden assessment. Introduction Atrial fibrillation (AF) is the most prevalent sustained arrhythmia seen in clinical practice and is associated with increased morbidity, mortality and healthcare utilization.1 Catheter ablation is an established therapy for symptomatic AF; however, effective post-procedural rhythm monitoring is critical to assess arrhythmia recurrence and quantify AF burden. Current guidelines acknowledge the limitations of traditional follow-up methods, yet the optimal monitoring strategy after ablation has not been clearly defined.2 Importantly, AF detection and AF burden assessment are highly dependent on the diagnostic modality applied,3 as emphasized in the recent EHRA position paper on AF screening strategies.4 Increasingly sophisticated approaches, including digital devices, are being developed to extend monitoring duration and improve diagnostic yield. Intermittent, short-duration ambulatory ECG monitoring—such as Holter or event monitors—is widely used but suffers from low sensitivity, often underdetecting asymptomatic or low-burden episodes and consequently overestimating procedural success.3,5 While implantable cardiac monitors (ICMs) provide continuous, high-fidelity rhythm surveillance and are considered the gold standard, their invasive nature, cost, and limited scalability restrict their routine clinical use. In recent years, smartwatches equipped with photoplethysmography (PPG) or electrocardiography (ECG) have emerged as promising alternatives for non-invasive, direct-to-consumer rhythm monitoring.6 With growing consumer usage, one in three U.S. adults reported using a wearable device—wearables represent an attractive, non-invasive, and scalable method of rhythm monitoring.7 However, their clinical utility for long-term arrhythmia monitoring and quantification of AF burden, particularly in the context of catheter ablation, is poorly defined. In this study, we used continuous rhythm data from the CIRCA-DOSE randomized trial to simulate how three commercial wearable algorithms would identify arrhythmia recurrence and estimate AF burden. Using ICM as the gold standard, we assessed the detection sensitivity and accuracy of AF burden in these wearable-based algorithms for post-AF ablation follow-up care. This study aimed to provide the first systematic evaluation of commercially available smartwatch-based AF detection algorithms against continuous monitoring in a post-AF ablation population, and to compare their performance with that of commonly used non-invasive intermittent monitoring strategies. Methods Study population We analysed 346 patients undergoing a first catheter ablation for symptomatic paroxysmal AF in the Cryoballoon or Irrigated Radiofrequency Catheter Ablation randomized controlled trial (CIRCA-DOSE NCT01913522).8,9 All patients underwent insertion of an ICM (Reveal LINQ, Medtronic, Minneapolis) before ablation. The institutional ethics committee approved the study, along with the review board of each participating centre. The data supporting this study's findings are available from the corresponding author upon reasonable request. Simulated AF detection by wearables The ICM-derived AF burden was calculated as the per cent time in AF from the end of the blanking period (day 90) to the end of the study (day 365). We then simulated the AF detection delivered by (i) the Apple Watch AF burden algorithm, (ii) the Apple Watch Irregular Rhythm Notification algorithm and (iii) the Fitbit Irregular Heart Rhythm Detection algorithm. The Apple Watch AF burden algorithm was simulated as follows: A 1-minute window was checked every 15 min. The AF burden was calculated as the number of checks with AF divided by the total number of checks.10 The Apple Watch Irregular Rhythm Notification (IRN) algorithm was simulated as follows: a 1-minute window was checked every 2 h. If AF was present, a 1-minute window was checked every 15 min. If 3 out of 5 of these checks had AF, an Irregular Rhythm Notification alert was generated. The AF burden was calculated as the number of alerts divided by the total number of q2-hour checks.10 The Fitbit Irregular Heart Rhythm Detection (IHRD) algorithm was simulated as follows: a 5-minute window was checked every 2.5 min. If AF was detected on 11 consecutive checks, an Irregular Heart Rhythm Detection alert was generated. The AF burden was calculated as the total number of alerts divided by the number of q-2.5 min checks.11 The simulations had patients wear the watch from 8:00 a.m. to 10:00 p.m., unless otherwise stated. Statistical analyses Continuous variables are presented as medians with interquartile ranges (IQR) and were compared using the Wilcoxon signed-rank test for paired data. Categorical variables are expressed as frequencies and percentages. Kaplan-Meier estimates were used to calculate 1-year arrhythmia-free survival rates, and comparisons between monitoring strategies were made using the log-rank test. The detection sensitivity of wearable algorithms for arrhythmia recurrence was calculated relative to the ICM reference standard. To assess the agreement between AF burden estimates derived from each wearable algorithm and the ICM, we used Pearson correlation coefficients. Analyses were repeated stratified by daily watch usage (1 to 15 h per day) and by simulated patient activity levels (0%, 25%, 50%, and 75% probability of missing rhythm checks). Differences in AF burden distributions between methods were assessed using paired nonparametric tests. A two-tailed P-value <0.05 was considered statistically significant. All statistical analyses were performed using Prism 10 (GraphPad Software; San Diego, CA) and Matlab R2023a (MathWorks; Natick, MA). The data underlying this article will be shared on reasonable request to the corresponding author. Results Arrhythmia recurrence detection and AF burden A total of 163 patients (47.1%) experienced a first recurrence of any (symptomatic or asymptomatic) atrial tachyarrhythmia (atrial fibrillation, atrial flutter, or atrial tachycardia) between post-ablation day 90 and 365. The arrhythmia-free survival at 1 year using ICM was 52.6% (95% CI, 47.5–58.0%). The arrhythmia-free survival at 1 year using the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm, and the Fitbit IHRD algorithm were 61.3% (95% CI, 55.9–66.2%, P = 0.0165 vs. ICM), 66.8% (95% CI, 61.5–71.5%, P < 0.0001 vs. ICM) and 69.7% (95% CI, 64.5–74.2%, P < 0.0001), respectively (Figure 1). This translated to detection sensitivities (vs. ICM) of 82.2%, 70.6% and 64.4%, respectively. As a comparator, the 1-year arrhythmia-free survival obtained in the same patient population by simulating three 24-hour, 48-hour and, 7-day and 14-day ambulatory ECG (AECG) monitors at 3, 6 and 12 months was 92.5% (95% CI, 90.2–95.8), 88.4% (95% CI, 85.5–92.3%), 79.8% (95% CI, 75.9–84.4%) and 69.4% (64.8–74.6%), respectively. The sensitivities for the AECG monitors were 15.8% (95% CI, 8.9–20.7%), 24.5% (95% CI, 16.2–30.6%), 41.6% (95% CI, 32.9–50.8%) and 64.6% (95% CI, 53.6–74.3%), respectively.5 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. The median ICM-derived AF burden was 0.00% (IQR, 0.00–0.13%). The median AF burden calculated using the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm was 0.00% (IQR, 0.00–0.08%), 0.00% (IQR, 0.00–0.09%), and 0.00% (IQR, 0.00–0.03%), respectively (Figure 2A). Considering only patients with AT/AF recurrence (non-zero AF burden), the median AF burden calculated with ICM was 0.15% (IQR, 0.02–0.68%), the Apple Watch AF burden algorithm was 0.13% (IQR, 0.01–0.64%; P = 0.0071 vs. ICM), the Apple Watch IRN algorithm was 0.09% (IQR, 0.00–0.73%; P = 0.0010 vs. ICM), and the Fitbit IHRD algorithm was 0.05% (IQR, 0.00–0.48%; P < 0.0001 vs. ICM), respectively (Figure 2B). 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. The median ICM-derived AF burden in those patients missed by the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm was 0.003% (IQR, 0.001–0.026%), 0.004% (IQR, 0.001–0.022%) and 0.004% (IQR, 0.002–0.034%), respectively (Figure 2C). There were 0/89 (0%), 1/89 (1.1%) and 2/89 (2.2%) patients with ICM-derived AF burden >0.1% that were missed by the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm. The correlation between the ICM-derived AF burden and the AF burden obtained with the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm was excellent for the study population as a whole (Pearson r, 0.98, 0.99 and 0.97, respectively; Figure 3) as well as when considering only patients with AT/AF recurrence (non-zero AF burden; Pearson r, 0.98, 0.99, 0.97, respectively; Supplementary material online, Figure S1). These results are summarized in Table 1. 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. Table 1Summary of AF recurrence detection and burden estimation performance across monitoring strategiesArrhythmia-free survival (%)Detection sensitivity (%)Median AF burden (Recurrences only; %)Correlation (Pearson r vs. ICM)Median AF burden in missed detections (%)ICM52.61000.15——Apple Watch AF Burden61.382.20.130.98<0.01Apple Watch IRN66.870.60.090.99<0.01Fitbit IHRD69.764.40.050.97<0.01AECG 24 h (x3)92.515.81.000.410.09AECG 48 h (x3)88.424.50.880.570.08AECG 7d (x3)79.841.60.360.750.06AECG 14d (x3)69.464.60.220.850.03Comparison of arrhythmia-free survival, detection sensitivity, AF burden (in patients with recurrence), correlation with ICM-derived burden, and AF burden in missed detections across different monitoring modalities. Wearable smartwatch algorithms (Apple Watch AF Burden, Apple Watch IRN, Fitbit IHRD) were simulated using daytime use (8:00–22:00) and benchmarked against implantable cardiac monitors (ICMs) and commonly used intermittent ambulatory ECG (AECG) strategies. Values represent performance from day 90 to 365 post-ablation. Summary of AF recurrence detection and burden estimation performance across monitoring strategies Comparison of arrhythmia-free survival, detection sensitivity, AF burden (in patients with recurrence), correlation with ICM-derived burden, and AF burden in missed detections across different monitoring modalities. Wearable smartwatch algorithms (Apple Watch AF Burden, Apple Watch IRN, Fitbit IHRD) were simulated using daytime use (8:00–22:00) and benchmarked against implantable cardiac monitors (ICMs) and commonly used intermittent ambulatory ECG (AECG) strategies. Values represent performance from day 90 to 365 post-ablation. Impact of daily watch usage on detection characteristics Simulations of 1-year arrhythmia-free survival rates detected by wearable devices decreased with increasing daily watch usage from 76.6% (95% CI, 71.8–80.7%), 78.9% (95% CI, 74.2–82.8%) and 80.3% (95% CI, 75.8–84.2%) with 1 h of daily usage of Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively, to 61.3% (55.9–66.2%), 66.8% (95% CI, 61.5–71.5%) and 69.1% (95% CI, 63.9–73.7%), respectively with 15 h of daily usage (Figure 4A). This translated to an increase in detection sensitivity (vs. ICM) from 0.50, 0.45 and 0.42 with the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively, with 1 h of daily watch usage to 0.82, 0.71 and 0.66, respectively, with 15 h of daily watch usage (Figure 4B). For the theoretical scenario of 24-hour of daily watch usage, the 1-year arrhythmia-free survival for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm would have been of 57.5%. (95% CI, 52.5–62.9%), 63.0% (95% CI, 57.7–67.9%) and 63.9% (95% CI, 58.6–68.7%) (see Supplementary material online, Figure S2), respectively, returning sensitivities of 90.2%, 78,5% and 76.7%, respectively. 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. The correlation (Pearson r) between ICM-derived and watch-derived AF burden ranged from 0.96, 0.96 and 0.94 with 1 h of daily watch usage to 0.99, 0.99 and 0.99 with 15 h of daily watch usage for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively (Figure 4C). The ICM-derived AF burden in those patients with missed recurrence detection was very low, ranging from 0.020% (25–75th percentile, 0.003–0.082%), 0.025% (25–75th percentile, 0.003–0.093%) and 0.028% (25–75th percentile, 0.003–0.99%) with 1 h of daily watch usage to 0.003% (25–75th percentile, 0.001–0.021%), 0.004% (25–75th percentile, 0.001–0.021%) and 0.004 (25–75th percentile, 0.002–0.029%), with 15 h of daily watch usage with the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively (Figure 4D). The number of patients with missed recurrence detection and ICM-derived AF burden >0.1% ranged from 17/89 (19.1%), 21/89 (23.6%) and 24 (26.9%) with 1 h of daily watch usage to 0/89 (0%), 1/89 (1.1%) and 2/89 (2.2%), with 15 h of daily watch usage, with the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively. Impact of patient activity on detection characteristics We simulated 4 different levels of patient activity corresponding to a 0% (‘None’), 25% (‘Low’), 50% (‘Intermediate’) and 75% (‘High’) probability that any given rhythm check would be missed because of patient activity. The calculated 1-year arrhythmia-free survival increased with increasing activity levels (Figure 5A), ranging from 61.3% (95% CI, 55.9–66.2%), 66.8% (95% CI, 61.5–71.5%) and 69.7% (95% CI, 64.5–74.2%) with no patient activity (‘None’) to 66.5% (95% CI, 61.2–71.2%), 75.1% (95% CI, 70.2–79.4%) and 76.6% (95% CI, 71.8–80.7%) with the highest level of patient activity (‘High’), for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively (Figure 5A). This corresponded to a decrease in detection sensitivity from 0.82, 0.71 and 0.64 with no patient activity (‘None’) to 0.71, 0.52 and 0.50 with the highest level of patient activity (‘High’), for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively (Figure 5B). 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. The correlation (Pearson r) between the ICM-derived AF burden and the wearable-estimated AF burden remained >0.96 across all levels of patient activity, for all 3 algorithms (Figure 5C). The median ICM-derived AF burden in those patients with missed recurrence detection was <0.01% across all algorithms and activity levels (Figure 5D). The number of patients with missed recurrence detection and ICM-derived AF burden >0.1% was 0/89 (0%), 1/89 (1.1%) and 2/89 (2.2%) with no patient activity, 1/89 (1.1%), 2/89 (2.2%) and 3/89 (3.4%) with low patient activity and 2/89 (2.2%), 14/89 (15.7%) and 13/89 (14.6%), with the highest level of patient activity, for the Apple Watch AF burden algorithm, the Apple Watch IRN algorithm and the Fitbit IHRD algorithm, respectively. In additional simulations restricted to nocturnal wear (00:00–08:00), sensitivities were 72.4% for the Apple Watch AF burden algorithm, 62.6% for Apple Watch IRN, and 65.6% for Fitbit IHRD. These values were similar to those obtained during daytime use and in some scenarios outperformed daytime monitoring under high-noise/activity conditions. Discussion The main findings of this study are that (i) wearable smartwatch algorithms specifically, the Apple Watch and Fitbit algorithms, showed strong sensitivity for detecting arrhythmia recurrence after AF ablation with detection sensitivities as high as 82%, when benchmarked against ICM; (ii) these algorithms delivered AF burden estimates which were strongly correlated with AF burden derived from continuous invasive monitoring (Pearson r > 0.97); and (iii) the AF burden in patients with missed detection by wearables was extremely low (median <0.01%). These findings support the integration of wearable devices as a viable, non-invasive strategy for rhythm monitoring following AF ablation. AF detection by wearables An ever-growing number of direct-to-consumer wearable devices are available on the market. Several studies have evaluated the accuracy of smartwatch algorithms in detecting AF.12–24 The Apple Heart Study evaluated the AF detection capabilities of the Apple Watch in 419 297 participants without known AF—individuals who received an irregular heart rate notification were sent a 7-day ambulatory ECG patch.25 In this study population with a very low AF prevalence, the positive predictive value of an irregular heart rate notification was high (0.84; 95% CI, 0.76–0.92). Similarly, the Fitbit Heart Study employed a similar study design to evaluate Fitbit’s PPG-based algorithm, yielding favourable detection metrics (positive predictive value, 98.2%; sensitivity, 67.5%; specificity, 98.4%).11 Along the same line, the WATCH-AF trial reported a sensitivity of 93.7% and specificity of 98.2% for detecting AF with PPG-based wearables, using single-lead ECG as the comparator.26 A recent meta-analysis of 15 studies totalling 12 802 participants found a pooled sensitivity and specificity for AF detection for PPG-based smartwatch algorithms of 97.4% (95% CI, 96.5–98.3%) and 96.6% (95% CI, 94.9–98.3%), respectively.27 Importantly, all these studies have used intermittent non-invasive monitoring, in the form of Holter monitors or patches, as the benchmark against which wearables were compared. Hence, these reports should be interpreted within the significant limitations of non-invasive monitoring for detecting AF.5 How the AF detection capabilities of smartwatch algorithms compare to those of an ICM, the gold standard, is unknown. Furthermore, a large proportion of the aforementioned studies evaluated wearables for AF screening in individuals without known AF (characterized by a very low AF prevalence). We found a single study28 that assessed the AF recurrence detection capabilities of a smartwatch system (JKwear 1, Chengdu CVhealth Science and Technology Co., Ltd; not available in North America or Europe) specifically in the post-AF ablation setting, albeit using a 24-hour ECG patch monitor as the comparator. To our knowledge, this is the first study to systematically evaluate the performance of wearable AF detection algorithms against ICMs in the post-AF ablation setting. Leveraging continuous, adjudicated rhythm data from the CIRCA-DOSE randomized trial, we simulated the performance of three commercially available smartwatch algorithms (Apple Watch AF Burden, Apple Watch IRN, and Fitbit IHRD) over a full year of follow-up. This design offers several key strengths. First, by benchmarking wearable detection against ICMs, the gold standard for arrhythmia surveillance, we provide the most rigorous validation of smartwatch-based monitoring to date. Second, we evaluated both recurrence detection and AF burden estimation, addressing the growing clinical emphasis on burden quantification as a therapeutic endpoint. Third, we examined the impact of real-world factors, such as variable daily watch usage and patient activity, on detection sensitivity and signal fidelity, thereby offering practical insights into algorithm robustness. Finally, by applying multiple algorithms to the same patient cohort, we were uniquely positioned to compare their operational characteristics under identical clinical conditions. Collectively, these strengths position our study as a critical step forward in validating wearable devices for post-ablation rhythm monitoring. Wearables vs. non-invasive intermittent monitoring We found that wearables delivered good 1-year arrhythmia-free survival, recurrence detection sensitivities, and AF burden estimation. All three algorithms matched or outperformed three 14-day AECG monitors (worn at 3, 6 and 12 months), which is a much more intensive monitoring strategy than routine clinical practice or clinical trials. Moreover, even minimal watch usage (1 h per day) yielded an arrhythmia recurrence detection sensitivity superior to that of three 7-day AECG monitors. These findings are clinically relevant, given the widespread limitations of patch monitors, including poor patient compliance and frequent signal noise. There is a growing interest in reporting AF burden in addition to the binary time-to-recurrence metric. We found that AF burden estimates derived from wearables are highly correlated (Pearson r > 0.97) with the gold-standard ICM-derived burden, even for high levels of patient activity or low daily watch usage. Furthermore, the AF burden in patients with missed recurrence detection was extremely low (median <0.02%), much below the 0.1% threshold for clinical relevance.29 Although we did not formally evaluate the cost-effectiveness of wearables for post-ablation arrhythmia detection, their superior arrhythmia detection characteristics and competitive pricing vs. non-invasive intermittent monitoring with Holter or ECG patches make wearables and a priori attractive option. Furthermore, smartwatches enable the tracking of other potentially clinically relevant metrics, such as pre- vs. post-ablation activity levels, which are not available with most Holter/patch systems.30 AF detection algorithms Our study was not designed to compare the algorithms with one another. All three algorithms displayed similar dependence on daily watch usage and patient activity level for optimal detection operational characteristics. The Apple Watch AF burden algorithm nevertheless had the highest arrhythmia detection across the different conditions evaluated in our study, whereas the Fitbit’s Irregular Heart Rate Detection algorithm had the lowest. This difference is likely the consequence of how the algorithms are designed. The Apple Watch algorithm checks for AF every 15 min and calculates the burden as the percentage of checks with an irregular rhythm. The Fitbit has a more stringent algorithm that requires 11 consecutive checks to have an irregular rhythm before generating a detection event. Given that a significant proportion of AF/AT episodes were of short duration (2 min), the Fitbit algorithm, which requires 11 consecutive 5-minute segments (staggered every 2.5 min), would miss arrhythmia episodes of duration less than 25 min. We were unable to directly evaluate how this more stringent criterion affects specificity. However, most reports evaluating the Apple Watch returned very high specificity values, suggesting that the less stringent Apple Watch algorithm retains good operational characteristics. Our data support the conclusion that frequent and systematic sampling—as modelled in the Apple Watch AF burden algorithm—yields more consistent estimates of AF burden than less frequent or alert-based strategies. A limitation of PPG-based detection algorithms is non-interpretable tracings due to patient activity or inadequate watch placement. The proportion of non-interpretable tracings has been reported to be in the 20%-range.31,32 Zhao and colleagues reported a significant difference in the proportion of day vs. nighttime invalid tracings, with almost 60% of daytime tracings being reported as invalid for interpretation.28 We evaluated a wide range of simulated patient activity levels, corresponding to 0% to 75% of input tracings being considered invalid for interpretation. We found that the Apple Watch AF burden algorithm was fairly robust in terms of maintaining good AF/AT detection sensitivity and excellent AF burden correlation even with significantly degraded input data. Conversely, the Fitbit IHRD algorithm’s detection characteristics were impacted considerably by patient activity. Newer artificial intelligence-based smartwatch algorithms may enhance the diagnostic accuracy and utility of smartwatch ECGs for detecting atrial arrhythmias.33,34 Our findings align with the EHRA Practical Guide on the use of digital devices for arrhythmia management, which emphasizes the importance of tailoring monitoring strategies to patient characteristics, device capabilities, and real-world adherence considerations.35 Incorporating such guidance reinforces the potential role of smartwatch-based AF detection as part of a broader digital health ecosystem for post-ablation follow-up. Clinical implications The findings of this study have important implications for post-ablation care in patients with AF. Smartwatch-based monitoring can significantly enhance post-ablation surveillance by improving detection of AF recurrence and accurately estimating AF burden compared with conventional strategies. Even modest daily use provided better recurrence detection than repeated Holter or patch monitors, which are resource-intensive and prone to patient non-compliance. The extremely low AF burden in missed cases (<0.02%) suggests that undetected episodes are unlikely to be clinically meaningful. These results support the integration of wearables into clinical care pathways as a cost-effective, patient-centred alternative to intermittent monitoring, with potential to optimize follow-up, reduce unnecessary testing, and inform guideline recommendations for post-ablation monitoring. In contrast to 14-day patch monitors or event recorders, which offer only intermittent snapshots, wearables provide an opportunity for continuous, scalable, and patient-directed rhythm surveillance. Furthermore, the deployment of smartwatches enables a wide range of digital AF surveillance paradigms, including tracking physical activity, sleep, and heart rate variability, which provides a multidimensional perspective on recovery following ablation. These functions can facilitate the early identification of AF recurrence, promote lifestyle interventions, and enable remote clinical decision-making to enhance patient outcomes and optimize healthcare resource utilization. The clinical importance of AF burden assessment has been underscored in the recent ESC/EHRA consensus statement, which defines AF burden as the proportion of time spent in AF during a specified monitoring period.36 AF burden has been linked to stroke risk, heart failure progression, cognitive decline, renal dysfunction, and mortality, and is increasingly recognized as a meaningful endpoint in clinical practice and research.29,37–42 Recent ablation trials have begun to incorporate AF burden reduction, rather than recurrence alone, as an outcome, reflecting that substantial reductions—even without complete elimination—can yield meaningful clinical benefit. Our findings, showing excellent correlation between smartwatch-estimated and ICM-derived AF burden, suggest that consumer wearables may offer a scalable, non-invasive method for capturing this clinically relevant endpoint in both trials and routine follow-up. Limitations This study has several limitations. First, the simulations were based on retrospective application of wearable detection algorithms to ICM-derived rhythm data. As such, this study represents an ‘ideal state’ comparison of the wearables, assuming ideal compliance with wearable use (i.e. worn daily) and performance of the PPG-based detection algorithms. A significant limitation of wearable devices is limited compliance (e.g. < 40% of those with cardiovascular disease report daily use), the relatively high rate of inconclusive diagnoses (up to 25% of episodes), as well as false-positive and false-negative detections of AF. The impact of over- and under-detection, especially in noisy signal conditions or during exercise, remains to be investigated. However, a recent real-world validation study in patients with established AF demonstrated that PPG-based monitoring was generally reliable, with the main limitation being a modest overestimation of heart rate.43 Second, we assumed wearing during daytime from 8:00 a.m. to 10:00 p.m., excluding overnight monitoring. This creates a bias against wearable-based detection, but it is also the most common usage of smartwatches, given that they must be charged. Third, the studied algorithms were based on public documentation descriptions. Although these simulations model their functioning, we cannot account for proprietary software improvements or future algorithmic developments. Fourth, for modelling purposes, all atrial tachyarrhythmias (AF, atrial flutter, and atrial tachycardia) were analysed collectively; however, the performance of the smartwatch algorithm may vary between AF and organized atrial tachyarrhythmias. However, only 3 of the 163 (1.8%) patients with arrhythmia recurrence had recurrence only in the form of AT/AFL. Moreover, the rate of recurrence in the form of AT/AFL is less after ablation of paroxysmal vs. persistent AF. Fifth, the patient population in this study had paroxysmal AF, and it remains unknown whether our findings can be extrapolated to those with persistent AF or those with different degrees of AF burden. Finally, estimating AF burden with smartwatches may require a certain degree of digital literacy, which could limit widespread implementation in real-world practice, particularly among older patients.44 Conclusion Wearable ECG-based smartwatches demonstrate an excellent correlation with implantable cardiac monitors for assessing AF burden and present reasonable sensitivity for detecting post-ablation arrhythmia recurrences. These findings support the integration of wearable ECG devices into post-ablation follow-up strategies, offering a scalable alternative to traditional monitoring. Supplementary Material Contributor Information Martin Aguilar, Montreal Heart Institute, Department of Medicine, Université de Montréal, 5000 Rue Bélanger, Montréal, QC, Canada H1T 1C8. Laurent Macle, Montreal Heart Institute, Department of Medicine, Université de Montréal, 5000 Rue Bélanger, Montréal, QC, Canada H1T 1C8. Ralph Chamieh, Montreal Heart Institute, Department of Medicine, Université de Montréal, 5000 Rue Bélanger, Montréal, QC, Canada H1T 1C8. Paul Khairy, Montreal Heart Institute, Department of Medicine, Université de Montréal, 5000 Rue Bélanger, Montréal, QC, Canada H1T 1C8. Marc W Deyell, Department of Medicine, Dilawri Cardiovascular Institute, 2775 Laurel St Vancouver, Canada BC V5Z 1M9; Center for Cardiovascular Innovation, Life Sciences Centre, 2350 Health Sciences Mall, Vancouver, Canada BC V6T 1Z3. Richard G Bennett, Department of Medicine, Dilawri Cardiovascular Institute, 2775 Laurel St Vancouver, Canada BC V5Z 1M9; Center for Cardiovascular Innovation, Life Sciences Centre, 2350 Health Sciences Mall, Vancouver, Canada BC V6T 1Z3. Jason G Andrade, Montreal Heart Institute, Department of Medicine, Université de Montréal, 5000 Rue Bélanger, Montréal, QC, Canada H1T 1C8; Department of Medicine, Dilawri Cardiovascular Institute, 2775 Laurel St Vancouver, Canada BC V5Z 1M9; Center for Cardiovascular Innovation, Life Sciences Centre, 2350 Health Sciences Mall, Vancouver, Canada BC V6T 1Z3. Supplementary material Supplementary material is available at Europace online. Funding Dr. Khairy is supported by the André Chagnon Research Chair in Electrophysiology and Congenital Heart Disease. References Associated Data Supplementary Materials