Two different studies were carried out to assess validity and reliability on two different groups separately. Sixteen Caucasian physically active volunteers (3 women and 13 men), between 19 and 32 years old, volunteered to participate in the validation study (Table 1). Another fourteen Caucasian physically active volunteers (3 women and 11 men), between 20 and 43 years old, agreed to participate in the reliability study (Table 1). All volunteers provided their written consent after being informed about the risks and benefits of the study, which was approved by the ethical committee of the University Hospital Dr. Negrin (Ref. 140187). All subjects were requested not to exercise and to refrain from drinking alcohol and beverages containing caffeine or taurine during the 48 h preceding the tests, as well as to eat a similar dinner the night before the tests.

Subjects reported to the laboratory between 7:00 and 9:30 A.M., following a 12-h overnight fast. Upon arrival, their body weight and height were measured to the nearest 0.1 kg and 0.1 cm, respectively. Measurements were performed while subjects wore light clothes and no shoes using a balance scale (Seca, Hamburg, Germany) calibrated using certified calibration masses of class M1 (Scheck, Germany).

The validity of the COSMED K5 was determined by comparison with the Vyntus CPX (Jaeger-CareFusion, Höchberg, Germany) both at rest and during submaximal exercise on a cycle ergometer (Corival, Lode, Netherlands). Both metabolic carts were equipped with relatively new (less than 3 months old) oxygen fuel cells). Before the measurements both devices were warmed up for a minimum of 15 min calibrated with high-grade calibration gases provided by the manufacturers and by pumping gas with a 3 L calibration syringe through the flow meters, following the recommendations of the manufacturers. Mask size was individually fitted prior to the first test and the same size was maintained for subsequent trials.

The combustion of butane (2 C₄H₁₀ + 13 O₂ → 8 CO₂ + 10 H₂O + heat) has a respiratory quotient of 0.615 (8/13). Thus, we built an air-tight Plexiglas box (70 cm × 50 cm × 50 cm) with two 3.05 cm circular openings. One opening was connected to a 3-L calibration syringe (Hans Rudolf, Shawnee, KS) and in the other the turbine of the Vyntus with the sample lines for gas analysis was placed. Inside the Plexiglas box, a small butane burner was ignited while a 5 liter stainless steel cooking pot was placed on top filled with 0.5 L of water and ice up to the top, to dissipate the heat generated by the butane combustion. Three different burning intensities were used to simulate low, medium and very high exercise intensity. Within 5 s after the ignition of the butane burner, air was pumped in and out continuously using the calibration syringe, adjusting the stroke rate between 15 and 80 strokes/min until an almost “steady state” was reached, as reflected by the F_EO₂ and F_ECO₂. The pumping rate was then maintained for 2 min. Butane combustion data were averaged every 10 s to determine the precision of the metabolic cart, and the coefficient of variation (CV) determined for a set of 5–6 consecutive averages with steady ventilation. The ventilation was considered steady when consecutive values differ by less than 10%. Some variability in ventilation was unavoidable since the air was pumped manually, although always by the same operator.

Data are reported as the mean (±SD), unless otherwise stated. Values were checked for normal distribution using the Shapiro-Wilks test. The agreement between methods was assessed by determining the bias in absolute values and as a percentage of the measured value with the reference method (Vyntus) and the corresponding limit of agreement upper limit of agreement (ULA) = bias + 1.96 × SD; lower limit of agreement (LLA) = bias – 1.96 × SD. In addition, the agreement between methods was further examined by determining the Concordance Correlation Coefficient (CCC; Lin, 1989). The accuracy was evaluated by determining the differences between the mean values obtained with each method using Student t-test. The reliability of COSMED K5 measurements was assessed by determining the CV as described by Forkman (2007). Finally, the relationship between VO₂ and VCO₂ during butane combustion at different rates was assessed by the correlation coefficient of Pearson using 10-s averaged data. A P-value ≤ 0.05 was considered to be statistically significant. All statistical analyses were performed using IBM SPSS v.21.0 for Mac Computers (IBM, New York, United States).

We used three different levels of combustion to simulate low, medium, and very high exercise intensities (Supplementary Table 1). The corresponding VO₂ values were 0.788, 1.314, and 6.395 L/min and the associated RQ values 0.624, 0.608, and 0.619, respectively. The RQ deviation from the theoretical value for the combustion of butane (0.615) were 1.5, -1.1, and 0.7%, respectively. The CV of RQ for 5–6 consecutive 10 s averaged-intervals with similar ventilation was always below 0.5%. VO₂ during butane combustion was closely related to VCO₂ (r = 0.9999).

During measurements at rest and 60 W Vyntus and COSMED K5 (B×B) reported similar mean VO₂, VCO₂, RER, and energy expenditure values (Supplementary Tables 2, 3 and Figures 1–4). Nevertheless, compared to Vyntus, COSMED K5 operated in mixing chamber mode (Mix) overestimated VO₂ by 13.4 and 5.8%, at rest and 60 W, respectively (Supplementary Table 2). Both Vyntus and COSMED K5 (Mix) reported similar V_E and VCO₂ values. Consequently, COSMED K5 (Mix) underestimated RER by 13.8 and 5.8%, at rest and at 60 W, respectively.

At the highest intensity (130 W and 160 W in women and men, respectively) COSMED K5 (B×B) underestimated VO₂ and VCO₂ by 6.6 and 6.9%, respectively, due to a 3.0% underestimation of V_E (B×B) combined with 1.8% F_EO₂ overestimation and 5.1% F_ECO₂ underestimation by COSMED K5 (B×B) (Supplementary Table 3 and Figures 5, 6). This resulted in a 6.6% underestimation of energy expenditure by COSMED K5 (B×B). Nevertheless, there was an excellent agreement between methods in RER and fat oxidation, while COSMED underestimated carbohydrate oxidation by 6.9% (Supplementary Table 3).

COSMED K5 (Mix) overestimated VO₂ during exercise by 5.8 and 4.8%, at 60 W and the highest intensity, respectively. Consequently, COSMED K5 (Mix) overestimated energy expenditure by 4.1%. Since VCO₂ was accurately measured during exercise by COSMED K5 (Mix), COSMED K5 (Mix) underestimated the RER by 5.8 and 4.8%, at 60 W and the highest intensity, respectively, causing a marked overestimation of fat oxidation (Supplementary Table 3).

Compared to B×B, the mixing chamber mode overestimated VO₂ by 12.7% at rest (Supplementary Table 4). This bias in the assessment of resting VO₂ is mostly explained by a 10.4% V_E overestimation in Mix mode. However, VCO₂ was accurately measured due to the compensation of the overestimation of V_E by an underestimation of F_ECO₂ in Mix mode. Consequently, the RER was underestimated by 11.8% and energy expenditure by 10.1% in Mix mode.

During exercise the Mix mode the K5 overestimated VO₂ by 9.9 and 12.3% at 60 W and the highest load, respectively (Supplementary Table 5). This was explained by a 2.2 and 2.4% F_EO₂ underestimation, respectively. At 60 W, there was no bias in the assessment of VCO₂, but it was overestimated by 7.1% at the highest intensity by the Mix mode, resulting in 7.5 and 4.7% underestimation of the RER at 60 W and the highest load, respectively. Consequently, the Mix mode overestimated energy expenditure by 8.3 and 11.5% at 60 W and highest load, respectively.

The COSMED K5 had excellent reliability during the field tests (Supplementary Table 6), despite small changes in walking speed, starting weight, and environmental conditions (Supplementary Table 7). Total energy expenditure per Km was determined with a CV for repeated measurements of 4.5% (CI: 3.2–6.9%) and a CCC of 0.91, similar to the variability of VO₂. This high reproducibility was explained by the low variation of F_EO₂ measurements, which had a CV of 0.9% (CI: 0.7–1.5%) combined with a slightly greater variability of F_ECO₂, V_E, VCO₂, and RER.

The precision of COSMED K5 was determined by using the mixing chamber mode during prolonged low intensity exercise in the field. Our data demonstrate that the COSMED K5 can assess VO₂ and energy expenditure during prolonged walking with a CV below 5%, i.e., performing at a similar level of precision as stationary metabolic carts (Carter and Jeukendrup, 2002; Crouter et al., 2006). It should be taken into consideration that this low CV was obtained despite small changes in body weight, environmental conditions, walking speed, and weight loss during the walks. Thus, even lower CVs may be achievable with the K5 under better controlled environmental conditions as during test carried out in laboratories. This level of precision exceeds that reported for its predecessor the K4B² for repeated walking measurements in separate days during treadmill walking (Darter et al., 2013; Howe et al., 2014) as well as during incremental exercise (Schrack et al., 2010; Brisswalter and Tartaruga, 2014).

Combustion of fuels like ethanol (Cooper et al., 1991; Rising et al., 2015), methanol (Damask et al., 1982; Miodownik et al., 1998), propane (Melanson et al., 2010; Rising et al., 2015; Ismail, 2017), and butane (Nunn et al., 1989; Joosten et al., 2000) are considered gold-standard methods to validate indirect calorimeters. A common feature of precedent propane and butane combustion studies has been the use of only low combustion rates. In the present investigation we have used a similar methodology but using a broad range of combustion rates eliciting VO₂ values spanning across the physiological VO₂ values reachable by humans with varied fitness levels. In addition, our simulation generated respiratory variables mostly within the physiological range for humans. Importantly, our data indicate that the Vyntus CPX metabolic cart is exceptionally accurate and precise measuring the stoichiometric RQ of butane combustion. This is only possible when VO₂ and VCO₂ are computed correctly from the inspired and expired O₂ and CO₂ fractions, even at high breathing frequencies (highest butane combustion rate) and despite differences in incurrent and excurrent airstreams (see Supplementary Table 1).

In summary, our data indicate that the COSMED K5 is an excellent portable metabolic cart which is almost as accurate as a state-of-art stationary metabolic cart, capable of measuring precisely energy expenditure in the field and showing a reliable performance during more than 2 h of continuous work. The mixing-chamber option allows for more accurate assessment of VO₂ during exercise at relatively higher intensities, but is less accurate at low exercise intensities, and should not be used to measure resting energy expenditure.