Agreement between Garmin and Ultrahuman wearables and their association with continuous glucose monitoring: an n-of-1 study
- 1Institute for Anthropological Research
Abstract
Background: Consumer wearables and continuous glucose monitors (CGM) are increasingly worn together outside clinical settings. Few studies have examined whether two wearables worn at the same time agree with each other, and whether either tracks glucose metrics derived from CGM. Objective: In one adult under free-living conditions, we examined (1) how well sleep, heart rate, heart rate variability (HRV) and activity metrics from a Garmin smartwatch and an Ultrahuman ring agreed with each other, and (2) how these metrics related to glucose measured by CGM. Methods: One healthy adult with no diagnosed diabetes wore a Garmin Fenix 8 watch on the wrist, an Ultrahuman Ring AIR and an Abbott FreeStyle Libre CGM at the same time for 12 days (29 July to 9 August 2026; 1101 glucose readings). We assessed agreement between the devices with Pearson correlation and Bland-Altman analysis (mean bias and 95% limits of agreement, LoA) for daily metrics and for heart rate within the day, and with Cohen's kappa for sleep stages, minute by minute. We related the wearable metrics to mean glucose, glucose variability (coefficient of variation, CV%) and time in range (TIR, 3.9 to 10.0 mmol/L), reading by reading (over the whole record and separately for sleep and waking hours) and day by day. All correlations are descriptive. Because the Ultrahuman ring recorded no data for 22 hours on 3 August and its daytime heart rate behaved differently afterwards, heart-rate analyses are reported for the whole period and separately before and after that gap. The daily results were checked by leaving out one day at a time and by excluding the first 24 hours of sensor wear. Results: The devices agreed closely on total sleep duration (r=0.94; bias -0.02 h, 95% LoA -0.56 to 0.51 h). Daily steps correlated strongly (r=0.99 over the 11 days with data from both devices), but Garmin counted more steps on every day (mean 11,157 vs. 8,929 steps per day; bias +2228 steps). Heart rate agreed moderately during sleep (r=0.68). While awake, agreement was moderate before the gap (r=0.54) and absent after it (r=0.08), when Ultrahuman's daytime heart rate stayed near 70 bpm. The devices assigned the same sleep stage in 50.0% of 4525 minutes (kappa=0.28). Mean glucose was 5.13 mmol/L (SD 0.82) and TIR was 97.5%. Of all wearable signals, heart rate was most closely associated with glucose (Garmin r=0.43 over all readings), but the correlation fell to 0.34 within sleep and 0.21 within waking hours, and before the gap the two devices' waking correlations were alike (Garmin 0.25, Ultrahuman 0.24). Across the 12 days, correlations between night-time wearable metrics and glucose variability weakened markedly when the first 24 hours of sensor wear were excluded (absolute r fell from 0.59 or more to 0.34 or less), and TIR was 100% on 8 of the 12 days, which left it little room to vary. Conclusions: The two wearables agreed on how long the participant slept but not on how the night was staged, and their heart-rate agreement depended on whether the participant was asleep or awake and on the period of wear. Of all wearable signals, heart rate was most closely associated with glucose, mainly because both are lower at night and higher during the day. The associations across days rested largely on the first night of sensor wear. In this single participant, the devices behaved differently by device and by period of wear, so studies that combine different wearables, or relate them to CGM, should check for such differences before pooling data.
Related
- Innovations and Challenges in Smart-Textile Wearables for Postural Monitoring: A Scoping Review
- Association Between Continuous Glucose Monitoring - Derived Time in Range and Diabetic Kidney Disease: A Systematic Review and Meta-Analysis of current evidence
- Association of an almond-based low-carbohydrate diet with continuous glucose monitoring metrics in patients with type 2 diabetes mellitus: a randomized controlled trial.
- In-race continuous glucose monitoring in highly trained male distance runners: a context-dependent reversal of the glucose–performance association between training and competition
- Agreement between upper-arm and wrist cuff occlusion for the assessment of vascular reactivity by digital thermal monitoring
BibTeX
@misc{lovric2026agreement,
title = {{Agreement between Garmin and Ultrahuman wearables and their association with continuous glucose monitoring: an n-of-1 study}},
author = {M. Lovric and D. Strbad and M. A. Rasmussen},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.13.758757},
url = {https://www.biorxiv.org/content/10.64898/2026.10.13.758757v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Agreement between Garmin and Ultrahuman wearables and their association with continuous glucose monitoring: an n-of-1 study
AU - M. Lovric
AU - D. Strbad
AU - M. A. Rasmussen
PY - 2026
DA - 2026/10/13
PB - bioRxiv
DO - 10.64898/2026.10.13.758757
UR - https://www.biorxiv.org/content/10.64898/2026.10.13.758757v1
AB - Background: Consumer wearables and continuous glucose monitors (CGM) are increasingly worn together outside clinical settings. Few studies have examined whether two wearables worn at the same time agree with each other, and whether either tracks glucose metrics derived from CGM. Objective: In one adult under free-living conditions, we examined (1) how well sleep, heart rate, heart rate variability (HRV) and activity metrics from a Garmin smartwatch and an Ultrahuman ring agreed with each other, and (2) how these metrics related to glucose measured by CGM. Methods: One healthy adult with no diagnosed diabetes wore a Garmin Fenix 8 watch on the wrist, an Ultrahuman Ring AIR and an Abbott FreeStyle Libre CGM at the same time for 12 days (29 July to 9 August 2026; 1101 glucose readings). We assessed agreement between the devices with Pearson correlation and Bland-Altman analysis (mean bias and 95% limits of agreement, LoA) for daily metrics and for heart rate within the day, and with Cohen's kappa for sleep stages, minute by minute. We related the wearable metrics to mean glucose, glucose variability (coefficient of variation, CV%) and time in range (TIR, 3.9 to 10.0 mmol/L), reading by reading (over the whole record and separately for sleep and waking hours) and day by day. All correlations are descriptive. Because the Ultrahuman ring recorded no data for 22 hours on 3 August and its daytime heart rate behaved differently afterwards, heart-rate analyses are reported for the whole period and separately before and after that gap. The daily results were checked by leaving out one day at a time and by excluding the first 24 hours of sensor wear. Results: The devices agreed closely on total sleep duration (r=0.94; bias -0.02 h, 95% LoA -0.56 to 0.51 h). Daily steps correlated strongly (r=0.99 over the 11 days with data from both devices), but Garmin counted more steps on every day (mean 11,157 vs. 8,929 steps per day; bias +2228 steps). Heart rate agreed moderately during sleep (r=0.68). While awake, agreement was moderate before the gap (r=0.54) and absent after it (r=0.08), when Ultrahuman's daytime heart rate stayed near 70 bpm. The devices assigned the same sleep stage in 50.0% of 4525 minutes (kappa=0.28). Mean glucose was 5.13 mmol/L (SD 0.82) and TIR was 97.5%. Of all wearable signals, heart rate was most closely associated with glucose (Garmin r=0.43 over all readings), but the correlation fell to 0.34 within sleep and 0.21 within waking hours, and before the gap the two devices' waking correlations were alike (Garmin 0.25, Ultrahuman 0.24). Across the 12 days, correlations between night-time wearable metrics and glucose variability weakened markedly when the first 24 hours of sensor wear were excluded (absolute r fell from 0.59 or more to 0.34 or less), and TIR was 100% on 8 of the 12 days, which left it little room to vary. Conclusions: The two wearables agreed on how long the participant slept but not on how the night was staged, and their heart-rate agreement depended on whether the participant was asleep or awake and on the period of wear. Of all wearable signals, heart rate was most closely associated with glucose, mainly because both are lower at night and higher during the day. The associations across days rested largely on the first night of sensor wear. In this single participant, the devices behaved differently by device and by period of wear, so studies that combine different wearables, or relate them to CGM, should check for such differences before pooling data.
ER -