Dynamic Sleep-Derived Heart Rate and Heart Rate Variability Features Associated with Glucose Metabolism Status: An Exploratory Feature-Selection Study Using Consumer Wearables.
Li L, Syed Taha S, et al. • Sensors (Basel, Switzerland) • 2026
Dynamic features capturing overnight trends and variability patterns of heart rate and HRV during sleep—derived from a consumer Fitbit device—showed stronger associations with nocturnal mean glucose than conventional static mean values, with two dynamic HRV features differing significantly between lower- and higher-glycemic-risk groups.
Key Findings
Results
Dynamic sleep-derived HR/HRV features showed stronger associations with nocturnal mean glucose than conventional static mean values.
Study analyzed 189 nights from 18 participants using a consumer wrist-worn Fitbit device
From 28 candidate HR/HRV variables, Elastic Net regression (α=0.5) was applied to identify features associated with nocturnal mean glucose
14 features retained non-zero coefficients after Elastic Net regularization
Dynamic features capturing overnight trends and variability patterns outperformed conventional static mean values in association with glucose
Results
The nocturnal trends of within-window standard deviation and variance of ln(RMSSD) emerged as prominent candidate features associated with glycemic status.
RMSSD (root mean square of successive differences between consecutive RR intervals) was estimated from PPG-derived inter-beat intervals via Fitbit
Both the overnight trend of within-window standard deviation of ln(RMSSD) and variance of ln(RMSSD) were among the most notable features retained
HR variability indices also appeared alongside these HRV features as prominent candidates
These dynamic features capture overnight trends rather than single-point measurements
Results
Two dynamic HRV features differed significantly between lower- and higher-glycemic-risk groups with large effect sizes.
Both features reached statistical significance (both p<0.05)
Effect sizes were large for both features (Cohen's |d| > 1.1)
7 participants were in the higher-glycemic-risk group (estimated HbA1c ≥ 5.5%) and 11 were in the lower-glycemic-risk group (estimated HbA1c < 5.5%)
Between-group comparisons were conducted independently from the Elastic Net feature selection
Results
The lower-glycemic-risk group exhibited decreasing overnight trends in HRV variability, while the higher-glycemic-risk group showed increasing variability trends.
The lower-glycemic-risk group showed 'decreasing overnight trends in HRV variability, consistent with progressive autonomic stabilization during sleep'
The higher-glycemic-risk group showed 'increasing variability trends, suggestive of persistent autonomic instability'
These directional patterns are described as 'consistent with prior evidence linking autonomic dysfunction to impaired glucose metabolism'
The authors characterize these patterns as physiologically plausible
Background
Impaired glucose metabolism is associated with dysregulation of the autonomic nervous system, providing a physiological rationale for using sleep HR/HRV as a metabolic marker.
Impaired glucose metabolism is described as 'a known precursor to type 2 diabetes'
The autonomic nervous system dysregulation associated with impaired glucose metabolism forms the theoretical basis for the study
Consumer wearables provide 'continuous, non-invasive physiological monitoring' that 'may capture autonomic signatures related to metabolic status'
Glycemic status in this study was estimated using HbA1c thresholds (≥5.5% vs. <5.5%)
Discussion
The study used estimated HbA1c from a consumer device rather than laboratory-measured values, which the authors identify as a limitation requiring confirmatory investigation.
Glycated hemoglobin was 'estimated' rather than laboratory-measured
The authors explicitly characterize findings as 'hypothesis-generating'
Authors call for 'confirmatory investigation in larger, independent cohorts with laboratory-measured HbA1c'
Sample size was limited to 18 participants and 189 nights
The study is described as 'exploratory'
Conclusions
Consumer-grade wearable devices may have utility beyond activity tracking for continuous, real-world assessment of cardiometabolic health.
The study used a widely available consumer wrist-worn Fitbit device for all physiological measurements
Authors highlight 'the potential of widely available, consumer-grade wearable devices to move beyond activity tracking'
The approach supports 'continuous, real-world assessment of cardiometabolic health' and 'everyday health monitoring and preventive medicine'
Free-living conditions were used, meaning participants wore devices in their normal environments
What This Means
This research suggests that the way your heart rate and its beat-to-beat variability change throughout the night—captured by a standard Fitbit wristband—may carry information about a person's blood sugar regulation health. The study tracked 18 adults over 189 nights and found that people with higher estimated blood sugar levels (a sign of pre-diabetic risk) tended to show increasingly erratic heart rate variability patterns as the night progressed, while people with healthier blood sugar levels showed a calming, stabilizing pattern. These differences were statistically significant and had large effect sizes, suggesting they are meaningful and not just random noise.
The key insight is that these 'dynamic' features—meaning how heart rate variability trends and fluctuates over the whole night—were more informative than simple averages. This makes biological sense because the autonomic nervous system, which controls heart rate during sleep, is known to be disrupted in people with impaired glucose metabolism. When that system is dysregulated, it may show up as unstable, irregular patterns overnight rather than the steady wind-down seen in healthier individuals.
This research suggests that consumer smartwatches and fitness trackers could potentially do more than count steps—they might one day help identify people at risk for type 2 diabetes by analyzing their overnight heart patterns in everyday life. However, the authors are careful to note that this is a small, exploratory study using estimated (not lab-confirmed) blood sugar values, and the findings need to be verified in larger studies with clinical-grade measurements before any practical health applications could be considered.
Li L, Syed Taha S, Nishinaka Y, Tan Y, Ohtsu H, Lee S, et al.. (2026). Dynamic Sleep-Derived Heart Rate and Heart Rate Variability Features Associated with Glucose Metabolism Status: An Exploratory Feature-Selection Study Using Consumer Wearables.. Sensors (Basel, Switzerland). https://doi.org/10.3390/s26041118