This paper is in the following e-collection/theme issue:

Wearable Devices and Sensors (912) mHealth for Data Collection and Research (988) Ubiquitous Health (uHealth) (127) Research Instruments, Questionnaires, and Tools (597) Novel Sensors and Data Acquisition Methods in Cardiology (203) Fitness Trackers and Smart Pedometers/Accelerometers (574) Cardiac Self-Management (109) e-Cardiology and Telemonitoring in Cardiology (175)

Published on in Vol 13 (2025)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/54871, first published .
Reliability and Accuracy of the Fitbit Charge 4 Photoplethysmography Heart Rate Sensor in Ecological Conditions: Validation Study

Reliability and Accuracy of the Fitbit Charge 4 Photoplethysmography Heart Rate Sensor in Ecological Conditions: Validation Study

Reliability and Accuracy of the Fitbit Charge 4 Photoplethysmography Heart Rate Sensor in Ecological Conditions: Validation Study

Authors of this article:

Maxime Ceugniez1 Author Orcid Image ;   Hervé Devanne1 Author Orcid Image ;   Eric Hermand1, 2 Author Orcid Image
Article Authors Cited by Tweetations Metrics

    1ULR 7369 - URePSSS - Unité de Recherche Pluridisciplinaire Sport Santé Société, Univ. Littoral Côte d’Opale, Univ. Lille, Univ. Artois, 189b, Avenue Maurice Schumann, Centre Universitaire des Darses, Dunkerque, France

    2UMR INSERM U1272 Hypoxie & Poumon, Département STAPS, Université Sorbonne Paris Nord, Bobigny, France

    Corresponding Author:

    Eric Hermand, PhD


    Background: Wrist-worn photoplethysmography (PPG) sensors allow for continuous heart rate (HR) measurement without the inconveniences of wearing a chest belt. Although green light PPG technology reduces HR measurement motion artifacts, only a limited number of studies have investigated the reliability and accuracy of wearables in non–laboratory-controlled conditions with actual specific and various physical activity movements.

    Objective: The purpose of this study was to (1) assess the reliability and accuracy of the PPG-based HR sensor of the Fitbit Charge 4 (FC4) in ecological conditions and (2) quantify the potential variability caused by the nature of activities.

    Methods: We collected HR data from participants who performed badminton, tennis, orienteering running, running, cycling, and soccer while simultaneously wearing the FC4 and the Polar H10 chest belt (criterion sensor). Skin tone was assessed with the Fitzpatrick Skin Scale. Once data from the FC4 and criterion data were synchronized, accuracy and reliability analyses were performed, using intraclass correlation coefficients (ICCs), Lin concordance correlation coefficients (CCCs), mean absolute percentage errors (MAPEs), and Bland-Altman tests. A linear univariate model was also used to evaluate the effect of skin tone on bias. All analyses were stratified by activity and pooled activity types (racket sports and running sports).

    Results: A total of 77.5 hours of HR recordings from 26 participants (age: mean 21.1, SD 5.8 years) were analyzed. The highest reliability was found for running sports, with ICCs and CCCs of 0.90 and 0.99 for running and 0.80 and 0.93 for orienteering running, respectively, whereas the ICCs and CCCs were 0.37 and 0.78, 0.42 and 0.88, 0.65 and 0.97, and 0.49 and 0.81 for badminton, tennis, cycling, and soccer, respectively. We found the highest accuracy for running (bias: 0.1 beats per minute [bpm]; MAPE 1.2%, SD 4.6%) and the lowest for badminton (bias: −16.5 bpm; MAPE 16.2%, SD 14.4%) and soccer (bias: −16.5 bpm; MAPE 17.5%, SD 20.8%). Limit of agreement (LOA) width and artifact rate followed the same trend. No effect of skin tone was observed on bias.

    Conclusions: LOA width, bias, and MAPE results found for racket sports and soccer suggest a high sensitivity to motion artifacts for activities that involve “sharp” and random arm movements. In this study, we did not measure arm motion, which limits our results. However, whereas individuals might benefit from using the FC4 for casual training in aerobic sports, we cannot recommend the use of the FC4 for specific purposes requiring high reliability and accuracy, such as research purposes.

    JMIR Mhealth Uhealth 2025;13:e54871

    doi:10.2196/54871

    Keywords

    photoplethysmography; physical activity; ecological conditions; accuracy; reliability; Fitbit Charge 4; Fitbit; exercise; ecological; wrist-worn device; device; sensor; wearables; usefulness; variability; sensitivity; heart rate; heart rate sensor 


    Over the recent years, connected bracelet and watch sales have been regularly increasing [1]. These tools allow individuals to monitor various active life parameters, such as time of activity, sleep quality, step numbers, and energy expenditure, among others. Most of them are computed through algorithms that use accelerometry and heart rate (HR) data. Although these new devices are used for training by both recreational athletes and elite athletes, chest belts remain the gold standard, especially among high-level athletes, for measuring HR, as they are based on R peak detection from the QRS electrocardiogram (ECG) complex [2,3]. In opposition to chest belts, other wearables perform additional measurements (eg, accelerometry and positioning measurements via a GPS), which, when combined with HR data and algorithm processing, provide data on other parameters, such as energy expenditure and quality of sleep. Moreover, wrist-worn devices could reduce the tolerance and acceptability issues observed with chest belts [4]. Wrist-worn devices usually estimate continuous HR through the photoplethysmography (PPG) technique, which was first used in the late 1930s [5]. PPG involves measuring light absorption through tissues of interest [6]; red and infrared lights are emitted by an LED through the skin, and a photoreceptor captures the remaining emissions after tissue absorption [7]. However, even if the concept remains similar, connected watches usually come with a green light PPG sensor for its ability to reduce motion artifacts, contrary to the red ones commonly used in the medical field for blood oxygen saturation evaluation [8,9]. The reason for this is that the deeper the light penetrates the tissue (eg, red wavelength), the more the pulse wave is affected by limb movements [10,11]. As light penetration depends on light wavelength, the shorter wavelength of green light provides less information from deeper nonpulsatile tissues [9,12]. Considering this, green light is less prone to motion artifacts during normal daily life [13-15]. In the case of HR monitoring via watches or bracelets during activities, signal accuracy and reliability may vary according to numerous factors. Among them, gear placement on skin, strap tightening (which induces skin compression), skin tone, and activity type and intensity can affect HR recording [15-21]. PPG HR sensor accuracy has been investigated during physical activity across a spectrum of intensities. However, researchers tend to measure PPG HR sensor accuracy with treadmill running or cycling ergometers in laboratory-controlled conditions [17,22-24]. To our knowledge, only a few evaluations of connected device accuracy (eg, accuracy of a smartwatch, as evaluated in this paper) were performed in ecological conditions across different physical activity types. As this type of device is meant to be used in non–laboratory-controlled conditions or free-living conditions, this study aimed to evaluate the accuracy and validity of the PPG HR data from the Fitbit Charge 4 (FC4; Fitbit LLC) across multiple physical activity types. Therefore, the objectives of this study were to (1) assess the accuracy and reliability of FC4 HR measurement in ecological conditions and (2) quantify the potential impact of activity type on accuracy.


    Participants

    A total of 26 healthy young adults from the Sport Sciences University of Calais, France, who were practicing physical activities on a weekly basis, volunteered and were included in this study, which was advertised on the university campus and via social networks. No inclusion or exclusion criteria were used.

    A minimum HR sample size was calculated with G*Power (version 3.1.9.6) [25] by using a significance level (α) of 5%, a statistical power of 1 – β = 80%, and an effect size of 0.075 (computed from the expected HR mean and SD). The number of necessary HR samples was below 6000, representing 100 minutes of recording (sample rate=1 s−1). Measurements were performed during participants’ regular training sessions for soccer, badminton, orienteering running, basketball, tennis, and road biking.

    Ethical Considerations

    This study was approved by the National Commission for Data Protection and Liberties (CNIL-France; registration number: 2224247). All participants gave their written informed consent, with the possibility to opt out of the protocol at any point. Data collected throughout the protocol were deidentified for privacy and confidentiality reasons. Finally, although participants could not be financially compensated for their participation, which did not impact their usual routine, each of them personally received an individualized analysis of their HR data, so that they could receive information about cardiac demand during various phases of their training sessions (intensity levels, duration, and cardiac work zone) and adjust their sessions’ contents if needed.

    Data Collection

    PPG HR signals from the FC4 were compared to those from the Polar H10 thoracic belt (Polar Electro Oy), which was used as the criterion sensor [26]. The assessment of skin tone was performed with the Fitzpatrick Skin Scale, which ranges from 1 (lightest tone) to 6 (darkest tone) [27].

    Participants were asked to wear both sensors simultaneously at each session. The FC4 was placed on the wrist of the nondominant arm (ie, around 2 cm away [proximal] from the ulnar styloid process), whereas the Polar H10 thoracic belt was placed under the thorax (ie, on the xyphoid process) and paired with Polar V800 wristwatches for HR recording. Each device was placed firmly against the skin, as recommended by the manufacturer’s instructions.

    Data Extraction and Analysis

    Data were extracted through both companies’ web services (ie, the Fitbit app [28] and Polar Flow website [29] for the FC4 and Polar H10, respectively). A MATLAB script (The MathWorks Inc) was then used to synchronize the two devices’ HR measurements. Record alignments were performed by using a least square method to minimize squared deviation between FC4 and Polar H10 records, and they were smoothed over a 10-second window to calibrate the sample rate from both devices, as previously described [18,30]. Data normality was verified by a Kolmogorov-Smirnov test.

    FC4 artifact data were defined as values that deviated from the criterion data by 20 beats per minute (bpm). Bland-Altman tests were performed on smoothed data to assess the accuracy of FC4 HR data by participant and by activity [31]. Means (bias) and SDs of the differences between the FC4 and H10 values were used to evaluate upper and lower limits of agreement (LOAs), per the following formula: upper/lower LOA = bias ± 1.96 × SD. Mean absolute error (MAE) and mean absolute percentage error (MAPE) were calculated to quantify mean differences between FC4 and Polar H10 HR data. Two tests were performed to evaluate the reliability of the FC4: (1) 2-way random intraclass correlation coefficients (ICCs) with an absolute consistency type were calculated and interpreted according to current guidelines (ICC<0.5: poor; 0.5<ICC<0.75: moderate; 0.75<ICC<0.90: good; ICC>0.90: excellent reliability) [32], and (2) a computation of Lin concordance correlation coefficients (CCCs) was performed, interpreted following McBride’s [33] recommendations (CCC<0.90: poor; 0.90<CCC<0.95: moderate; 0.95<CCC<0.99: very good; CCC>0.99: almost perfect strength of agreement).

    All statistical analyses were stratified by activity type, and a Kruskal-Wallis test was performed to compare activity bias, with activity types as independent groups. Mann-Whitney U tests were implemented to compare bias from 0 among activity types (independent groups). Further, a linear univariate model was used to estimate the effect of skin tone on bias while controlling the impact of activity type. Statistics were performed using IBM SPSS statistics 25 software (IBM Corp).


    Participants’ Characteristics

    A total of 26 young adults (11 women and 15 men) were included in this study. Their characteristics are compiled in Table 1. In total, 77.5 hours of practice were recorded, distributed across 55 sessions (Table 2).

    Table 1. Participants’ characteristics.
    Male (n=15), mean (SD)Female (n=11), mean (SD)All participants (N=26), mean (SD)
    Age (y)21.2 (7.0)20.8 (3.7)21.1 (5.8)
    Weight (kg)75.8 (9.6)57.6 (8.9)68.1 (12.9)
    Height (cm)183 (6)166 (8)176 (11)
    BMI (kg/m2)22.6 (1.8)20.9 (1.7)21.9 (1.9)
    Skin tone (Fitzpatrick Skin Scale score)2.9 (0.6)2.8 (0.6)2.9 (0.6)
    Table 2. Descriptive data of recorded sessions.
    ActivitySessions, nRecorded time, hParticipants, n
    Racket sports
    Badminton1015.077
    Tennis34.902
    Total1319.969
    Running sports
    Orienteering running57.025
    Run1114.093
    Total1621.108
    Other sports
    Bike1318.392
    Soccer1318.0112

    Accuracy and Artifact Percentage

    Biases, LOAs, and artifact percentages are shown in Table 3.

    Table 3. Bland-Altman analyses, mean absolute error (MAE), and mean absolute percentage error (MAPE) by activity.
    ActivityBland-Altman analyses, bpmaArtifact ratios, %bpm, MAE (SD)bpm, MAPE (SD)
    BiasUpper LOAb; lower LOALOA width
    Racket sports
    Badminton−16.535.2; −68.2103.539.321.7 (22.3)16.2 (14.4)
    Tennis−6.224.8; −37.262.022.712.8 (11.2)8.9 (7.4)
    Total−14.034.3; −62.396.635.219.5 (20.5)14.4 (13.4)
    Running sports
    Orienteering running−8.626.0; −43.369.417.011.7 (15.9)9.5 (10.4)
    Run0.110.9; −10.721.62.51.7 (5.2)1.2 (4.6)
    Total−2.820.5; −26.146.67.35.0 (11.1)4.0 (8.1)
    Other sports
    Bike4.836.8; −27.364.118.110.4 (10.4)8.1 (11.0)
    Soccer−16.526.7; −59.686.335.519.2 (19.7)17.5 (20.8)

    abpm: beats per minute.

    bLOA: limit of agreement.

    Biases were different between each activity (P<.001), with all of them also being different from 0 (P<.001). The lowest bias values, MAEs, and MAPEs were found for running and cycling, and the highest ones were found for badminton and soccer. Furthermore, the narrowest LOA width was found for running, and the widest was found for badminton (Figure 1). We found similar results by grouping activities; the lowest bias, MAE, MAPE, and LOA width were found for running activities (running and orienteering running), and the largest ones were found for racket sports (badminton and tennis; Table 3).

    ‎
    Figure 1. Bland-Altman plots for (A) badminton and (B) running. Badminton exhibits the highest bias, and the bias for running is the closest to the origin. Each activity represents, respectively, 14 and 15.07 hours of recording. bpm: beats per minute; FC4: Fitbit Charge 4; HR: heart rate; LOA: limit of agreement; Polar: Polar H10.

    Reliability

    ICCs and CCCs are presented in Table 4. ICCs and CCCs indicated poor reliability (<0.50 and <0.90, respectively) for racket sports and soccer but excellent reliability for running overall (the total ICC and CCC for all running activities overall were >0.90 and >0.99, respectively). The largest percentages of artifact HR data were found for badminton and soccer, whereas running exhibited the lowest (Table 3). The results were similar when pooling activities; the highest rate of artifact HR data was found for racket sports, and the lowest was found for running sports (Table 3).

    Table 4. Intraclass correlation coefficients (ICCs) and Lin concordance correlation coefficients (CCCs) by single and pooled activities.
    ActivityICC (95% CI)CCC
    Racket sports
    Badminton0.365 (0.517-0.154)0.778
    Tennis0.421 (0.320-0.421)0.884
    Total0.435 (0.237-0.574)0.883
    Running sports
    Orienteering running0.801 (0.865-0.801)0.932
    Run0.900 (0.898-0.900)0.999
    Total0.926 (0.915-0.935)0.996
    Other sports
    Bike0.658 (0.702-0.603)0.971
    Soccer0.487 (0.158-0.487)0.809

    Effect of Skin Tone

    Mean Fitzpatrick Skin Scale scores are shown in Table 1. No overall interaction and no interaction in each activity were found between bias and skin tone, while being standardized by activity type.


    Main Results

    In this study, we evaluated the HR accuracy and reliability of the FC4 by comparing it to the Polar H10 chest belt (criterion sensor) in non–laboratory-controlled conditions. Our results showed negative biases for most activities (except running and cycling; Table 3), the presence of artifact data, and HR underevaluation by the FC4 (Figure 2). These results are similar to earlier findings that show the tendency of PPG wrist sensors to overestimate or underestimate HR [34-36]. ICCs and CCCs were fluctuant, mainly depending on the activity type. The FC4 shows good reliability for running activities, with almost perfect and moderate CCCs and excellent and good ICCs for running and orienteering running, respectively. Additionally, running was the activity with the lowest bias, the lowest MAPE, and the smallest LOA width. On the other hand, we found lower ICCs for badminton, soccer, and cycling, which also showed higher artifact ratios. Thus, we suggest that the excessive amount of arm movement in these activities could affect HR recording, as previously shown [37]. Badminton and tennis are characterized by “sharp” movements and rotations of the nondominant arm, whereas soccer can induce some instability of the sensor due to random arm and wrist actions, which can result in watches sliding over skin and the transient loss of HR signals (Figure 2). Furthermore, although cycling shows an overall good CCC (0.971), the ICC (0.658), MAPE (8.1%, SD 11.0%), and LOA width (64.03 bpm) indicate a lack of reliability during this activity. Since cycling remains a lower limb cyclic activity, there are little to no arm movements that may affect sensor placement. However, wrist position on handlebars (eg, during road cycling) and the contractility of wrist muscle flexors and extensors could alter vascular arteriovenous system detection and lower signal quality while enhancing compression forces [6,8,16,17]. ICCs and CCCs were calculated for pooled activities—running sports (running and orienteering running) and racket sports (tennis and badminton). These activities mostly rely on the same corporal pattern, and grouping them allowed us to equilibrate the time of practice between other activities. Our data showed no differences in ICC or CCC parameters, with those for running sports and racket sports showing excellent and poor reliability, respectively (ICC: 0.926 vs 0.435; CCC: 0.996 vs 0.883; Table 4).

    ‎
    Figure 2. Examples of synchronized heart rate signals (Polar H10: light grey; Fitbit Charge 4: black; A: orienteering running; B: soccer). Soccer heart rate data show recurrent sudden uncoupling between Fitbit Charge 4 and Polar H10 data, as well as heart rate underevaluations by the Fitbit Charge 4 throughout the recording session. bpm: beats per minute.

    Overall, artifacts and random movements could highly affect a sensor’s precision. Some studies tried to reduce motion artifacts by using novel techniques, such as accelerometry coupling or algorithm-based processing, but there is no consensus yet on which one should be used [38-42]. Even by varying numbers of diodes, lights colors, and algorithms, the technical aspect for measuring HR remains similar, suggesting that some FC4 characteristics could increase the presence of motion artifacts. For example, the materials used for the strap was slippery on skin, amplified by exercise-induced sudation. In addition, from a general standpoint, algorithms used by manufacturers can also affect the recordings, but we did not have access to these proprietary processing scripts.

    Limitations

    We observed no influence of skin tone on bias, unlike previous works that highlighted some effects on HR error rates [21]. However, our study is in line with another paper that was based on analyses of the signal to noise ratio, which showed no effect of skin tone when using a proper PPG wavelength (520 nm) [15]. More recently, another study did not find an effect of skin tone on beat-to-beat interval quality while separating skin types into two major groups (group 1: 1 to 4 on the Fitzpatrick Skin Scale; group 2: 5 and 6 on the Fitzpatrick Skin Scale) [43]. Even if our study findings are consistent with these results, no generalization can be made, since participants’ skin types ranged from 2 (n=6) to 4 (n=3) on the Fitzpatrick Skin Scale. Moreover, we did not take into account physiological and environment factors, such as local temperature, humidity, and sudation, which may impact peripheral vasomotricity and therefore increase or decrease PPG signal intensity [44-46]. It would also have been relevant to assess the influence of motion parameters, such as acceleration measured at the wrist, on HR accuracy and reliability of the FC4; however, we could not access raw data produced by the proprietary processes for further analyses, and it was not possible to use inertial units because these could have resulted in discomfort for the participants, and inertial units are not a commodity among the public. Furthermore, we chose to place the FC4 on the wrist of the nondominant arm, as this placement is recurrent in daily life and research, although we knew that the movements would be inferior to those of the dominant wrist and therefore would reduce the artifact rate. Finally, the low number of participants included for the running activity (n=3) should be considered while interpreting our results, although the pooled analysis lowered this potential bias while providing similar reliability and accuracy results.

    Comparison With Previous Works

    A recent study evaluated the accuracy of the FC4 HR sensor against an ECG Holter monitor for activities of daily living (sitting, walking, typing, lying down, etc) and showed acceptable HR measurement capabilities [47]. The added value of our protocol comes from the ecological approach to the gear sensor validation, which included additional physical activities with various intensities and limb movements. To our knowledge, no study has evaluated the FC4 in this manner during multiple sports or activities yet. However, further studies should be conducted to measure the reliability of the FC4 and its parameters for activities of daily living (number of steps, number of stairs climbed, number of calories burned, and quality of sleep), especially among persons or patients whose physical activities are restricted, as observed in sedentary individuals, people with obesity, people with heart failure, etc.

    Considering our results (ie, the lack of precision and reliability in specific activities), the FC4 should not be used for research or athlete training purposes, including those related to running, which showed the lowest LOA width and artifact ratio. However, the FC4 could be useful for tracking HR during daily activities, which does not require such accurate monitoring, and it may be considered for patients’ reeducation. However, for the latter, further studies should inspect the FC4’s reliability according to the characteristics of the population. For example, obesity affects the physiological factors necessary for proper PPG signal intensity and quality, such as capillary density and recruitment, blood flow, and skin thickness [44].

    Conclusion

    The FC4 shows excellent reliability for measuring HR during activities with slow and predictive arms movements, such as running. However, it should not be used for activities with “sharp” and random arm and wrist movements, such as soccer and racket sports, due to its sensitivity to motion artifacts. Hence, in ecological conditions, this device should not be used for research or training purposes due to the high artifact rate and LOA width.

    Authors' Contributions

    MC and EH designed the research. MC collected and analyzed the data. MC wrote the manuscript. EH and HD revised the manuscript. All authors have read and agreed to publish this version of the paper.

    Conflicts of Interest

    None declared.

    1. Costello K. Gartner says worldwide wearable device sales to grow 26 percent in 2019. Gartner. Nov 29, 2018. URL: https:/​/www.​gartner.com/​en/​newsroom/​press-releases/​2018-11-29-gartner-says-worldwide-wearable-device-sales-to-grow-#:~:text=Gartner%2C%20Inc.​,billion%20will%20be%20on%20smartwatches [Accessed 2024-12-12]
    2. Gilgen-Ammann R, Schweizer T, Wyss T. RR interval signal quality of a heart rate monitor and an ECG Holter at rest and during exercise. Eur J Appl Physiol. Jul 2019;119(7):1525-1532. [CrossRef] [Medline]
    3. Speer KE, Semple S, Naumovski N, McKune AJ. Measuring heart rate variability using commercially available devices in healthy children: a validity and reliability study. Eur J Investig Health Psychol Educ. Jan 10, 2020;10(1):390-404. [CrossRef] [Medline]
    4. Andre D, Wolf DL. Recent advances in free-living physical activity monitoring: a review. J Diabetes Sci Technol. Sep 2007;1(5):760-767. [CrossRef] [Medline]
    5. Hertzman AB. Photoelectric plethysmography of the fingers and toes in man. Exp Biol Med (Maywood). Dec 1, 1937;37(3):529-534. [CrossRef]
    6. Alian AA, Shelley KH. Photoplethysmography. Best Pract Res Clin Anaesthesiol. Dec 2014;28(4):395-406. [CrossRef] [Medline]
    7. Bartels K, Thiele RH. Advances in photoplethysmography: beyond arterial oxygen saturation. Can J Anesth. Dec 2015;62(12):1313-1328. [CrossRef] [Medline]
    8. Allen J. Photoplethysmography and its application in clinical physiological measurement. Physiol Meas. Mar 2007;28(3):R1-R39. [CrossRef] [Medline]
    9. Maeda Y, Sekine M, Tamura T, Moriya A, Suzuki T, Kameyama K. Comparison of reflected green light and infrared photoplethysmography. Presented at: 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society; Aug 20-25, 2008:2270-2272; Vancouver, BC. [CrossRef]
    10. Anderson RR, Parrish JA. The optics of human skin. J Invest Dermatol. Jul 1981;77(1):13-19. [CrossRef] [Medline]
    11. Cui WJ, Ostrander LE, Lee BY. In vivo reflectance of blood and tissue as a function of light wavelength. IEEE Trans Biomed Eng. Jun 1990;37(6):632-639. [CrossRef] [Medline]
    12. Giltvedt J, Sira A, Helme P. Pulsed multifrequency photoplethysmograph. Med Biol Eng Comput. May 1984;22(3):212-215. [CrossRef] [Medline]
    13. Maeda Y, Sekine M, Tamura T. Relationship between measurement site and motion artifacts in wearable reflected photoplethysmography. J Med Syst. Oct 2011;35(5):969-976. [CrossRef] [Medline]
    14. Lee J, Matsumura K, Yamakoshi KI, Rolfe P, Tanaka S, Yamakoshi T. Comparison between red, green and blue light reflection photoplethysmography for heart rate monitoring during motion. Presented at: 2013 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC); Jul 3-7, 2013:1724-1727; Osaka, Japan. [CrossRef]
    15. Fallow BA, Tarumi T, Tanaka H. Influence of skin type and wavelength on light wave reflectance. J Clin Monit Comput. Jun 2013;27(3):313-317. [CrossRef] [Medline]
    16. Rafolt D, Gallasch E. Influence of contact forces on wrist photoplethysmography--prestudy for a wearable patient monitor. Biomed Tech (Berl). 2004;49(1-2):22-26. [CrossRef] [Medline]
    17. Jo E, Lewis K, Directo D, Kim MJ, Dolezal BA. Validation of biofeedback wearables for photoplethysmographic heart rate tracking. J Sports Sci Med. Aug 5, 2016;15(3):540-547. [Medline]
    18. Hermand E, Cassirame J, Ennequin G, Hue O. Validation of a photoplethysmographic heart rate monitor: Polar OH1. Int J Sports Med. Jul 2019;40(7):462-467. [CrossRef] [Medline]
    19. Spierer DK, Rosen Z, Litman LL, Fujii K. Validation of photoplethysmography as a method to detect heart rate during rest and exercise. J Med Eng Technol. 2015;39(5):264-271. [CrossRef] [Medline]
    20. Dooley EE, Golaszewski NM, Bartholomew JB. Estimating accuracy at exercise intensities: a comparative study of self-monitoring heart rate and physical activity wearable devices. JMIR Mhealth Uhealth. Mar 16, 2017;5(3):e34. [CrossRef] [Medline]
    21. Shcherbina A, Mattsson CM, Waggott D, et al. Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. J Pers Med. May 24, 2017;7(2):3. [CrossRef] [Medline]
    22. Stahl SE, An HS, Dinkel DM, Noble JM, Lee JM. How accurate are the wrist-based heart rate monitors during walking and running activities? Are they accurate enough? BMJ Open Sport Exerc Med. Apr 25, 2016;2(1):e000106. [CrossRef] [Medline]
    23. Hough P, Glaister M, Pledger A. The accuracy of wrist-worn heart rate monitors across a range of exercise intensities. J Phys Act Res. Nov 25, 2017;2(2):112-116. [CrossRef]
    24. Lee CM, Gorelick M. Validity of the Smarthealth watch to measure heart rate during rest and exercise. Meas Phys Educ Exerc Sci. Jan 29, 2011;15(1):18-25. [CrossRef]
    25. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods. May 2007;39(2):175-191. [CrossRef] [Medline]
    26. Schaffarczyk M, Rogers B, Reer R, Gronwald T. Validity of the Polar H10 sensor for heart rate variability analysis during resting state and incremental exercise in recreational men and women. Sensors (Basel). Aug 30, 2022;22(17):6536. [CrossRef] [Medline]
    27. Fitzpatrick TB. The validity and practicality of sun-reactive skin types I through VI. Arch Dermatol. Jun 1988;124(6):869-871. [CrossRef] [Medline]
    28. Fitbit. Google Play. URL: https://play.google.com/store/apps/details?id=com.fitbit.FitbitMobile&hl=en [Accessed 2024-12-23]
    29. Polar Flow. URL: https://flow.polar.com/ [Accessed 2024-12-23]
    30. Mühlen JM, Stang J, Lykke Skovgaard E, et al. Recommendations for determining the validity of consumer wearable heart rate devices: expert statement and checklist of the INTERLIVE Network. Br J Sports Med. Jul 2021;55(14):767-779. [CrossRef] [Medline]
    31. Altman DG, Bland JM. Measurement in medicine: the analysis of method comparison studies. Journal of the Royal Statistical Society: Series D (The Statistician). Sep 1983;32(3):307-317. [CrossRef]
    32. Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med. Jun 2016;15(2):155-163. [CrossRef] [Medline]
    33. McBride GB. A proposal for strength-of-agreement criteria for Lin’s concordance correlation coefficient. National Institute of Water and Atmospheric Research; 2005. NIWA client report: HAM2005-062
    34. Gillinov S, Etiwy M, Wang R, et al. Variable accuracy of wearable heart rate monitors during aerobic exercise. Med Sci Sports Exerc. Aug 2017;49(8):1697-1703. [CrossRef] [Medline]
    35. Benedetto S, Caldato C, Bazzan E, Greenwood DC, Pensabene V, Actis P. Assessment of the Fitbit Charge 2 for monitoring heart rate. PLoS One. Feb 28, 2018;13(2):e0192691. [CrossRef] [Medline]
    36. Boudreaux BD, Hebert EP, Hollander DB, et al. Validity of wearable activity monitors during cycling and resistance exercise. Med Sci Sports Exerc. Mar 2018;50(3):624-633. [CrossRef] [Medline]
    37. Ahmadi AK, Moradi P, Malihi M, Karimi S, Shamsollahi MB. Heart rate monitoring during physical exercise using wrist-type photoplethysmographic (PPG) signals. Presented at: 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC); Aug 25-29, 2015:6166-6169; Milan, Italy. [CrossRef]
    38. Zhu S, Tan K, Zhang X, Liu Z, Liu B. MICROST: a mixed approach for heart rate monitoring during intensive physical exercise using wrist-type PPG signals. Presented at: 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC); Aug 25-29, 2015:2347-2350; Milan, Italy. [CrossRef]
    39. Mashhadi MB, Asadi E, Eskandari M, Kiani S, Marvasti F. Heart rate tracking using wrist-type photoplethysmographic (PPG) signals during physical exercise with simultaneous accelerometry. IEEE Signal Process Lett. Feb 2016;23(2):227-231. [CrossRef]
    40. Fujita Y, Hiromoto M, Sato T. PARHELIA: particle filter-based heart rate estimation from photoplethysmographic signals during physical exercise. IEEE Trans Biomed Eng. Jan 2018;65(1):189-198. [CrossRef] [Medline]
    41. Biswas D, Simoes-Capela N, Van Hoof C, Van Helleputte N. Heart rate estimation from wrist-worn photoplethysmography: a review. IEEE Sensors J. Aug 15, 2019;19(16):6560-6570. [CrossRef]
    42. Arunkumar KR, Bhaskar M. Heart rate estimation from wrist-type photoplethysmography signals during physical exercise. Biomed Signal Process Control. Mar 2020;57:101790. [CrossRef]
    43. Puranen A, Halkola T, Kirkeby O, Vehkaoja A. Effect of skin tone and activity on the performance of wrist-worn optical beat-to-beat heart rate monitoring. Presented at: 2020 IEEE SENSORS; Oct 25-28, 2020:1-4; Rotterdam, Netherlands. [CrossRef]
    44. Fine J, Branan KL, Rodriguez AJ, et al. Sources of inaccuracy in photoplethysmography for continuous cardiovascular monitoring. Biosens (Basel). Apr 16, 2021;11(4):126. [CrossRef] [Medline]
    45. Jeong IC, Yoon H, Kang H, Yeom H. Effects of skin surface temperature on photoplethysmograph. J Healthc Eng. 2014;5(4):429-438. [CrossRef] [Medline]
    46. Johnson JM. Exercise in a hot environment: the skin circulation. Scand J Med Sci Sports. Oct 2010;20 Suppl 3:29-39. [CrossRef] [Medline]
    47. Nissen M, Slim S, Jäger K, et al. Heart rate measurement accuracy of Fitbit Charge 4 and Samsung Galaxy Watch Active2: device evaluation study. JMIR Form Res. Mar 1, 2022;6(3):e33635. [CrossRef] [Medline]

    ‎
    bpm: beats per minute
    CCC: Lin concordance correlation coefficient
    ECG: electrocardiogram
    FC4: Fitbit Charge 4
    HR: heart rate
    ICC: intraclass correlation coefficient
    LOA: limit of agreement
    MAE: mean absolute error
    MAPE: mean absolute percentage error
    PPG: photoplethysmography

    Edited by Lorraine Buis; submitted 25.11.23; peer-reviewed by Luca Ardigò, Michael Nissen, Muhammad Etiwy, Shusuke Okita; final revised version received 10.10.24; accepted 24.10.24; published 08.01.25.

    Copyright

    © Maxime Ceugniez, Hervé Devanne, Eric Hermand. Originally published in JMIR mHealth and uHealth (https://mhealth.jmir.org), 8.1.2025.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.