This study analyzed heart rate variability (HRV) data collected via wearable BioPatch devices from 458 patients with chronic kidney disease (CKD), revealing the impact of clinical factors such as diabetes and proteinuria on HRV. Using SDNN (Standard Deviation of NN Intervals) as the core HRV metric, the study evaluated diurnal HRV variation and disease-related associations in CKD patients. The results confirmed that diabetes and high proteinuria are significantly associated with reduced SDNN, laying an important foundation for future biomarker development aimed at predicting cardiovascular disease risk and improving clinical outcomes in CKD patients.

1. Study Background and Rationale

The prognostic value of continuous ambulatory biomonitoring of cardiovascular function among patients with chronic kidney disease (CKD) has not been well characterized. Both cardiac and renal function are regulated by molecular biological clocks and maintain healthy homeostasis by adapting to physiological demands that vary on a 24-hour cycle. For instance, non-dipping hypertensive patients have higher mortality than those whose blood pressure dips at night, illustrating that the loss of time-dependent physiological phenomena is deeply associated with various diseases. In night-shift models, disrupted sleep-wake rhythms increase blood pressure and inflammatory markers, suppress heart rate variability (HRV), and blunt immune-related gene expression profiles.

While wearable device data is increasingly used in clinical medicine for disease risk prediction and classification, the prognostic value of ambulatory biomonitoring of cardiovascular function in CKD patients had remained an unexplored domain. Previous work had found that HRV measured via 12-lead resting ECG predicts mortality in CKD patients, but this study went further by focusing on collecting and deeply analyzing HRV data from patients' daily lives using wearable devices.

2. Study Participants and Data Collection

This study was conducted with participants from the Chronic Renal Insufficiency Cohort (CRIC) study. From 2019 to 2023, ECG data averaging 50.3 +/- 9.3 hours was collected from 458 participants across seven CRIC centers using wearable BioPatch devices. This research-grade wearable records cardiovascular and respiratory activity. Participants were instructed to wear the device for at least 48 hours to capture a minimum of two complete 24-hour circadian cycles, enabling more accurate characterization of biological rhythms.

The average age of participants was 68.6 +/- 9.7 years, 45% were female, 49.3% had diabetes, and 30.8% had a history of cardiovascular disease. The study population was generally healthier than other patients who consented to CRIC Phase 4 but were not included. After data quality control, records from 125 participants were excluded, mostly due to HRV trace gaps exceeding 5 consecutive hours.

3. Key Metric: SDNN and Its Significance

This study used SDNN (Standard Deviation of NN Intervals) as the primary HRV measurement. SDNN represents the standard deviation of normal RR intervals (intervals between heartbeats), an important indicator reflecting the heart's autonomic nervous system regulatory capacity. Low SDNN values indicate impaired autonomic regulation, which is associated with increased cardiovascular disease risk.

Participants' SDNN data was divided into tertiles for analysis:

• Low SDNN group: <= 33.7 ms (152 participants)
• Middle SDNN group: 33.8 - 48.8 ms (153 participants)
• High SDNN group: >= 49.0 ms (153 participants)

The clinically established normal reference for short-term SDNN is 50 +/- 16 ms, with low SDNN values associated with increased cardiovascular risk. The average SDNN value in this study (50.3 +/- 9.3 ms) sits at a critical threshold where the steep association between low SDNN and elevated mortality risk (51%) begins. In other words, even a small further decrease in SDNN could significantly increase mortality risk.

4. Association Between Disease Burden and Low Heart Rate Variability

This study examined in detail how various clinical factors -- diabetes, proteinuria, and kidney function -- affect SDNN.

4.1. Diabetes and SDNN Reduction

Approximately half of all participants had diabetes. Of participants in the low SDNN group, 65.8% had diabetes, compared to only 36.6% in the high SDNN group. This suggests that diabetes is closely associated with SDNN reduction. Hemoglobin A1C (HbA1C) levels were also significantly higher in the low SDNN tertile group (6.85 +/- 1.69%) compared to the middle and high SDNN groups (6.18 +/- 0.98% and 6.05 +/- 1.10%, respectively; p < 0.0001).

Multivariable linear regression analysis revealed that even after adjusting for other covariates, diabetes status was significantly associated with a 7.4 ms reduction in SDNN (p = 0.001).

"In multivariable linear regression, participants with diabetes had SDNN that was 7.4ms lower than those without diabetes (p = 0.001)."

This provides important evidence that diabetic patients have impaired cardiac autonomic regulatory function.

4.2. Proteinuria and SDNN Reduction

Participants with high urine protein-to-creatinine ratio (uPCR >= 0.2) had SDNN that was 5.73 ms lower than those with low uPCR (< 0.2) (p = 0.011). These results once again underscore that proteinuria is an independent risk factor for elevated cardiovascular disease risk in CKD patients.

4.3. Interaction Between Kidney Function and SDNN

Participants with eGFR below 45 mL/min/1.73 m2 had SDNN 5.32 ms higher than those with eGFR 45 or above, but this was primarily explained by a significant interaction between uPCR and eGFR (p = 0.023). Specifically, participants with both impaired kidney function and high proteinuria (eGFR < 45, uPCR >= 0.2) had substantially lower SDNN than those with impaired kidney function but low proteinuria (eGFR < 45, uPCR < 0.2): 41.9 ms versus 53.6 ms. This clearly demonstrates that proteinuria further amplifies cardiovascular disease risk in CKD patients.

5. Disease-Specific Circadian Patterns of HRV

The study analyzed how diurnal HRV variation patterns differ by sex, disease status, and medication use.

5.1. Sex Differences in HRV Amplitude

Cosine modeling showed that male participants exhibited higher SDNN amplitude over 24 hours compared to female participants (p < 0.0001). This is consistent with previously known physiological phenomena. In multivariable linear regression, females had SDNN amplitude 2.23 ms lower than males (p = 0.005).

Figure 2: Disease-specific cyclical patterns in heart rate variability. Estimated SDNN curves showing time-specific differences by sex, proteinuria status, diabetic status, cardiovascular disease history, and beta blocker use.

5.2. Association Between Diabetes and HRV Amplitude

Diabetic patients showed a tendency toward 1.45 ms lower SDNN amplitude compared to non-diabetic patients (p = 0.074). This suggests that diabetes affects not only overall HRV variability but also diurnal patterns.

5.3. Factors Affecting HRV Acrophase

In the SDNN acrophase analysis (the time at which the rhythm reaches its peak), BMI and beta blocker use were significantly associated with acrophase shifts.

• BMI increase: Each 1-unit BMI increase was associated with a 0.11-hour (6.6-minute) advance in acrophase (p = 0.007)
• Beta blocker use: Patients on beta blockers showed a 1.36-hour delay in acrophase (p = 0.018)

Diabetic patients showed a tendency toward a 1-hour delay in acrophase (p = 0.06), though this was not statistically significant. These acrophase shifts suggest changes in cardiac autonomic status and may represent rhythm-based cardiovascular disease risk factors.

Spline analysis revealed additional patterns:

• Sex-related differences in curve shape supporting acrophase changes were distinct (p < 0.0001).
• Participants with high uPCR or diabetes showed overall lower SDNN levels than non-diabetic participants, but similar acrophases.
• The curve shape for participants with cardiovascular disease history was significantly blunted compared to those without (p = 0.01). Notably, SDNN levels decreased during nighttime hours.
• SDNN hour-by-hour curve shapes also differed significantly by beta blocker use (p = 0.0005). Patients on beta blockers tended to show increased SDNN, interpreted as beta blockers increasing parasympathetic activity and thereby raising SDNN.

6. Conclusion and Future Outlook

This study established and demonstrated the importance of the largest SDNN-based HRV analysis dataset from wearable devices in a chronic kidney disease (CKD) patient cohort. We confirmed disease-specific HRV regulation by sex, disease (particularly diabetes and HbA1C levels), and kidney damage (uPCR).

• Diabetes and high proteinuria are significantly associated with SDNN reduction, which may be linked to increased cardiovascular disease risk.
• Time-specific SDNN variability differed by sex, disease status, and use of treatments such as beta blockers. Notably, beta blocker use was associated with increased SDNN through enhanced parasympathetic activity.
• Diabetic patients had low SDNN levels throughout the day, while patients with cardiovascular disease history showed decreased SDNN levels primarily during nighttime hours. These findings may represent important clues for time-specific cardiovascular disease risk.

This study laid an important foundation for exploring the predictive power of HRV metrics for future cardiovascular events and other clinical outcomes in CKD patients. It established a new paradigm for monitoring time-dependent cardiovascular health in CKD patients using wearable digital health technology, providing an essential platform for developing clinically actionable digital biomarkers to improve cardiovascular disease management in CKD patients.

Of course, this is a correlational study, and the short follow-up period for some participants limits the ability to identify associations between HRV parameters and clinical outcomes. Despite these limitations, the study clearly demonstrates the potential of wearable devices for cardiovascular health monitoring in CKD patients and points the way for future research.