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

This study demonstrates that wearable-derived AI-ECG age exhibits reproducible circadian patterns in AF-naive individuals, where the disruption of a normal nocturnal decline is significantly associated with left atrial enlargement, suggesting that circadian electrophysiological aging captures structural atrial vulnerability beyond static assessments.

The Gist

The Big Idea: Your Heart Has a "Biological Clock"

Imagine your heart isn't just a pump that runs at a steady speed. Instead, think of it like a smart building that has a daily routine. Just as you feel energetic in the morning and sleepy at night, your heart goes through its own daily cycle of "working hard" and "resting deeply."

In a healthy heart, the "resting" phase at night is like a deep, rejuvenating sleep. The heart's electrical system slows down and resets, getting ready for the next day. This paper is about a new way to measure whether your heart is getting that good night's sleep or if it's staying "awake" and stressed when it should be resting.

The New Tool: The "Heart Age" Watch

Scientists have developed a special AI (Artificial Intelligence) that looks at your heart's electrical signals (from a simple wearable patch, like a smartwatch) and tells you your "Heart Age."

• Static Heart Age: Usually, this is like taking a single photo of your heart. It tells you, "Your heart looks like it belongs to a 60-year-old."
• Circadian Heart Age: This study used a video camera instead of a photo. It watched the heart age change every 10 minutes for two days straight.

What They Found: The "Rejuvenator" vs. The "Insomniac"

By watching the heart age change over 48 hours, the researchers noticed that people fall into two main groups, similar to how people sleep:

1. The "Rejuvenators" (Healthy Pattern):

   • The Analogy: Imagine a battery that charges up fully at night.
   • What happens: During the day, their "Heart Age" is a bit higher (maybe 60 years old). But when they sleep, the number drops significantly (maybe down to 58).
   • Meaning: Their heart is successfully "recharging" and resetting its electrical system while they sleep. This is a sign of a healthy, flexible heart.

2. The "Disrupted" or "Insomniacs" (Unhealthy Pattern):

   • The Analogy: Imagine a battery that stays plugged in but never actually charges, or even gets hotter at night.
   • What happens: Their "Heart Age" stays high during the day, and it doesn't drop at night. In some cases, it even goes up slightly while they sleep.
   • Meaning: Their heart is stuck in "work mode" even when it should be resting. It's like a car engine that keeps revving while parked.

The Big Discovery: The "Insomniac" Heart is Stiffer

The most important finding of this study is what happens to the heart structure in these two groups.

• The Problem: The researchers looked at the size of the Left Atrium (the heart's main receiving room). When this room gets too big, it's a sign of "atrial cardiomyopathy"—basically, the heart muscle is getting stretched and scarred, which is a major risk factor for developing Atrial Fibrillation (AFib), a dangerous irregular heartbeat.
• The Result: People with the "Disrupted" (Insomniac) pattern had significantly larger, more stretched-out heart rooms compared to the "Rejuvenators."
• The Twist: This was true even if the people had the same "static" heart age, the same blood pressure, and the same history of diabetes or high blood pressure. The pattern of the heart's daily rhythm was the secret clue that the standard check-up missed.

Why This Matters (The "So What?")

Think of it like checking a house for damage.

• Old Way: You walk in and look at the walls. They look fine. You say, "This house is safe."
• New Way: You watch the house for two days. You notice that every night, the foundation cracks slightly and doesn't fully heal by morning. Even though the walls look fine, the house is actually unstable.

This study suggests that how your heart behaves at night is a crucial warning sign. If your heart doesn't "dip" (get younger/younger) at night, it might be a sign that your heart is starting to get damaged, even before you feel any symptoms or before you get a diagnosis of irregular heartbeat.

The Takeaway for You

1. Your heart has a rhythm: It should get "younger" and more relaxed at night.
2. Wearables are powerful: A simple patch on your chest can detect these subtle daily rhythms that a standard 10-second ECG in a doctor's office would miss.
3. Early Warning: If your heart doesn't rest at night, it might be a sign that your heart structure is changing. This could help doctors catch heart problems before they become serious, allowing for earlier intervention.

In short: A healthy heart knows how to sleep. If your heart's "digital age" doesn't drop while you sleep, it might be telling you that your heart is under stress and needs attention.

Technical Summary

1. Problem Statement

• Context: Physiological aging is a primary driver of cardiovascular disease and atrial fibrillation (AF). Artificial Intelligence (AI) has enabled the creation of "Electrocardiographic Age" (AI-ECG age), a digital biomarker that predicts mortality and cardiovascular outcomes.
• Gap: While AI-ECG age is typically derived from static, hospital-based 12-lead ECGs, recent advances allow for its estimation from continuous, single-lead wearable ECGs. However, the circadian behavior of AI-ECG age in free-living conditions has not been defined.
• Hypothesis: Cardiovascular physiology exhibits circadian organization (e.g., blood pressure dipping, heart rate variability). The authors hypothesized that AI-ECG age exhibits reproducible circadian patterns in AF-naïve individuals and that the disruption of these patterns (specifically, a lack of nocturnal decline) is associated with adverse structural cardiac remodeling, specifically Left Atrial (LA) enlargement.

2. Methodology

Study Design & Population:

• Data Source: Two independent prospective cohorts: S-Patch (n=286, 72h monitoring) and Memo Patch (n=213, 7–14 days monitoring).
• Inclusion Criteria: AF-naïve individuals (no prior diagnosis and no AF detected during monitoring) with ≥48 hours of high-quality single-lead ECG data.
• Total Analyzed: 499 participants after exclusions for AF episodes and data quality.

Data Processing & Feature Extraction:

• AI-ECG Age Estimation: Utilized the PROPHECG-Age Single model to estimate AI-ECG age every 10 minutes over the first 48 hours of monitoring.
• Circadian Feature Derivation:
  • Cosinor Analysis: Extracted amplitude, trough phase (bathyphase), and $R^2$.
  • Frequency Analysis: Used Fast Fourier Transform (FFT) to identify dominant periods.
  • Stability Metrics: Calculated self-similarity (Pearson correlation between day 1 and day 2) and within-hour variability.
• Clustering: Applied Kernel PCA (RBF kernel) followed by K-means clustering on standardized features to identify intrinsic circadian trajectories without prior assumptions.

Phenotype Classification:

• Binary Definition: Defined the day–night gap ($\Delta$Age) as: $\text{Mean Day (10:00–16:00)} - \text{Mean Night (00:00–06:00)}$.
• Threshold: A cutoff of 0.6 years was established to classify participants into:
  • Restorative (Rejuvenator): $\Delta$Age > 0.6 (Normal nocturnal decline).
  • Disrupted (Riser/Irregular): $\Delta$Age $\le$ 0.6 (Blunted decline or paradoxical increase).

Statistical Analysis:

• Reproducibility: Assessed using Cohen's $\kappa$, Fleiss' $\kappa$, and Intraclass Correlation Coefficient (ICC) across consecutive 48-hour epochs.
• Association with Remodeling: Analyzed the association between the Disrupted phenotype and Left Atrial Volume Index (LAVI) using multivariable linear regression (adjusted for age, sex, BMI, baseline AI-ECG age) and logistic regression for severe LA enlargement (LAVI > 39 mL/m²).
• Meta-analysis: Combined results from both cohorts using inverse-variance weighting.

3. Key Contributions

1. First Characterization of Circadian AI-ECG Age: Demonstrated that AI-ECG age is not static but follows distinct, reproducible circadian trajectories in AF-naïve individuals.
2. Novel Phenotyping: Identified three unsupervised clusters (Rejuvenator, Riser, Irregular) and translated them into a clinically actionable binary phenotype (Restorative vs. Disrupted).
3. Structural Correlate Discovery: Established a link between the attenuation of nocturnal AI-ECG age decline and Left Atrial Remodeling, independent of conventional comorbidities and static AI-ECG age metrics.
4. Sex-Specific Insights: Revealed that the association between the Disrupted phenotype and LA enlargement is significantly stronger in females than in males.

4. Key Results

Circadian Patterns:

• Unsupervised learning consistently identified three patterns across both cohorts: Rejuvenator (pronounced nocturnal decline), Riser (blunted/paradoxical increase), and Irregular (no rhythmicity).
• The Restorative phenotype (nocturnal decline) was observed in approximately 48–50% of participants.
• Reproducibility: The binary phenotype showed moderate temporal stability (Cohen's $\kappa$ = 0.518; ICC = 0.51–0.54), suggesting it captures a recurring physiological trait rather than random noise.

Association with Left Atrial Remodeling:

• LAVI Association: Participants with the Disrupted phenotype had significantly higher Left Atrial Volume Index (LAVI) compared to the Restorative group.
  • Pooled Adjusted Mean Difference: +6.09 mL/m² (95% CI: 1.46–10.72; $p=0.010$).
  • Heterogeneity: Negligible ($I^2 = 0\%$).
• Severe Enlargement: The Disrupted phenotype was associated with a 4.17-fold increased odds of severe LA enlargement (LAVI > 39 mL/m²) (95% CI: 1.58–10.99; $p=0.004$).
• Sex Stratification: The association was robust in females (Mean diff: +10.11 mL/m²; $p=0.001$) but not statistically significant in males.

Specificity & Independence:

• The Disrupted phenotype was not associated with traditional comorbidities (hypertension, diabetes, heart failure) or static mean AI-ECG age.
• While HRV metrics (SDNN, RMSSD) were lower in the Disrupted group in one cohort, the circadian AI-ECG age phenotype provided complementary information not captured by standard HRV or static age metrics.

5. Significance and Implications

• New Biomarker Dimension: This study introduces "circadian electrophysiological aging" as a distinct dimension of cardiac health. It suggests that the pattern of aging (how the heart recovers at night) is as critical as the level of aging (static age).
• Early Detection of Atrial Cardiomyopathy: The findings suggest that circadian disruption in AI-ECG age may serve as an upstream, non-invasive digital biomarker for early atrial structural vulnerability (LA remodeling) before the onset of clinical AF.
• Clinical Utility: The "Disrupted" phenotype identifies a high-risk subgroup that might be missed by standard risk scores or single-timepoint ECGs. This supports the potential for wearable devices to monitor nocturnal electrophysiological recovery as a prognostic tool.
• Future Directions: The authors call for prospective longitudinal studies to determine if the Disrupted phenotype predicts the incident development of AF and to investigate the mechanisms behind the observed sex-specific differences.

Limitations Noted:

• Cross-sectional design (no causal inference).
• Reliance on fixed clock times for day/night windows rather than individual sleep cycles.
• Echocardiographic data available only in a subset (n=122).
• Predominantly East Asian cohort.