hrv.expert

Wearable HRV accuracy: the validation studies

Studies that compared a consumer device’s HRV with an ECG or a validated chest strap, with the number of people and the agreement figures copied from each paper.

Last updated

This page lists the evidence; what each brand measures and where to find it is in HRV on Oura, Whoop, Garmin and Apple Watch.

What “validated” means for HRV

A validation study records the same heartbeats twice, once with the device and once with a reference, usually an ECG, and then measures how close the two sets of numbers are. A device is validated for one measure, in one situation, against one reference, in the people who took part.

Studies report agreement with a handful of statistics:

  • Bias is the average difference between the device and the reference, such as 5 ms too low. A small bias can hide large errors, because readings that are too high and too low cancel out.
  • Limits of agreement are the range that holds 95% of the differences for single readings. This is the number to look at if you read one night or one morning at a time. Wide limits mean a single reading can be far off, so only a trend over several days tells you something.
  • MAPE, the mean absolute percentage error, is how far a typical reading is from the reference, as a percentage, in either direction. At an RMSSD of 50 ms, a MAPE of 6% means a typical reading is off by about 3 ms.
  • Concordance (CCC) and intraclass correlation (ICC) go up to 1, where 1 is perfect agreement and 0 is none (negative values are possible but rare). They reward readings that match the reference, not only readings that rise and fall with it.

Correlation is not agreement

Correlation (r) only says whether two sets of numbers go up and down together. A device that always showed twice the true value would have a perfect correlation and very poor agreement. A study that reports only r has not shown that the device gives the right number.

Three questions for every row

  • Which number was compared? Some studies took the HRV the app shows. Others took the device’s beat data and worked out HRV themselves, which tests the sensor, not the number you see.
  • When was it measured? Optical sensors do best when you are still. Agreement during sleep or quiet rest says little about readings during exercise or a busy day.
  • Who took part? Most studies tested a few dozen healthy adults and one model. Results can differ with age, heart rhythm, another model of the same brand, or a software update.

The studies

Each row is one device in one study, so a study that tested five devices has five rows. Figures are copied from the paper with the paper’s own signs. Papers differ on which device they subtract from which, so compare the size of a bias across studies, not its sign.

Device Measure When Compared with People (n) Agreement Study
Oura Ring Gen 4 Ring Tested: the device’s own HRV RMSSD Sleep ECG chest strap (Polar H10) 13 139 nights
  • CCC 0.99
  • MAPE 5.96%
 Dial et al., 2025 [1] Physiological Reports
Oura Ring Gen 3 Ring Tested: the device’s own HRV RMSSD Sleep ECG chest strap (Polar H10) 13 470 nights
  • CCC 0.97
  • MAPE 7.15%
 Dial et al., 2025 [1] Physiological Reports
WHOOP 4.0 Wrist Tested: the device’s own HRV RMSSD Sleep ECG chest strap (Polar H10) 13 289 nights
  • CCC 0.94
  • MAPE 8.17%
 Dial et al., 2025 [1] Physiological Reports
Garmin Fenix 6 Wrist Tested: the device’s own HRV RMSSD Sleep ECG chest strap (Polar H10) 13 150 nights
  • CCC 0.87
  • MAPE 10.52%
 Dial et al., 2025 [1] Physiological Reports
Polar Grit X Pro Wrist Tested: the device’s own HRV RMSSD Sleep ECG chest strap (Polar H10) 13 206 nights
  • CCC 0.82
  • MAPE 16.32%
 Dial et al., 2025 [1] Physiological Reports
Apple Watch Series 6 Wrist Tested: the device’s own HRV RMSSD Sleep Sleep-lab ECG 53
  • Bias −9.6 ms
  • Limits of agreement ±55.2 ms
  • ICC 0.67
Values read through a third-party app. 		Miller et al., 2022 [2] Sensors
Garmin Forerunner 245 Music Wrist Tested: the device’s own HRV RMSSD Rest Sleep-lab ECG 53
  • Bias −22.4 ms
  • Limits of agreement ±92.0 ms
  • ICC 0.24
 Miller et al., 2022 [2] Sensors
Oura Ring Gen 2 Ring Tested: the device’s own HRV RMSSD Sleep Sleep-lab ECG 53
  • Bias −10.2 ms
  • Limits of agreement ±77.2 ms
  • ICC 0.63
 Miller et al., 2022 [2] Sensors
WHOOP 3.0 Wrist Tested: beat data analysed by the researchers RMSSD Sleep Sleep-lab ECG 53
  • Bias −4.5 ms
  • Limits of agreement ±7.6 ms
  • ICC 0.99
Beat data supplied by the maker. 		Miller et al., 2022 [2] Sensors
Oura Ring Ring Tested: not reported in the abstract HRV, formula not reported in the abstract Sleep ECG, type not reported in the abstract 49
  • r² 0.980 (nightly averages)
  • Bias −1.2 ms (nightly averages)
  • r² 0.765 (5-minute segments)
 Kinnunen et al., 2020 [3] Physiological Measurement
Oura Ring Ring Tested: the device’s own HRV RMSSD Sleep Portable ECG (Shimmer3) 35
  • r 0.96 (nightly averages)
  • Bias −15.88 ms (nightly averages)
  • Limits of agreement −33.29 to 1.53 ms (nightly averages)
 Cao et al., 2022 [4] Journal of Medical Internet Research
RingConn Gen 2 Ring Tested: beat data analysed by the researchers RMSSD Rest ECG chest strap (Polar H10) 64
  • r 0.9564
  • CCC 0.9438
  • ICC 0.9449
  • Bias 3.237 ms
  • Limits of agreement −8.591 to 15.065 ms
  • MAE 4.620 ms
Beat data supplied by the maker. 		Tao et al., 2026 [5] BMC Sports Science, Medicine and Rehabilitation
WHOOP 2.0 Wrist Tested: beat data analysed by the researchers lnRMSSD Sleep Sleep-lab ECG 6 15 sleep periods
  • Bias 1.66% (best of the filters tested)
  • Limits of agreement ±5.93% (best of the filters tested)
Fewer than ten people. 		Bellenger et al., 2021 [6] Sensors
Samsung Gear Sport Wrist Tested: beat data analysed by the researchers RMSSD Sleep Portable ECG (Shimmer3) 28
  • r 0.778 (5-minute segments)
  • Bias −18.24 ms (5-minute segments)
  • Limits of agreement −68.49 to 32.01 ms (5-minute segments)
 Sarhaddi et al., 2022 [7] PLOS ONE
Samsung Gear Sport Wrist Tested: beat data analysed by the researchers RMSSD Daily life Portable ECG (Shimmer3) 28
  • r 0.251 (5-minute segments)
  • Bias 6.15 ms (5-minute segments)
  • Limits of agreement −79.59 to 91.89 ms (5-minute segments)
 Sarhaddi et al., 2022 [7] PLOS ONE
Fitbit Versa 4 Wrist Tested: the device’s own HRV RMSSD Sleep Portable ECG (Firstbeat Bodyguard 3) 8 26 nights
  • r 0.944 (nightly averages)
  • Median 21.660 ms on the device, 20.558 ms on the ECG
Pilot study.Fewer than ten people.Mostly older adults (average age 64) waiting for shoulder surgery. 		Nasr et al., 2025 [8] Digital Health
Apple Watch Series 9 and Ultra 2 Wrist Tested: the device’s own HRV SDNN Rest ECG chest strap (Polar H10) 39 316 readings
  • Bias −8.31 ms
  • MAPE 28.88%
  • MAE 20.46 ms
 O'Grady et al., 2024 [9] Sensors
Apple Watch Wrist Tested: beat data analysed by the researchers Beat intervals (RR) Rest ECG chest strap (Polar H7) 20
  • CCC 0.989 (at rest)
  • Beats missed 10.98% (at rest)
Gaps of 3 to 10 seconds with no beats detected. 		Hernando et al., 2018 [10] Sensors
Garmin vivoactive 4 Wrist Tested: beat data analysed by the researchers RMSSD Rest Lab ECG 263
  • CCC 0.6617
Most people had had a heart attack or a stroke. 		Theurl et al., 2023 [11] European Heart Journal – Digital Health
Polar OH1 Upper arm Tested: beat data analysed by the researchers RMSSD Rest ECG chest strap (Polar H10) 31
  • ICC 0.955 (lying down)
  • Bias −3.21 ms (lying down)
  • ICC 0.834 (sitting)
  • Bias −8.05 ms (sitting)
 Coste et al., 2025 [12] Sensors
Oura Ring Gen 2 Ring Tested: the device’s own HRV RMSSD Rest Lab ECG (Shimmer3) 5 25 trials
  • MAPE 6.84%
  • CCC 0.91
  • Bias −2.33 ms
Fewer than ten people. 		Stone et al., 2021 [13] Frontiers in Sports and Active Living
Polar H10 with HRV4Training Chest strap Tested: the device’s own HRV RMSSD Rest Lab ECG (Shimmer3) 5 21 trials
  • MAPE 4.10%
  • CCC 0.98
  • Bias −0.74 ms
Fewer than ten people. 		Stone et al., 2021 [13] Frontiers in Sports and Active Living
iPhone 8 camera with CameraHRV Phone camera Tested: the device’s own HRV RMSSD Rest Lab ECG (Shimmer3) 5 25 trials
  • MAPE 112.36%
  • CCC 0.04
  • Bias 78.33 ms
Fewer than ten people. 		Stone et al., 2021 [13] Frontiers in Sports and Active Living
Smartphone camera with CameraHRV 5.0.9 Phone camera Tested: the device’s own HRV RMSSD Rest Lab ECG 37
  • MAPE 17.49%
  • Bias −3.0 ms
  • Limits of agreement −29.7 to 23.7 ms
 Johansson et al., 2026 [14] Frontiers in Physiology
Polar H10 Chest strap Tested: beat data analysed by the researchers RMSSD Rest Lab ECG 37
  • MAPE 2.16%
  • Bias 0.4 ms
  • Limits of agreement −3.3 to 4.0 ms
 Johansson et al., 2026 [14] Frontiers in Physiology
Smartphone camera and Polar H7 Phone camera Tested: not reported in the abstract RMSSD Rest ECG, type not reported in the abstract 29
  • r 1.00
  • Typical error 6.35% (both devices and breathing patterns together)
The figures pool the phone camera and the chest strap. 		Plews et al., 2017 [15] International Journal of Sports Physiology and Performance
Smartphone camera (Happitech software) Phone camera Tested: not reported in the abstract lnRMSSD Rest Lab ECG (NeXus) 57
  • ICC > 0.90
 van Dijk et al., 2023 [16] Psychosomatic Medicine
Polar H10 Chest strap Tested: beat data analysed by the researchers RMSSD Rest Lab ECG 43
  • CCC ≥ 0.99
  • MAPE < 1%
  • Limits of agreement −1.5 to 1.7 ms (widest of all measures)
 Blalock et al., 2026 [17] Autonomic Neuroscience
Polar H10 Chest strap Tested: beat data analysed by the researchers Beat intervals (RR) Exercise ECG, every beat checked by eye 10
  • Beats detected correctly 99.6%
  • Beats detected correctly 99.4% (at high intensity)
 Gilgen-Ammann et al., 2019 [18] European Journal of Applied Physiology
Polar H10 Chest strap Tested: beat data analysed by the researchers Beat intervals (RR) Exercise Lab ECG 25
  • r 1.00 (at rest)
  • r 1.00 (during exercise)
 Schaffarczyk et al., 2022 [19] Sensors
Movesense Medical Chest strap Tested: beat data analysed by the researchers RMSSD Exercise Lab ECG 21
  • r 0.99 (at rest)
  • r 0.98 (during exercise)
 Rogers et al., 2022 [20] Sensors

Sources

  1. Dial MB, Hollander ME, Vatne EA, et al. (2025). Validation of nocturnal resting heart rate and heart rate variability in consumer wearables. Physiological Reports. doi:10.14814/phy2.70527 Checked in the full text
  2. Miller DJ, Sargent C, Roach GD (2022). A Validation of Six Wearable Devices for Estimating Sleep, Heart Rate and Heart Rate Variability in Healthy Adults. Sensors. doi:10.3390/s22166317 Checked in the full text
  3. Kinnunen H, Rantanen A, Kenttä T, Koskimäki H (2020). Feasible assessment of recovery and cardiovascular health: accuracy of nocturnal HR and HRV assessed via ring PPG in comparison to medical grade ECG. Physiological Measurement. doi:10.1088/1361-6579/ab840a Checked in the abstract
  4. Cao R, Azimi I, Sarhaddi F, et al. (2022). Accuracy Assessment of Oura Ring Nocturnal Heart Rate and Heart Rate Variability in Comparison With Electrocardiography in Time and Frequency Domains: Comprehensive Analysis. Journal of Medical Internet Research. doi:10.2196/27487 Checked in the full text
  5. Tao L, Sun H, Li Y, et al. (2026). Method comparison of a ring-based PPG device and a chest-strap monitor for heart rate and heart rate variability in shooting athletes. BMC Sports Science, Medicine and Rehabilitation. doi:10.1186/s13102-026-01960-x Checked in the full text
  6. Bellenger CR, Miller DJ, Halson SL, et al. (2021). Wrist-Based Photoplethysmography Assessment of Heart Rate and Heart Rate Variability: Validation of WHOOP. Sensors. doi:10.3390/s21103571 Checked in the full text
  7. Sarhaddi F, Kazemi K, Azimi I, et al. (2022). A comprehensive accuracy assessment of Samsung smartwatch heart rate and heart rate variability. PLOS ONE. doi:10.1371/journal.pone.0268361 Checked in the full text
  8. Nasr A, Behbehani K, Wang J, et al. (2025). Agreement between Fitbit Versa 4 and a reference standard on time-domain nocturnal heart rate variability: A pilot study. Digital Health. doi:10.1177/20552076251394960 Checked in the full text
  9. O'Grady B, Lambe R, Baldwin M, et al. (2024). The Validity of Apple Watch Series 9 and Ultra 2 for Serial Measurements of Heart Rate Variability and Resting Heart Rate. Sensors. doi:10.3390/s24196220 Checked in the full text
  10. Hernando D, Roca S, Sancho J, et al. (2018). Validation of the Apple Watch for Heart Rate Variability Measurements during Relax and Mental Stress in Healthy Subjects. Sensors. doi:10.3390/s18082619 Checked in the full text
  11. Theurl F, Schreinlechner M, Sappler N, et al. (2023). Smartwatch-derived heart rate variability: a head-to-head comparison with the gold standard in cardiovascular disease. European Heart Journal – Digital Health. doi:10.1093/ehjdh/ztad022 Checked in the full text
  12. Coste A, Millour G, Hausswirth C (2025). A Comparative Study Between ECG- and PPG-Based Heart Rate Sensors for Heart Rate Variability Measurements: Influence of Body Position, Duration, Sex, and Age. Sensors. doi:10.3390/s25185745 Checked in the full text
  13. Stone JD, Ulman HK, Tran K, et al. (2021). Assessing the Accuracy of Popular Commercial Technologies That Measure Resting Heart Rate and Heart Rate Variability. Frontiers in Sports and Active Living. doi:10.3389/fspor.2021.585870 Checked in the full text
  14. Johansson H, Adderley E, Clarke S, et al. (2026). An observational study of the reliability and concurrent validity of heart rate variability devices in athletes. Frontiers in Physiology. doi:10.3389/fphys.2025.1707318 Checked in the full text
  15. Plews DJ, Scott B, Altini M, et al. (2017). Comparison of Heart-Rate-Variability Recording With Smartphone Photoplethysmography, Polar H7 Chest Strap, and Electrocardiography. International Journal of Sports Physiology and Performance. doi:10.1123/ijspp.2016-0668 Checked in the abstract
  16. van Dijk W, Huizink AC, Oosterman M, et al. (2023). Validation of Photoplethysmography Using a Mobile Phone Application for the Assessment of Heart Rate Variability in the Context of Heart Rate Variability–Biofeedback. Psychosomatic Medicine. doi:10.1097/PSY.0000000000001236 Checked in the abstract
  17. Blalock ACE, Riemann BL, Flatt AA (2026). Validity of the Polar H10 for heart rate variability and cardiac autonomic reflex tests. Autonomic Neuroscience. doi:10.1016/j.autneu.2026.103447 Checked in the abstract
  18. Gilgen-Ammann R, Schweizer T, Wyss T (2019). RR interval signal quality of a heart rate monitor and an ECG Holter at rest and during exercise. European Journal of Applied Physiology. doi:10.1007/s00421-019-04142-5 Checked in the abstract
  19. Schaffarczyk M, Rogers B, Reer R, Gronwald T (2022). Validity of the Polar H10 Sensor for Heart Rate Variability Analysis during Resting State and Incremental Exercise in Recreational Men and Women. Sensors. doi:10.3390/s22176536 Checked in the full text
  20. Rogers B, Schaffarczyk M, Clauß M, et al. (2022). The Movesense Medical Sensor Chest Belt Device as Single Channel ECG for RR Interval Detection and HRV Analysis during Resting State and Incremental Exercise: A Cross-Sectional Validation Study. Sensors. doi:10.3390/s22052032 Checked in the full text

How we check a study

Every row was last checked against its paper on 20 Oct 2026. We resolved each DOI and matched the title, authors, journal and year with the publisher’s record. The figures come from the abstract or, where the paper is open access, from the full text, and the list of sources says which. Nothing is estimated, converted or averaged across studies.

The table leaves out conference papers, studies of heart rate alone, and studies whose figures we could not read in the paper itself. A device missing from the table has not been shown to be inaccurate. We may not yet have found or added a study we could check.

Suggest a study

Know a peer-reviewed study we missed, or one on a newer model? Send us its DOI through the contact form. We add a row only after reading the paper ourselves.