Abstract
Little is known about the prognostic value of out-of-clinic biometric monitoring of cardiovascular function in chronic kidney disease (CKD). Using real-world sampling strategies, a mean (±SD) of 50.3±9.3 hours of ECG recordings from wearable BioPatch ECG devices was collected in a cohort consisting of 458 participants from seven Chronic Renal Insufficiency Cohort (CRIC) centers. The presence of diabetes was associated with a 7.4 ms lower Standard Deviation of NN Intervals (SDNN) compared to non-diabetic participants (p=0.001). Multivariable linear regression revealed that participants without proteinuria (uPCR<0.2) had a 5.15 ms higher SDNN compared to participants with proteinuria (uPCR≥0.2, p=0.027). Cosinor modeling suggested differences in SDNN acrophase quartiles for diabetes (p=0.02), history of cardiovascular disease (p=0.003), eGFR (p=0.04), systolic blood pressure (p=0.04), and beta blocker use (p=0.0002). In the spline analysis, the SDNN curve differed between participants with and without cardiovascular disease (p=0.0005). This study assembled the largest dataset to date of SDNN as an index for heart rate variability from wearable digital health technology in the CRIC. The study demonstrates that several clinical and demographic factors are associated with SDNN in participants with CKD. This sets the stage to determine the predictiveness of time-specific HRV metrics for future clinical outcomes.