Introduction
Heart rate (HR) is one of the main signs for clinical examination and is used to assess the prognosis and response to treatment.1 Peripheral oxygen saturation of hemoglobin (SpO2) is a useful variable to monitor a patient’s oxygen status.2 The reference method for evaluating oxygen saturation is cooximetry or using the calculated arterial oxygen saturation from an arterial blood gas.3 However, the technique requires needle puncture of a peripheral artery and may be challenging for inexperienced veterinarians.4
Smartwatches are increasingly used to evaluate physiologic biomarkers.5 Apple Watch 6 (Apple Inc [AppW6]) is a reliable way to obtain HR and SpO2 in humans.6,7 The HR values obtained using Fenix 5X Plus (Garmin Ltd [GF5xp]) correlate well with the reference method,8 but overall, it tends to overestimate SpO2.9 Furthermore, GF5xp accurately measures HR in dogs.10 However, the reliability of GF5xp and AppW6 to measure HR and SpO2 in cats remains to be demonstrated. The aim of the study reported here was to evaluate the accuracy for 2 smartwatches with oximetry technology and optical wrist HR or single-lead ECG technology (Fenix 5X Plus [GF5xp], Garmin Ltd and Apple Watch 6 [AppW6], Apple Inc, respectively) versus gold standards (ECG and transmittance pulse oximetry [TPO], respectively) in measuring HR and SpO2 in cats.
Materials and Methods
The Atatürk University Local Board of Ethics Committee for Animal Experiments approved the experimental protocol. Written informed consent was obtained from the owners of the participating cats. The study was performed based on the guidelines outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals.11
A total of 10 male mixed-breed cats that were American Society of Anesthesiologists physical status 1, aged 8 to 12 months, and weighing 3.2 to 4.5 kg that were referred to Atatürk University Animal Hospital for elective castration between March 10 and April 15, 2022, were eligible for enrollment. Food was withheld for 6 hours and water was withheld for 2 hours before surgery. Animals received butorphanol (0.2 mg kg−1, IM), followed with dexmedetomidine (0.02 mg kg−1, IM) for sedation. Following the placement of the cephalic venous catheter, cefazolin (25 mg kg−1, IV) was administered. Induction of anesthesia was achieved with propofol (2 to 6 mg kg−1, IV) bolus. Each cat was intubated with a 4.5-mm endotracheal tube after its larynx was desensitized with 0.1 mL of 2% lidocaine. Then, oxygen (2 L/min) was supplied till extubation. Approximately 3 to 5 mL kg h−1 sterile saline (0.9% NaCl) solution was intravenously administered throughout anesthesia. A multiparameter monitor (Cardell 9405; Sharn Veterinary Inc) was used to monitor the HR (via 3-lead ECG), respiratory rate, rectal temperature, end-tidal carbon dioxide, non-invasive blood pressure, and SpO2 (via TPO). Meloxicam (0.2 mg kg−1, IV) was administered immediately after castration.
Measurements
All animals were placed in sternal recumbency throughout the experiment. An ECG with 3-lead systems (assumed as the reference method for HR)12 was used to collect HR measurements. The HR was also measured via thoracic auscultation with a stethoscope for 30 seconds to verify the accuracy of the ECG measurements. The SpO2 measurements were obtained from a probe attached to the tongue with TPO (assumed as the reference method for SpO2).13
Two smartwatches were used to measure the HR and SpO2 values: GF5xp and AppW6. An area over the proximolateral aspect of each tibia was clipped with a clipper (blade no: 40), and the GF5xp and AppW6 were then randomly attached with a watch strap over these prepared sites (Supplementary Figure S1). Each HR and SpO2 measurement obtained from the devices was recorded by the same person; 4 people (1 for ECG, 1 for stethoscope, 1 for GF5xp, and 1 for AppW6) simultaneously recorded measurements. Each person recorded their device’s HR measurements on the datasheet at an interval of 30 seconds for 10 minutes, whereas SpO2 measurements were collected at any point during the respiratory cycle without any specific time intervals. All data were then transferred to a spreadsheet file (Excel; Microsoft Corp) for statistical analysis. The results were classified as device failure when the HR or SpO2 were not displayed.
Statistical analysis
Power analysis (PS-Power and Sample Size calculation; version 3.1.2) was employed to determine the minimum number of measurements required in each device to consider a 2% difference in the HR significant, thus 10 animals (200 measurements) were needed. The level of significance was 0.05 (Type I) and a power of 80% (Type II). The analysis was based on data by previous study which compares HR measurement in dogs using different smartwatches.10
Results are presented in terms of the mean difference (%) and the 95% limits of agreement (LoA) with appropriate measurements of precision for each estimated parameter. For each comparison pair, we also regressed the differences on the mean of the combined measurements to assess the relationship between bias and the magnitude of the measurement.14 Finally, for each comparison method we have also reported the proportion of measurements falling within the LoA.
Commercial software (Med-Calc; version 20.110; MedCalc Software Ltd) was used for statistical analysis. The normality of the data was tested prior to analysis using the Shapiro–Wilk test. Non-normal distributed data (ECG and stethoscope) were subjected to the Mann–Whitney U test. The failure rate for the SpO2 measurements was subjected to cross-tabulation for χ2 analysis. Agreement between results obtained with ECG or TPO versus those obtained with the tested smartwatches were evaluated using the Bland-Altman plot, in which the differences (%) between methods were plotted against their mean HR or SpO2 (reference method measurement – test device measurement) and the LoA (mean ± 1.96 × SD).15 Values of P < .05 were considered statistically significant. Data were presented as mean ± SD or as median (range).
Results
All animals recovered from anesthesia without any complications. A total of 800 HR measurements were obtained from ECG (n = 200 measurements), stethoscope (200 measurements), AppW6 (200 measurements), and GF5xp (200 measurements). No failure rate was recorded for HR in any of the devices. No significant (P = .843) difference was observed between the HR values of the ECG (median, 166 beats/min; range, 148 to 181 beats/min) and stethoscope (median, 164 beats/min; range, 152 to 180 beats/min]). Compared with ECG measurements of HR, GF5xp had superior bias (–0.1%) and LoA (3.0% to –3.3%) versus the AppW6 (bias, 0.2%; LoA, 3.7% to –3.4%; ; Figure 1). GF5xp (3.04) achieved a lower bias than AppW6 (3.71) in estimating HR. The SE of estimate for HR was quite low for GF5xp (0.0124) and AppW6 (0.014).
Bland-Altman plots of differences between measurements of heart rate (HR; beats/min) obtained with 3-lead ECG connected to a multiparameter monitor (reference method) versus a smartwatch with either single-lead ECG technology (Apple Watch 6 [AppW6], Apple Inc; A) or optical wrist HR technology (Fenix 5X Plus [GF5xp], Garmin Ltd; B) attached over the lateral aspect of the tibial regions, presented as percentages and plotted against the mean of the 2 measurements, for 10 client-owned cats undergoing elective castration between March 10 and April 15, 2022, with 200 paired measurements for each device (ECG and AppW6; ECG and GF5xp). For each plot, each asterisk represents the difference between ECG and AppW6 or GF5xp measurements at each measurement time point, the uppermost and lowermost dashed lines represent the upper and lower limits of agreement (LoA), the solid straight line represents the mean bias (0.2 for AppW6; –0.1 for GF5xp), the dotted line represents 0% difference, the dotted-and-dashed lines represent the best fit, the curvilinear solid lines represent 95% Cl of the best fit, and the error bars represent the 95% CIs of the mean bias and LoA.
Citation: Journal of the American Veterinary Medical Association 261, 2; 10.2460/javma.22.08.0357
A total of 216 measurements were obtained from 3 different devices for SpO2. The failure rate for SpO2 measurement was significantly (P < .001) lower in GF5xp (94 in 140 attempts) than in AppW6 (110 in 140 attempts). No failure rate for SpO2 was observed in the TPO (140 in 140 attempts). Compared with TPO measurements of SpO2, AppW6 had superior bias (0.2%) and LoA (3.0% to –2.5%) versus those of the GF5xp (bias, –2.1%; LoA, 0.2 to –4.4%; Figure 2). The GF5xp (0.18) achieved a lower bias than AppW6 (–2.50) in estimating SpO2. The SE of estimate for SpO2 was quite low in both GF5xp (0.136) and AppW6 (0.165).
Bland-Altman plots of differences between measurements of peripheral oxygen saturation of hemoglobin measured with pulse oximetry (SpO2) by transmittance pulse oximetry (TPO) of a multiparameter monitor (reference method; n = 140/140 measurements) versus 2 smartwatches with oximetry technology: AppW6 (A; 110/140 measurements) or GF5xp (B; 94/140 measurements) for the cats described in Figure 1. Each asterisk represents the difference between TPO and AppW6 or GF5xp measurements at each time point, the uppermost and lowermost dashed lines represent the upper and lower LoA, the solid straight line represents the mean bias (0.2 for AppW6; –2.1 for GF5xp), the dotted line represents 0% difference, the dotted-and-dashed lines represent the best fit, the curvilinear solid lines represent 95% Cl of the best fit, and the error bars represent the 95% CIs of the mean bias and LoA.
Citation: Journal of the American Veterinary Medical Association 261, 2; 10.2460/javma.22.08.0357
Discussion
The present study determined the accuracy of 2 smartwatch devices for HR and SpO2 in anesthetized cats. Our findings indicated that data from GF5xp and AppW6 were overall acceptable and consistent with that from ECG over the range of HR and SpO2 observed. GF5xp was superior to AppW6 as an alternative measurement, with a lower LoA for detecting HR compared with that of the reference method. A previous study in dogs has reported that GF5xp has high accuracy for detecting HR.10 To our knowledge, no other study has investigated the accuracy of HR in animals using AppW6. However, our results were consistent with a previous study that showed that AppW6 has a high level of accuracy for HR monitoring in humans.16
To the authors’ knowledge, this study was the first to evaluate the validity of GF5xp and AppW6 in measuring SpO2 in cats. Therefore, comparing the current findings with previous reports on animals may not be possible. Nevertheless, a previous study that utilized AppW6 in humans have found that the mean biases of HR and SpO2 are 0% and 0.8%,6 respectively, which was consistent with our findings (mean biases of HR and SpO2 are –0.1% and 0.2%, respectively).
The accuracy of SpO2 measurements obtained from GF5xp exhibits minimal overestimation (mean bias, 3.3%; LoA, −1.9 to 8.6%),17 whereas minimal underestimation was acquired in the present work (mean bias, –2.1%; LoA, 0.2 to –4.4%). Schiefer et al9 evaluated the accuracy of GF5xp to monitor SpO2 at high altitudes. Although GF5xp showed high validity (mean bias, 0.1%; LoA, –10.7 to 10.9%), arterial blood gas analysis showed a mean bias of 7.0% with a wide LoA (−6.5 to 20.5%). Therefore, using GF5xp for monitoring SpO2 for predictive health status is not recommended.9 This finding is likely related to the extreme values obtained at high altitudes. Smartwatches exhibit poor accuracy during rigorous exercises.18,19 The low LoA in the current study may be associated with anesthesia, which is related to stable conditions in cats. Notably, false values may be recorded in humans when motion and hand movement occur.20 Future studies should be performed in cats during physical activity to understand whether both devices can be surrogates to reference methods.
Although we demonstrated that GF5xp and AppW6 could be used to assess SpO2 in cats, both smartwatches frequently failed to display values for pulse oximetry. This result might restrict the clinical use of devices in cats. The tightness of the watch and the environmental light may affect successful measurements.21 These factors were not considered in the current study and might be considered as limitations. We conclude that HR measurements can easily be obtained when smartwatches are tightly wrapped over the proximal tibia. Nevertheless, this procedure may not enable us to record most of the SpO2 measurements. Future research can be conducted by placing smartwatches on different anatomic regions to examine how correct measurements can be obtained.
In conclusion, various brands of smartwatches have been used in human healthcare. Apple and Garmin are both well known in the world of smartwatches; however, no study has evaluated the use of these devices in cats to monitor HR and SpO2. Based on our findings, both GF5xp and AppW6 exhibit high accuracy in evaluating HR and SpO2 when compared with the reference methods. Both devices are cost-effective when used to monitor HR in cats and can be used for follow-up assessment for screening heart disorders. Thus, both wearable devices may not only serve to humans but also cats. However, the technology may not be suitable for screening SpO2 in cats because of high failure rates. Given that the cats used in this study are free of any cardiac disease, future studies are required to determine the effectiveness of both devices in detecting cardiovascular problems.
Supplementary Materials
Supplementary materials are posted online at the journal website: avmajournals.avma.org
Acknowledgments
No third-party funding or support was received in connection with this study or the writing or publication of the manuscript. The authors report no conflicts of interest related to this study.
References
- 1.↑
Atkins CE, Gallo AM, Kurzman ID, Cowen P. Risk factors, clinical signs, and survival in cats with a clinical diagnosis of idiopathic hypertrophic cardiomyopathy: 74 cases (1985–1989). J Am Vet Med Assoc. 1992;201(4):613–618.
- 2.↑
Hoffmeister EH, Read MR, Brainard BM. Evaluating veterinarians’ and veterinary students’ knowledge and clinical use of pulse oximetry. J Vet Med Educ. 2005;32(2):272–277.
- 3.↑
Farrell KS, Hopper K, Cagle LA, Epstein SE. Evaluation of pulse oximetry as a surrogate for PaO2 in awake dogs breathing room air and anesthetized dogs on mechanical ventilation. J Vet Emerg Crit Care (San Antonio). 2019;29(6):622–629.
- 4.↑
Thawley V, Waddell LS. Pulse oximetry and capnometry. Top Companion Anim Med. 2013;28(3):124–128.
- 5.↑
Rienzo MD, Mukkamala R. Wearable and nearable biosensors and systems for healthcare. Sensors (Basel). 2021;21(4):1291. doi:10.3390/s21041291
- 6.↑
Pipek LZ, Nascimento RF, Acencio MM, Teixeria LR. Comparison of SpO2 and heart rate values on Apple Watch and conventional commercial oximeters devices in patients with lung disease. Sci Rep. 2021;11(1):18901. doi:10.1038/s41598-021-98453-3
- 7.↑
Spaccarotella C, Polimeni A, Mancuso C, Pelaia G, Esposito G, Indolfi C. Assessment of non-invasive measurements of oxygen saturation and heart rate with an Apple smartwatch: comparison with a standard pulse oximeter. J Clin Med. 2022;11(6):1467. doi: 10.3390/jcm11061467
- 8.↑
Serantoni C, Zimatore G, Bianchetti G, Abeltino A, De Spirito M, Maulucci G. Unsupervised clustering of heartbeat dynamics allows for real time and personalized improvement in cardiovascular fitness. Sensors (Basel). 2022;22(11):3974. doi: 10.3390/s22113974
- 9.↑
Schiefer LM, Treff G, Treff F, et al. Validity of peripheral oxygen saturation measurements with the Garmin Fēnix 5X plus wearable device at 4559 m. Sensors (Basel). 2021;21(19):6363. doi: 10.3390/s21196363
- 10.↑
Yanmaz LE, Okur S, Ersoz U, Senocak MG, Turgut F. Accuracy of heart rate measurements of three smartwatch models in dogs. Top Companion Anim Med. 2022;49:100654.
- 11.↑
National Research Council. Guide for the Care and Use of Laboratory Animals. 8th ed. National Academy Press; 2011.
- 12.↑
Kraus MS, Gelzer AR, Rishniw M. Detection of heart rate and rhythm with a smartphone-based electrocardiograph versus a reference standard electrocardiograph in dogs and cats. J Am Vet Med Assoc. 2016;249(2):189–194.
- 13.↑
Mair A, Martinez-Taboada F, Nitzan M. Effect of lingual gauze swab placement on pulse oximeter readings in anaesthetised dogs and cats. Vet Rec. 2017;180(2):49.
- 14.↑
Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8(2):135–160.
- 15.↑
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–310.
- 16.↑
Hajj-Boutros G, Landry-Duval MA, Comtois AS, Gouspillou G, Karelis AD. Wrist-worn devices for the measurement of heart rate and energy expenditure: a validation study for the Apple Watch 6, Polar Vantage V and Fitbit Sense. Eur J Sport Sci. Published online January 31, 2022. doi:10.1080/17461391.2021.2023656
- 17.↑
Lauterbach CJ, Romano PA, Greisler LA, Brindle RA, Ford KR, Kuennen MR. Accuracy and reliability of commercial wrist-worn pulse oximeter during normobaric hypoxia exposure under resting conditions. Res Q Exerc Sport. 2021;92:549–558.
- 18.↑
Evenson KR, Goto MM, Furberg RD. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act. 2015;12:159. doi: 10.1186/s12966-015-0314-1
- 19.↑
Fuller D, Colwell E, Low J, et al. Reliability and validity of commercially available wearable devices for measuring steps, energy expenditure, and heart rate: systematic review. JMIR Mhealth Uhealth. 2020;8(9):e18694. doi: 10.2196/18694
- 20.↑
Raja JM, Elsakr C, Roman S, et al. Apple watch, wearables, and heart rhythm: where do we stand? Ann Transl Med. 2019;7(17):417. doi: 10.21037/atm.2019.06.79
- 21.↑
Castaneda D, Esparza A, Ghamari M, Soltanpur C, Nazeran H. A review on wearable PPG sensors and their potential future applications in health care. Int J Biosens Bioelectron. 2018;4(4):195–202.
