Validation of Aerobic Capacity (VO2max) and Lactate Threshold in Wearable Technology for Athletic Populations

s wearable technology (WT) has evolved, devices have developed the ability to track a range of physiological variables. These include maximal aerobic capacity (VO 2max) and lactate threshold (LT). With WT quickly growing in popularity, independent evaluation of these devices is important to determine the appropriate use-cases for the devices. Therefore, the purpose of this study was to determine the validity of WT in producing estimates of VO 2maxand LT in athletic populations. METHODS: 21 participants completed laboratory LT and VO 2maxtesting, as well as an outdoor testing session guided by the WT being tested (Garmin f¯enix 6 ® watch and accompanying heart rate monitor). Statistical analysis was completed, using hypothesis testing (ANOVA,t-test), correlation analysis (Pearson's r, Lin's Concordance Correlation [CCC]), error analysis (mean absolute percentage error [MAPE]), equivalence testing (TOST test), and bias assessment (Bland–Altman analysis). RESULTS: The Garmin watch was found to have acceptable agreement for VO 2maxwhen compared to the 1 min averaged values (MAPE = 6.85%, CCC = 0.7) and for LT and the onset of blood lactate accumulation (OBLA), (MAPE = 7.52%, CCC = 0.79; MAPE = 8.20%, CCC = 0.74, respectively). Therefore, the Garmin f¯enix 6 ® produces accurate measurements of VO 2maxand LT in athletic populations and can be used to make training decisions among athletes. Keywords:Garmin; training;

averaged values (MAPE = 6.85%, CCC = 0.7) and for LT and the onset of blood lactate accumulation (OBLA), (MAPE = 7.52%, CCC = 0.79; MAPE = 8.20%, CCC = 0.74, respectively). Therefore, the Garmin f¯enix 6 ® produces accurate measurements of VO 2maxand LT in athletic populations and can be used to make training decisions among athletes. Keywords:Garmin; training; endurance; running; elite athletes; OBLA 1. Introduction Among the two most important parameters for predicting endurance performance are maximal aerobic capacity (VO2max), and lactate threshold (LT) [1–4]. VO2maxrepresents the highest amount of oxygen an individual is capable of bringing into the body and utilizing to produce energy [4]. The lactate threshold is the point just prior to an exponential rise in lactate concentrations, a metabolic byproduct of anaerobic metabolism that increases in concentration during exercise, especially intense exercise [4]. Traditionally, LT has been obtained in a laboratory setting with the use of blood lactate analyzers and a graded exercise test, either on a bike or treadmill [4]. To determine VO2max, the use of a metabolic cart to measure oxygen consumption during a graded exercise test can determine an individual's maximal aerobic capacity [4]. While eld-based tests have been developed for the estimation of VO2maxthey are not as accurate as laboratory measurements [5–7]. While the eld tests do have the bene t of increased accessibility as a result of not requiring expensive equipment and trained technicians to administer the tests, it also opens up the ability to test multiple people simultaneously. As wearable technology (WT) continues to evolve, it may serve as another tool to determine VO2maxand LT as predictive measures of endurance performance. Wearable technology utilization continues to grow in popularity and prevalence, both in recreational and higher-level athletics [5–14]. As WT becomes more sophisticated, its us- age will increase to a greater portion in each population [15]. Wearable technology devices are usually worn around the wrist or chest but can vary in terms of sensor placement [16]. Some of the ways WT can be used are varied, including medical [17], biomechanical [18], Technologies2023,11, 71.

As WT becomes more sophisticated, its us- age will increase to a greater portion in each population [15]. Wearable technology devices are usually worn around the wrist or chest but can vary in terms of sensor placement [16]. Some of the ways WT can be used are varied, including medical [17], biomechanical [18], Technologies2023,11, 71.

Technologies2023,11, 71 2 of 10 physiological [16], tactical [19], and for training decisions [20]. Some of the variables mea- sured or estimated by current WT include factors such as heart rate, VO2max, LT, blood oxygen saturation, energy expenditure, sleep quantity, heart rate variability (HRV), and ground contact time. A review of technology in team sports identi ed four general cate- gories of use for integrated technology, including [21] quantifying movement patterns, [15] assessing the demands of training and competition, [22] measuring physiological and metabolic responses, and [23] determining velocity and sprint effort [24]. Wearable devices that can estimate LT using heart rate, muscle oxygen and sweat sensors have been intro- duced [25–29]. Another potential bene t of WTs is that they can estimate VO2maxand LT in a non-invasive manner. The values produced by the technology can be used to make training decisions based on individual physiological responses. Obtaining these values via WT in the eld could offer an advantage to athletes and teams in terms of cost and availability [16]. As we have established, VO2maxand LT are critically important values in determining endurance performance, as an increase in these metrics allows athletes to sustain a higher intensity of exercise for longer, which is particularly important for endurance-based perfor- mances. While WT has its advantages and disadvantages in estimating these values, up to this point, studies have primarily been conducted using the general population rather than athletic populations. It remains unknown how well this technology works with higher-level athletes compared to the general population. Trained individuals and athletes have higher VO2maxvalues and reach LT at higher running speeds than untrained individuals. This may present a challenge for wearable technology in determining the VO2maxand LT of highly trained individuals, compared to untrained or lightly trained individuals. While previous research has validated VO2max[16,30,31] and LT [25,26] in WT using the general population, it is important to determine whether limitations exist in athletic populations for use in collegiate and professional athletics. Recruitment of high-level athletes makes this work unique and will allow athletes, coaches, researchers, and others to better understand the use-case of this

or lightly trained individuals. While previous research has validated VO2max[16,30,31] and LT [25,26] in WT using the general population, it is important to determine whether limitations exist in athletic populations for use in collegiate and professional athletics. Recruitment of high-level athletes makes this work unique and will allow athletes, coaches, researchers, and others to better understand the use-case of this technology and who may bene t from its use. Therefore, the purpose of this study was to determine the validity of wearable technology to estimate VO2maxand LT in athletic populations. 2. Materials and Methods 2.1. Study Design Prior to data collection occurring for this investigation, the protocols were approved by the University of Nevada, Las Vegas Institutional Review Board (IRB, 1525606-12). All participants signed an informed consent and lled out pre-assessment documents prior to completing the study. Data collection occurred over two days and included a laboratory testing day and an outdoor/ eld testing day. After consenting to the study, demographic data were obtained (24.24 6.30 years, 11 male, 10 female, 171.68 8.01 cm, 65.14 9.41 kg , BMI = 22.01 1.91, 17.04 5.69% fat mass, 39.25 3.26% muscle mass, 42.49 22.96 km per week, all reported as mean SD). Next, a treadmill-based graded exercise test utilizing speed and grade progression every two to three minutes was performed to determine both LT and VO2max. Blood lactate levels were measured via the handheld Lactate Plus analyzer (Nova Biomedical Corp, Waltham, MA, USA). Oxygen consumption was determined by the ParvoMedics TrueOne 2400 metabolic cart (ParvoMedics Inc, Salt Lake City, UT, USA). The lactate threshold was determined by graphing the data and determining the point just prior to an exponential rise in lactate concentration (>1 mmol/L rise) that also corresponded to a nal concentration above 4 mmol/L. The onset of blood lactate accumulation (OBLA) was determined by solv- ing the slope-intercept equation for speed when lactate concentration equaled 4 mmol/L. VO2maxwas determined by taking the highest average oxygen consumption during the graded exercise test for a set timeframe. VO2maxvalues for 4-breath, 15 s, 30 s, and 1 min average timeframes were obtained by

concentration above 4 mmol/L. The onset of blood lactate accumulation (OBLA) was determined by solv- ing the slope-intercept equation for speed when lactate concentration equaled 4 mmol/L. VO2maxwas determined by taking the highest average oxygen consumption during the graded exercise test for a set timeframe. VO2maxvalues for 4-breath, 15 s, 30 s, and 1 min average timeframes were obtained by the metabolic cart to compare to the wearable device.

Technologies2023,11, 71 3 of 10 After laboratory values were obtained for LT and VO2max, participants returned between two and seven days (5.56 2.53) after the laboratory-based test to complete the outdoor testing session. The outdoor run was conducted in one of two places, the University track or a at area of campus, depending on track availability. Ten participants ran the track, and eleven completed the protocol on campus. The altitude was ~686 m, and the average temperature during outdoor testing was 22.01 9.57 C. The outdoor testing involved completing two separate runs while wearing the tness tracker watch (Garmin f¯enix 6 ® , Garmin Ltd., Olathe, KS, USA) and accompanying heart rate monitor (Garmin HRM-Run ® ). A factory reset on the watch was performed before each test so that previous data did not in uence the estimate of VO2maxor LT. The rst run was a 10–15 min run at above 70% of the estimated max heart rate (MHR). This gave the device enough data to estimate VO2max, using a linear extrapolation of heart rate (HR) and running speed [21,22]. For the 10–15 min run to determine VO2max, the average distance, time, pace, and HR were 2.52 0.37 km, 12.63 3.19 min, 5.1 1.43 min/km, and 154.8 10.28 bpm, respectively. After the 10–15 min run, participants were given up to 10 min to rest before the next run, which was a graded exercise test guided by the watch. Participants were provided with a HR range via the watch and instructed to run at a pace that could be maintained within that intensity range. The HR window progressively increased every 3–4 min, and participants were required to speed up to match the new HR window. This continued until the watch concluded the test or the participants voluntarily stopped prematurely, which concluded the outdoor data collection. For the progressive exercise test to determine LT, the watch utilizes HRV during exercise to identify LT [23]. The average distance, time, pace, and HR for the graded exercise test was 3.42 0.98 km, 16.71 5.41 min, 4.99 1.45 min/km, 164.25 9.81 bpm, respectively. If the

the test or the participants voluntarily stopped prematurely, which concluded the outdoor data collection. For the progressive exercise test to determine LT, the watch utilizes HRV during exercise to identify LT [23]. The average distance, time, pace, and HR for the graded exercise test was 3.42 0.98 km, 16.71 5.41 min, 4.99 1.45 min/km, 164.25 9.81 bpm, respectively. If the f¯enix 6 was not able to produce an estimate, either because the participant had to end early or the watch failed to produce an estimate for unknown reasons, participants were asked to return on a different day to perform the outdoor test again. If the device was unable to produce an estimate after two different attempts, participants were not tested a third time. There were two participants for whom the watch was unable to generate an estimate of LT. There was one participant whose data were not recorded prior to resetting the watch and was lost. Therefore, while the total number of participants was 21, LT analysis was performed with 18 subjects, and VO2max with 20. 2.2. Participants For this study, apparently healthy individuals who exercised regularly (>3 times per week) were recruited. Of those that were tested, 21 scored in the 95th percentile or above for their VO2maxvalues, based on their age and biological sex, and were included in the athletic population dataset for the current investigation. 2.3. Data Analysis Data for lactate concentration and speed were input directly into Google Sheets (Al- phabet Inc., Mountain View, CA, USA), and further analysis to determine LT and OBLA for each participant was completed within Google Sheets. VO2maxand associated percentile for each timeframe (4-breath, 15 s, 30 s, and 1 min) was determined by the ParvoMedics soft- ware and input into Google Sheets. All granular calculations were completed within Google Sheets. All hypothesis testing, summary statistics, validation measures, and gures were completed and generated in jamovi (jamovi project, version 2.2,, accessed on 22 March 2023). These include ANOVA's with post hoc pairwise comparisons with Tukey adjustments for multiple comparisons (when appropriate), descriptive statistics, error analysis (mean absolute percentage error), correlation analysis

Google Sheets. All granular calculations were completed within Google Sheets. All hypothesis testing, summary statistics, validation measures, and gures were completed and generated in jamovi (jamovi project, version 2.2,, accessed on 22 March 2023). These include ANOVA's with post hoc pairwise comparisons with Tukey adjustments for multiple comparisons (when appropriate), descriptive statistics, error analysis (mean absolute percentage error), correlation analysis (Pearson's r, Lin's Concordance Correlation Coef cient [CCC]), equivalence testing (TOST Paired Samples Test), and bias assessment (Bland–Altman analysis). TOST test lower and upper bounds were set at +0.5 and 0.5 Cohen's D for each test. Data analysis for VO2maxwas completed by comparing the f¯enix 6 estimates of VO2maxto each laboratory timeframe. Data analysis

Technologies2023,11, 71 4 of 10 for LT was completed by comparing the f¯enix 6 estimates of speed at LT and HR at LT to the laboratory values (speed at LT, speed at OBLA, and HR at LT). Determination of validation was pre-determined, and any device that produced a CCC 0.7 and a MAPE < 10% was considered valid [16]. 3. Results 3.1. VO2max The 21 participants used for this analysis had an average VO2maxpercentile of98.24 1.3% , based on the 30 s averaged VO2maxvalues. The one-way ANOVA for VO2maxat each time showed a signi cant difference for the global test (F = 5.59,p< 0.001, 2 = 0.19). Further post hoc pairwise comparisons with Tukey adjustments for multiple comparisons were performed. The f¯enix 6 estimate was signi cantly different from the 4-breath average (t = 4.52 ,p< 0.001, Cohen's D = 1.43), but not different for any of the other time compar- isons (see Table). Error analysis showed that the f ¯enix 6 VO2maxestimate had a MAPE of less than 10% when compared to the 30 s and 1 min averaged time parameters (see Table). Correlation analysis produced a CCC 0.7 for the 1 min averaged time only (see Table). Equivalence testing via TOST test was violated for all four time parameters (see Table). Bland–Altman bias values and 95% con dence intervals can be found in Table, and associated plots can be found for all time parameters in Figure. Table 1.VO 2maxdescriptive and validation statistics results. F¯enix 6 VO2max Estimate Lab VO2max—4 Breath Avg Lab VO2max—15 Sec Avg Lab VO2max—30 Sec Avg Lab VO2max—1 Min Avg Mean (mL/kg/min) 54.00 64.73 59.43 57.88 56.89 Standard Deviation 5.18 8.83 7.80 7.62 7.60 MAPE 16.91% 10.04% 7.67% 6.85% Pearson Correlation 0.81 0.82 0.82 0.81 Lin's Concordance 0.34 0.58 0.67 0.70 Bland–Altman Bias 10.485 ( 13.09, 7.88) 5.18 ( 7.33, 3.03) 3.62 ( 5.68, 1.56) 2.65 ( 4.75, 0.55) TOST Test (Upper) <0.001 <0.001 <0.001 <0.001 TOST Test (Lower) 1.00 0.994 0.917 0.653 n = 20. MAPE = Mean Absolute Percentage Error, TOST Test = Two One-Sided T-Tests. Bland–Altman bias values and 95% con dence

Lin's Concordance 0.34 0.58 0.67 0.70 Bland–Altman Bias 10.485 ( 13.09, 7.88) 5.18 ( 7.33, 3.03) 3.62 ( 5.68, 1.56) 2.65 ( 4.75, 0.55) TOST Test (Upper) <0.001 <0.001 <0.001 <0.001 TOST Test (Lower) 1.00 0.994 0.917 0.653 n = 20. MAPE = Mean Absolute Percentage Error, TOST Test = Two One-Sided T-Tests. Bland–Altman bias values and 95% con dence intervals provided. Values that met the predetermined validation criteria are bolded.Technologies 2023, 11, x FOR PEER REVIEW 5 of 10 Figure 1. A representative sampling of TOST test results for VO 2max and LT data. Far left = Lab HR at LT to fēnix 6 HR at LT, middle left = speed at OBLA to fēnix 6 LT speed, middle right = fēnix 6 VO 2max to 30 s avg VO2max, far right = fēnix 6 VO 2max to 1 min avg VO2max. Upper and lower bounds set at +0.5 and −0.5 Cohen’s D. All tests shown violated equivalence testing parameters except lab HR at LT to fēnix 6 HR at LT (far left). Figure 2. Bland–Altman plots for VO 2max time parameters. Top left = fēnix 6 and 4-breath avg, top right = fēnix 6 and 15 s avg, bottom left = fēnix 6 and 30 s avg, and bottom right = fēnix 6 and 1 min avg. The horizontal solid line represents a theoretical difference of 0, the middle-dotted line repre- sents the average difference, and the top and bottom dotted lines represent upper and lower bounds of the 95% limit of agreement interval. The diagonal solid blue line represents the proportional bias line with associated confidence intervals in gray shading. Figure 1. A representative sampling of TOST test results for VO 2maxand LT data. Far left = Lab HR at LT to f¯enix 6 HR at LT, middle left = speed at OBLA to f¯enix 6 LT speed, middle right = f¯enix 6 VO 2max to 30 s avg VO 2max, far right = f¯enix 6 VO 2maxto 1 min avg VO 2max. Upper and lower bounds set at +0.5 and 0.5 Cohen's D. All tests shown

Lab HR at LT to f¯enix 6 HR at LT, middle left = speed at OBLA to f¯enix 6 LT speed, middle right = f¯enix 6 VO 2max to 30 s avg VO 2max, far right = f¯enix 6 VO 2maxto 1 min avg VO 2max. Upper and lower bounds set at +0.5 and 0.5 Cohen's D. All tests shown violated equivalence testing parameters except lab HR at LT to f¯enix 6 HR at LT (far left).

The study assesses the accuracy of Garmin fēnix 6 in measuring VO2max and lactate threshold.