Circadian Patterns of Wearable-Derived Electrocardiographic Age and Left Atrial Remodeling in AF-Naïve Individuals

Abstract

Background: AI-enabled electrocardiographic age (AI-ECG age) is a digital biomarker of electrophysiological cardiac health. Although cardiovascular physiology exhibits circadian organization, the circadian behavior of AI-ECG age and its structural correlates have not been defined in AF-naive individuals. Objectives: To determine whether AI-ECG age exhibits reproducible circadian patterns and whether disruption of these patterns is associated with left atrial (LA) remodeling, a marker of atrial myopathy. Methods: Continuous single-lead wearable ECGs were analyzed from two independent prospective cohorts (S-Patch [ClinicalTrials.gov: NCT05119725, registered November 2021]; Memo Patch [ClinicalTrials.gov: NCT05355948, registered May 2022]). In AF-naive participants with 48 hours of data, AI-ECG age was estimated every 10 minutes. Unsupervised clustering was used to identify intrinsic circadian trajectories. For clinical interpretability, participants were classified using a day-night difference cutoff (Age 0.6 years) as Restorative (Age >0.6) or Disrupted (Age 0.6). We assessed phenotype reproducibility and examined associations with left atrial volume index (LAVI) using multivariable regression and meta-analysis. Results: Unsupervised learning consistently identified three circadian trajectory patterns across cohorts. Under the simplified binary classification, the Restorative phenotype was observed in approximately half of the participants (47.6-50.2%). Phenotype reproducibility was moderate (Cohen's 0.518; ICC=0.51-0.54) and was not fully explained by conventional heart rate variability measures. Among participants with echocardiography (n=122), the Disrupted phenotype was associated with higher LAVI (adjusted mean difference 6.09 mL/m2; 95% CI 1.46-10.72; p=0.010) and higher odds of severe LA enlargement (adjusted OR 4.17; 95% CI 1.58?10.99; p=0.004), with negligible heterogeneity (I2=0%). Conclusions: Wearable-derived AI-ECG age exhibits circadian patterns in AF-naive individuals, with unsupervised learning identifying distinct trajectories. Attenuation of a nocturnal decline the Disrupted phenotype is associated with left atrial enlargement, independent of conventional comorbidities and static AI-ECG age metrics. These findings suggest that circadian electrophysiological aging phenotyping may capture a dimension of atrial structural vulnerability not reflected by point-in-time assessments, and support prospective studies to evaluate its clinical utility.