Study design
Each participant visited the lab twice. The visits were at least 2 days but no more than 2 weeks apart and took place at the same hour of the day. The participants were asked to avoid intense and strenuous training the day before the tests. Furthermore, the participants were asked to abstain from alcohol 24 h and from food and drinks with caffeine for the 6 h before each test.
During their first visit, the athletes were informed about the study procedures, anthropometric data were measured, and the preliminary test was performed. The height and weight measurements were taken to the nearest 0.01 m using a stadiometer and to the nearest 0.01 kg using a calibrated scale (Model 213 and Model 877, respectively; seca GmbH, Hamburg, Germany). The two running trials were performed on a treadmill (Model Mercury, h/p/cosmos sports & medical GmbH, Nussdorf-Traunstein, Germany) with an increment of 1% to simulate outdoor running [14]. First, the participants participated in a submaximal incremental exercise test of maximally ten 5 min stages, starting at 5 km/h and with an incremental increase of 1.5 km/h per stage [15, 16]. The test was stopped when the participants reached a respiratory exchange ratio (RER) of ≥1.0 (mean over 1 min). Afterwards, the participants rested for 8 min. Second, the participants performed an all-out test to assess their HRmax and VO2peak. The all-out test started at 7 km/h, the first three stages lasted 1 min each, and the incremental increase was 1 km/h. The following stages lasted 30 s each, with 0.5 km/h incremental increases until volitional exhaustion [17]. During the last 15 s of each running stage, the participants were asked to rate their perceived exertion on a Borg scale ranging from 6 to 20 [18]. From the speed at VO2peak (vVO2peak), the individual’s relative speeds for the test on the second visit were calculated at 30%, 50%, 70%, 90%, and 110% of vVO2peak. To measure breath-by-breath automatic gas exchange, the Moxus Modular Metabolic System (AEI Technologies, Pittsburg PA, USA) was used. Several authors previously validated the Moxus Modular Metabolic System against the Douglas bag method and reported adequate to high reliability and reasonable validity during submaximal and maximal activities [4, 19].
On the second testing day, the participants were each fitted with three sports watches (Suunto Ambit2, Suunto Oy, Vantaa, Finland; Garmin Forerunner920XT, Garmin International Inc., Olathe KS, USA; Polar V800, Polar Electro Oy, Kempele, Finland) – and their corresponding HR monitors. The watches were set according to each individual’s age, height, weight, HRmax, and sex (Polar V800 only). The participants wore all three watches at the same time. Each participant wore two watches on the left wrist and forearm, the third watch on the right wrist, and the heart rate monitors (paired with the corresponding watch) around the chest. The positioning of the watches and the localization of the paired heart rate monitors was randomized. First, the participants were asked to stand still on the treadmill for 2 min, during which a baseline measurement was taken before the treadmill test began. The first three stages were performed at individual running speeds of 30%, 50%, and 70% of vVO2peak and lasted 10 min each, with a 2 min standing break in-between the stages. The last two stages, performed at 90% and 110% of vVO2peak, lasted 90 s each, with the same standing break in-between. All measurement devices were calibrated before each test and used in accordance with the manufacturer recommendations. The training profile “running” and for Garmin Forerunner920XT “indoor running” was selected from each watch’s menu. The watches were simultaneously started and stopped directly before and after each stage. The data were saved on the watch and synchronized using the proprietary online software (Suunto Movescount, Suunto Oy, Vantaa, Finland; Garmin Connect, Garmin International Inc., Olathe KS, USA; Polar Flow, Polar Electro Oy, Kempele, Finland) on a computer after each test. From there, the individual caloric values from the five stages were transferred to a database for further analysis.
Participants
Twenty healthy participants (12 males and 8 females) volunteered to participate in this study (age 23.90 ± 1.92 years, height 1.74 ± 0.08 m, weight 66.90 ± 10.02 kg, HRmax 193.10 ± 4.88 bpm, VO2peak 55.75 ± 7.33 ml/min/kg). All participants were recreational or competitive runners, and none of them had experienced any injury to their lower extremities within the past year. Before the first test, the participants were informed about the procedure and aims of the study and signed a written informed consent form that had been previously approved by the Institutional Review Board of the Swiss Federal Institute of Sport Magglingen. This study meets the principals outlined in the Declaration of Helsinki.
Data analysis – EE estimation during low to moderate running intensity
All the data from the watches was normalized to the unit of kcal/min. Missing values resulting from unsystematic HR monitor failure or malfunction were replaced using the relative difference (slope) from the reference mean to the specific watch mean from the corresponding running stage. For the EE measurements from the criterion measure, the formula of Elia and Livesey [20] was used to compute the total EE from the gas exchange data in kcal/min for the three submaximal categories (stage 1: 30% vVO2peak, stage 2: 50% vVO2peak, and stage 3: 70% vVO2peak). These formulas are commonly accepted for estimating EE during aerobic or submaximal intensities [6, 20,21,22,23,24,25]. However, very few studies have validated these formulas for anaerobic activities.
Data analysis – EE estimation during high-intensity running
The few studies that have examined high-intensity exercises generally reported low validity with regard to the criterion measure of indirect calorimetry [6, 26, 27]. Therefore, other methods were needed to overcome these measurement problems during vigorous physical activity. Medbo and colleagues [15] first proposed a new way to assess anaerobic proportions of EE during high-intensity physical activities. By assuming a linear relationship between running speed and oxygen uptake, they were able to interpolate to intensities greater than the maximal oxygen uptake [15]. From the intrapolated value at a certain speed or intensity, the measured oxygen consumption can be subtracted. The difference, integrated over the duration of the activity, can be used to estimate the maximal accumulated oxygen deficit (MAOD). Several authors reported MAOD to be the most accurate, non-invasive method for determining the anaerobic proportion of EE during high-intensity activities [16, 28, 29]. Therefore, the MAOD method was applied to compute the difference between the measured breath-by-breath gas exchange and the theoretically necessary oxygen uptake [15, 28] for the near-maximal and the supramaximal categories (stage 4: 90% vVO2peak and stage 5: 110% vVO2peak). Considering the high intensity of these two bouts and the measured RER values of ≥1.0 following these exercises, pure carbohydrates can be assumed as the muscle energy source. Therefore, the oxygen values, measured in ml/min, were multiplied by 5.04 kcal/l oxygen [25, 30].
Statistical analysis
The data were tested for normality using the Shapiro-Wilk test and mean values and standard deviations (SD) were calculated. The data were analyzed using a repeated-measures ANOVA with a Bonferroni post-hoc analysis. The validity of the three watches was initially investigated using Pearson’s correlation analyses. Furthermore, mean absolute error (MAE) and mean absolute percentage error (MAPE) of each watch compared to the criterion measure were calculated. As the threshold for accurate EE estimations, a MAPE ≤10% was defined, similar to the definition used by other researchers [11, 31]. The individual error, which was used specifically to assess inter-individual differences, was computed with the root mean square error (RMSE). Bland-Altman plots including 95% limits of agreement (±1.96 times SD) with their corresponding intercept and slope were created to graphically represent the data and to visualize systematic differences in EE estimation [32]. The level of significance was set at p < 0.05, and the statistical analyses were performed using SPSS 23 (IBM Corporation, Armonk NY, USA).