Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Validity, reliability and minimum detectable change of COSMED K5 portable gas exchange system in breath-by-breath mode

  • Laura Guidetti,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Validation, Visualization, Writing – original draft

    Affiliation Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy

  • Marco Meucci,

    Roles Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft

    Affiliation Vascular Biology and Autonomic Studies Laboratory, Appalachian State University, Boone, North Carolina, United States of America

  • Francesco Bolletta,

    Roles Data curation, Formal analysis, Methodology

    Affiliation Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy

  • Gian Pietro Emerenziani,

    Roles Conceptualization, Data curation, Validation, Writing – review & editing

    Affiliation Department of Experimental and Clinical Medicine, University of Magna Græcia of Catanzaro, Catanzaro, Italy

  • Maria Chiara Gallotta,

    Roles Conceptualization, Methodology, Writing – review & editing

    Affiliation Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy

  • Carlo Baldari

    Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Writing – review & editing

    Affiliations Department of Movement, Human and Health Sciences, University of Rome “Foro Italico”, Rome, Italy, eCampus University, Novedrate (Como), Italy

Validity, reliability and minimum detectable change of COSMED K5 portable gas exchange system in breath-by-breath mode

  • Laura Guidetti, 
  • Marco Meucci, 
  • Francesco Bolletta, 
  • Gian Pietro Emerenziani, 
  • Maria Chiara Gallotta, 
  • Carlo Baldari



This study aimed to examine the validity, reliability and minimum detectable change (MDC) of the Cosmed K5 in breath by breath (BxB) mode, against VacuMed metabolic simulator. Intra and inter-units reliability was also assessed.


Fourteen metabolic rates (from 0.9 to 4 L.min-1) were reproduced by a VacuMed system and pulmonary ventilation (VE), oxygen consumption (VO2) and carbon dioxide production (VCO2) were measured by two different K5 units. Validity was assessed by ordinary least products (OLP) regression analysis, Bland-Altman plots, intraclass correlation coefficients (ICC), mean percentage differences, technical errors (TE) and MDC for VE, VO2, and VCO2. Intra- and inter-K5 reliability was evaluated by absolute percentage differences between measurements (MAPE), ICCs, TE, and MDC.


Validity analysis from OLP regression data and Bland- Altman plots indicated high agreement between K5 and simulator. ICC values were excellent for all variables (>0.99). Mean percentage differences in VE (-0.50%, p = 0.11), VO2 (-0.04%, p = 0.80), and VCO2 (-1.03%, p = 0.09) showed no significant bias. The technical error (TE) ranged from 0.73% to 1.34% (VE and VCO2 respectively). MDC were lower than 4% (VE = 2.0%, VO2 = 3.8%, VCO2 = 3.7%). The intra and inter K5 reliability assessment reveled excellent ICCs (>0.99), MAPE <2% (no significant differences between trials), TE < or around 1%, MDC <or around 3%.


K5 in BxB mode is a valid and reliable system for metabolic measurements. This is the first study assessing the MDC accounting only for technical variability reporting intra- and inter-units MDCs <3.3%.


The use of automated metabolic systems to measure oxygen consumption (VO2) and carbon dioxide production (VCO2) has become an essential tool for the analysis of physical performance and clinical diagnoses. Over the last three decades, the development of technology has facilitated the transition from laboratory to field measurements by introducing a variety of portable systems able to measure the energy cost of outdoor activities [1, 2].

COSMED recently launched a new portable metabolic system, (Cosmed K5) culminating in significant hardware, firmware and software improvements from the previous model. The K5 is a single unit device (174×64×114 mm dimensions and ~900 g weight) combining breath-by-breath (BxB) technology from the COSMED K4b2 and the dynamic mixing chamber system used in the COSMED Fitmate series. This option, called ‘IntelliMET’ (Intelligent Dual Metabolic Sampling Technology), allows users to select either the dynamic mixing chamber or the BxB sampling modality to measure either steady-state metabolic rates or oxygen kinetics during transients. This technology is supported by a series of significant hardware and firmware/software updates that aim to improve the reliability of its’ measures: 1) a dynamic mixing chamber that uses a constant flow pump; 2) a 4th generation opto-electronic reader and high performance turbine flowmeter with 0.08–16 L/s flow range; 3) an external scrubber to obtain real zero carbon dioxide and allow for more accurate gas calibration; 4) an external ambient temperature sensor for the calculation of the inspiratory BTPS factor and a capacitive ambient humidity and piezo-resistive pressure sensors inside the K5 unit for the calculation of the expiratory BTPS and STPD factors. Additional functions have been included to improve flexibility and durability of the product such as a 3.5″ TFT back-lit LCD touch-screen; a 4h Li-ion “smart battery”, an integrated 10 Hz GPS receiver for navigation/motion, integrated ANT+ technology for optional wireless sensors, a weatherproof case (IP54 standard), a standard or long-range Bluetooth 2.1 and an SDHC card for additional data storage [2].

The aim of the present study is to evaluate the validity and reliability of the COSMED K5 with BxB measurements, testing two different portable units over a wide range of metabolic rates. Douglas bag methods have been used to determine accuracy and precision of cardiopulmonary exercise testing (CPET) equipment, however previous studies clearly state limits to this method. Inherent biological variability disproportionately contributes to the overall error with Douglas bags whereas only a small part of the variability is caused by the measurement itself [3, 4]. On the other hand, metabolic simulators [5] are able to remove the biological variability and isolate measurement errors by systematically reproducing the human breath [6, 7]. Preliminary data produced in our laboratory, published as congress abstracts, suggest that the system is adequately reliable and valid when compared against a criterion VacuMed metabolic simulator [8, 9], however, a systematic validation study is necessary.


The study was conducted within the Department of Health Sciences at the University of Rome “Foro Italico”. COSMED and VacuMed were not involved in designing the study, data collection, analysis, interpretation or preparation of the manuscript.


Two COSMED K5 units, s/n 2015060002 (K5_02) and 2015060018 (K5_18), were used in this study. The K5 system uses a galvanic fuel cell and a non-dispersive infrared sensor for the analysis of oxygen (O2) and carbon dioxide (CO2) in the inhaled and exhaled air and an opto-electronic reader with a high performance turbine flowmeter to measure flow rate. After 30 minutes of warm up; flowmeter, gas, scrubber and delay time calibrations were performed following manufacturer’s recommendations. The two-point gas calibration was completed sampling the ambient air and the gas from a certified tank containing 16% O2, 5% CO2 and standard atmospheric Nitrogen. A 0% CO2 sampling was performed using a CO2 scrubber to obtain an accurate 0% CO2 reading and adjust for the CO2 and O2 values in the atmospheric air. Flowmeter calibration was performed connecting the turbine to a calibrated Hans Rudolph 3-liter syringe and completing six full strokes at a respiratory frequency of 20–25 b/min. Delay time calibration was performed with the flowmeter and the sampling line connected to the face mask and by executing six breaths at a given rhythm while breathing in the facemask.

VacuMed automated system

A commercially available metabolic simulator, the VacuMed automated system model #17056 (VacuMed, USA), was used in this study. This system uses a motor-drive syringe able to vary tidal volume and respiratory frequency to reproduce different ventilations (VE), and a gas tank containing air with the 79% N2 and 21% CO2 that is used to reproduce the VO2 and VCO2 in the exhaled air at different metabolic rates by diluting the gas tank with room air inside the piston pump [6]. Therefore, VO2 and VCO2 are proportional to the gas flow of calibration gas from the tank. Simulated volumes have been automatically corrected by the VacuMed software that compensated for temperature, barometric pressure and humidity measured in room air since VO2 and VCO2 are expressed in STPD while the simulator system utilizes known mixtures of a dry tank gas with a partially humidified room air [10]. The manufacturer certifies a system accuracy of 1% for simulated VO2 and VCO2 and a 0.25% for tidal volume.

Study design

The flowmeter and sampling line of each COSMED K5 system were connected directly to the outlet of the VacuMed automated system. Fig 1 shows a schematic of the system.

The K5_02 and K5_18 were tested separately and on different days under similar atmospheric conditions. The COSMED K5 units measured the volume and the concentrations of O2 and CO2 in each breath exhaled by the VacuMed system. Fourteen metabolic rates (from 0.9 to 4 L.min-1) were simulated by the VacuMed system and measured by the two K5 units. All tested metabolic rates are reported in Table 1.

Table 1. Protocol used to simulate 14 different metabolic rates (step) using VacuMed simulator.

Atmospheric pressure, ambient temperature, and relative humidity were measured by the K5 units before each test. Expired gases were sampled at the turbine through a semipermeable Nafion sampling line (0.75 m in length), and analyzed into the COSMED K5 portable units through an electro-galvanic fuel O2 cell and an infrared CO2 analyzer. All data were transmitted by Bluetooth from the portable unit to a personal computer and controlled in real time. Data from each metabolic rate were measured BxB for 70 s and the values were entered into a spreadsheet for later analysis. Raw data were reduced by removing the first 10 s of measurement to eliminate data related to the wash-out of the gas-filled dead space of the simulator, and performing a 60 s average of the remaining breaths. The accuracy and reliability of the K5 units were assessed for the main ventilatory and gas exchange variables: VE (L.min-1), VO2 (mL.min-1), and VCO2 (mL.min-1).

Statistical analyses


Agreement between the COSMED K5 and the VacuMed systems were assessed for VE, VO2, and VCO2 parameters by ordinary least products (OLP) regression analysis, which account for measurement error in both devices [11]. Regression parameters (slope and intercept), coefficients of determination (R2), and 95% confidence intervals (95% CI) were calculated for the OLP regression equations to determine fixed and proportional biases. The 95% confidence intervals containing the value 1 for the slope and the 0 for the intercept allows rejecting the hypothesis of proportional and fixed differences respectively. Bland-Altman plots [12] were constructed to determine the 95% limits of agreement (LoA) between the COSMED K5 and the VacuMed systems. Intraclass correlation coefficients (ICC) were used as parameters for criterion validity of the Cosmed K5 compared to the VacuMed simulator. A single measure, two-way random model, type absolute intra-class correlation coefficient was used to calculate ICCs. The strength of criterion reliability for ICC was classified in accordance with Hopkins (2000) [13].

Lastly, accuracy was quantified as the percentage differences (error) between the COSMED K5 s and VacuMed simulator [100*(COSMED K5-VacuMed)/VacuMed] and reported as mean and range values. COSMED K5 validity was also assessed by comparing the measured VE, VO2, VCO2 values vs simulated values with a paired samples t-test. Measurement error was expressed in “typical percentage error” (TE) and “minimum detectable change” (MDC). Typical error was calculated by dividing the standard deviation of the difference score by √2. This typical percentage error is a coefficient of variation and is considered highly reliable if less than 5% [13]. MDC values [also referred to as the “smallest detectable difference (SMD)], which reflects the magnitude of change necessary to provide confidence that the change was not resultant of random variation or measurement error, were calculated as 1.96*√2*TE.


To verify intra- and inter-K5 reliability the ICCs were determined on the 14 simulated metabolic rates measured twice by the same system or by two different K5 systems, respectively. A single measure, two-way mixed model, type absolute intra-class correlation coefficient was used to calculate ICCs [13, 14]. Intra and inter COSMED K5 system differences were quantified as the absolute percentage differences between measurements of the same K5 or between two different K5 systems, respectively. Due to the lack of a reference system, percentage difference was calculated as absolute percentage difference divided by the average intra or inter-system values and multiplied by 100. COSMED K5 BxB reliability was also assessed by comparing both intra and inter-system measures of VE, VO2, VCO2 with a paired samples t-test. Measurement error of intra- and inter-K5 systems was expressed in TE and MDC and calculated as reported in the validity section.

Statistical analyses were performed using the SPSS software package version 24.0 (SPSS Inc., Chicago, IL, USA), with a significance level set at p < 0.05.



Table 2 details R2, parameters of the OLP regression equation (slope and intercept) and the mean percentage of the difference between the values generated by the VacuMed simulator and measured by the K5. The agreement between values generated by the VacuMed simulator and measured by the K5 for the main gas exchange variables is presented in Table 2.

Table 2. Agreement between values generated by the VacuMed simulator and values measured by the K5 as assessed by OLP regression analysis.

VacuMed and K5 values demonstrated high correlation (R2 > 0.99; ranging from 0.9969 to 0.9995) in VE, VO2 and VCO2 variables. OLP regression analysis reported slope and intercept values that always include the 1 and the 0, respectively. ICC values were excellent for all variables (>0.99). The OLP regression analysis and the Bland-Altman plots of the averaged VE, VO2, and VCO2 values obtained during the 14 simulated metabolic rates between the Simulator (VacuMed) and K5 (BxB modality) are graphically shown in Fig 2A, 2B and 2C respectively.

Fig 2. Regression and difference plots of pulmonary ventilation (VE), oxygen uptake (VO2) and carbon dioxide production (VCO2) measured by K5 and generated by the VacuMed simulator.

For each one of the 3 main graphs is the OLP regression plot (the outside panel), with the linear regression (solid line), the identity (dashed line) and the equation with the Pearson’s determinant coefficient (R2), and the Bland-Altman plot (upper-left panel) with the mean difference (solid lines) and the 95% CI (dashed lines).

Mean percentage differences in VE (-0.50%, p = 0.11), VO2 (-0.04%, p = 0.80), and VCO2 (-1.03%, p = 0.09) showed no significant bias (Table 2). The typical percentage error (TE) ranged from 0.73% to 1.34% (VE and VCO2 respectively). MDC was lower than 4% (VE = 2.0%, VO2 = 3.8%, VCO2 = 3.7%).


ICC was excellent (>0.99) in all conditions (Table 3). For all variables, the mean absolute percentage error (MAPE), calculated intra- and inter- K5 system were around 1% with the 95% CI values below 3%, and no significant differences were found between trials. The typical percentage error was below or around 1%. The MDC values were similar or slightly lower in the intra- (VE = 2.0%, VO2 = 2.3%, VCO2 = 2.6%) when compared to the inter-K5 system reliability (VE = 2.0%, VO2 = 3.2%, VCO2 = 3.3%).


The first aim of this study was to test the accuracy of the COSMED K5 portable metabolic measurement system using the BxB setting against the criterion VacuMed simulator. The second aim assessed the intra- and inter-K5 system reliability. To our knowledge, this is the first study assessing accuracy and reliability of the K5 portable metabolic system in comparison with a gas exchange simulator. However, metabolic simulators have been widely used in the past to assess the validity of different metabolic systems [1517].


The results indicate high agreement between the K5 measurements and the simulated values over a wide range of simulated exercise intensities (VO2 up to 4 L.min-1). The OLP regression equations indicated that neither fixed or proportional biases were present. All ICCs and 95%CI values showed excellent agreement > 0.99. The OLP regression analysis revealed no proportional or fixed differences between measured (K5) and simulated (VacuMed) in all variables (VE, VO2, VCO2). Measurement differences were within ±4% and in agreement with the suggested range of differences for VO2 [18]. Mean differences were less than -1% for VE (-0.5%) and VO2 (-0.04%) and -1% for VCO2, with no statistical significance between measured and simulated values. Previous studies reported higher differences when comparing simulated and measured values when the K4b2 (VE 4.2%, VO2 3.6%, VCO2−2.2%) (16), the Quark CPET or an automated on-line system (4–12% range) were used [15, 17]. Moreover, our results reported a typical percentage error of 0.7% for VE and 1.4% and 1.3% for VO2 and VCO2, respectively, which is lower than the < 3% VO2 and <5% VE recommended by Hodges at al. (2005) [19]. Even in the case when it could be extremely likely that these two reference percentages were erroneously attributed, (VE < 3% and VO2 < 5% since VO2 derives from a calculation that includes VE and FO2; [20, 21], these results will still be considered below the acceptable percentage of error. We hypothesized the 1% error we observed in VE may depend on the small temperature difference between simulated inspired and expired air as these are close to room temperature, contrasted by in vivo measurements where the expired temperature is set by default to 34°C. This may suggest that a real temperature measurement of expired air could positively influence the accuracy of VE values during in vivo measurements, even though it has been noted that a 1.0°C difference in the estimated expiratory temperature from the actual temperature would result in only a 0.6% error in VE having only a minor effect on the calculation of VO2 [22].

To our knowledge, previous studies used the MDC to evaluate the reliability of a test protocol and a metabolic system [23, 24] but it has never been used versus a criterion system (simulator) to quantify the minimum change attributing the difference to the measurement error and not to the result of random variation. If the difference between a single measurement and a criterion is smaller than the smallest detectable change, it is likely due to measurement error while any difference larger than the MDC should be considered as real difference. In our study, the MDC of the simulator for VE (2%) was lower than the accuracy limits reported by ATS/ERS guidelines for spirometry of 3.5% [21]. Moreover, most certification bodies tolerate a maximal error of 4% in VO2 [19, 25] and a difference <5% versus the reference method would be considered as acceptable [1]. In our study the MDC in VO2 (3.8%) and VCO2 (3.7%) were lower than the reference values from literature for VO2 and reflect the low MDC value in VE (2%). Lastly, the low MDC values reported in VO2 and VCO2 reflect the low MDC in VE, since that the accuracy of VO2 and VCO2 in the BxB calculation is influenced by both VE and time delay errors [17].


K5 in breath-by-breath mode showed an excellent intra- and inter- device reliability in VO2 and VO2 with ICCs >0.99 and a MAPE <1.5%. In previous studies the ICCs test-retest reproducibility ranged between 0.90 to 0.97 for stationary metabolic carts [26, 27], and between 0.88 to 0.95 for portable systems such as K4b2 [28]. The highest ICCs were obtained for mixing chamber systems (0.98 and 0.98, VO2 and VCO2 respectively) and stationary breath by breath apparatus (0.97 and 0.96, O2 and VCO2, respectively) [29].

Studies that assessed test-retest variability in humans, obtained MAPE and coefficient of variation (CV) of 1.8 to 7.4% for VO2 and 4.1 to 7.7% for VCO2 including both technological errors and biological fluctuations [15, 2628]. To our knowledge, the intra- and inter-unit technological variability was previously evaluated by connecting a metabolic system to a gas simulator [30] or attaching two devices to the exercising subject for simultaneous sampling [31]. These studies showed low intra-unit and inter-unit variability with relative percentage errors < 2% for VE, VO2 and VCO2 and TEM< 1.5%, and MAPE = 2.1% for VO2 and a CV = 1.5%, respectively. These results are similar or slightly higher than those obtained from our intra- and inter-device TEM (<1% and <1.2% for all VE, VO2 and VCO2), comparable to the 1% relative error generated from an automated calibration system [7] and considerably below the TEM reliability limit of 3% recommended by the Australian Sports Commission [32].

Previous reliability studies reported an MDC of 7–10% using human subjects which included both technical and biological variability [23, 33]. This is the first study assessing the MDC accounting only for technical variability. The inter-unit MDC of the K5, representing the smallest change detectable by the instrument beyond the variability of the technical measurement, was low (2 to 3.3%) and similar to the 2–2.6% intra-unit MDC for VE, VO2 and VCO2.

Limitations of this study are represented by the limited range of simulated metabolic rates and by the use of a gas exchange simulator. Some caution should be taken for metabolic rates lower and higher than those used in this study. Further, a physiological scope outside of the parameters we tested will affect the generalizability of these results. Moreover, despite the clear advantages of using a gas exchange simulator (e.g., exclusion of biological variability and reliable simulation of gas volumes in a wide range of measurements), limitations are represented by the production of ambient temperature and dry gases only mathematically corrected using the manufacturer’s software [10]. Despite these limitations, this study was able to assess validity and reliability of the COSMED K5 using BxB mode while only accounting for the variability caused by technical errors.


The COSMED K5 in BxB mode is a valid system for the measurement of VE, VO2 and VCO2 for a wide range of metabolic rates as indicated by: a) the absence of systematic and proportional errors, b) the very high ICCs and excellent agreement c) a typical percentage error lower than 1.5%, d) the low MDC of 2–3.8% versus the reference system. The K5 was also found to be a reliable system as shown by the very high ICCs and the low intra and inter-device variability (TEM < 1.2%). Moreover, the low intra- and inter-device MDC (<2.6% and <3.3%, respectively) in repeated measurements is useful to discriminate measurement error from true change.

Supporting information

S1 File. Supporing information file.

VCO2_sim = simulated carbon dioxide production; VCO2_K5_18_2 = carbon dioxide production measured by K5 unit 2015060018 as first measurement; VO2_sim = simulated oxygen uptake; VO2_K5_18_2 = uptake measured by K5 unit 2015060018 as first measurement; VE_sim = simulated pulmonary ventilation; VE_K5_18_2 = pulmonary ventilation measured by K5 unit 2015060018 as first measurement; VE_K5_18_1 = pulmonary ventilation measured by K5 unit 2015060018 as second measurement; VO2_K5_18_1 = oxygen uptake measured by K5 unit 2015060018 as second measurement; VCO2_K5_18_1 = carbon dioxide production measured by K5 unit 2015060018 as second measurement; VE_K5_02_1 = pulmonary ventilation measured by K5 unit 2015060002 as first measurement; VE_K5_02_2 = pulmonary ventilation measured by K5 unit 2015060002 as second measurement; VO2_K5_02_1 = oxygen uptake measured by K5 unit 2015060002 as first measurement; VO2_K5_02_2 = oxygen uptake measured by K5 unit 2015060002 as second measurement; VCO2_K5_02_1 = carbon dioxide production by K5 unit 2015060002 as first measurement; VO2_K5_02_2 = carbon dioxide production measured by K5 unit 2015060002 as second measurement.



  1. 1. Overstreet BS, Bassett DR Jr., Crouter SE, Rider BC, Parr BB. Portable open-circuit spirometry systems. J Sports Med Phys Fitness. 2017; 57(3): 227–37. pmid:26861831
  2. 2. Macfarlane DJ. Open-circuit respirometry: a historical review of portable gas analysis systems. Eur J Appl Physiol. 2017; 117(12): 2369–86. pmid:29043499
  3. 3. Shephard RJ. A critical examination of the Douglas bag technique. J Physiol. 1955; 127(3): 515–24. pmid:14368545
  4. 4. Katch VL, Sady SS, Freedson P. Biological variability in maximum aerobic power. Med Sci Sports Exerc. 1982; 14(1): 21–5. pmid:7070252
  5. 5. Bunn JA, Pittsley JL, Baker SV, Yates J. Assessment of accuracy of the Vacu-Med 17053 calibrator for ventilation, oxygen uptake (V(O(2))), and carbon dioxide production (V(CO(2))). Respir Care. 2011; 56(4): 472–6. pmid:21255508
  6. 6. Huszczuk A, Whipp BJ, Wasserman K. A respiratory gas exchange simulator for routine calibration in metabolic studies. Eur Respir J. 1990; 3(4): 465–8. pmid:2114308
  7. 7. Gore CJ, Catcheside PG, French SN, Bennett JM, Laforgia J. Automated VO2max calibrator for open-circuit indirect calorimetry systems. Med Sci Sports Exerc. 1997; 29(8): 1095–103. pmid:9268968
  8. 8. Baldari C, Meucci M, Bolletta F, Gallotta MC, Emerenziani GP, Guidetti L. Accuracy and reliability of COSMED K5 portable metabolic device versus simulating system. Sport Sci Health. 2015; 11(1): 58.
  9. 9. Bolletta F, Meucci M, Emerenziani GP, Gallotta MC, Guidetti L, Baldari C. Accuracy and reliability of the K5 portable metabolic system in micro-mixing chamber mode using a gas exchange simulator system. Sport Sci Health. 2016; 12(1): 34.
  10. 10. USA V. Operating & service manual; Metabolic Calibrator / Simulator 17057. Serial number 1135–001. 2013.
  11. 11. Ludbrook J. Linear regression analysis for comparing two measurers or methods of measurement: but which regression? Clin Exp Pharmacol Physiol. 2010; 37(7): 692–9. pmid:20337658
  12. 12. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999; 8(2): 135–60. pmid:10501650
  13. 13. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med. 2000;30(1):1–15. pmid:10907753
  14. 14. McGraw KO, Wong SP. Forming inferences about some intraclass correlations coefficients (vol 1, pg 30, 1996). Psychol Methods. 1996; 1(4): 390–.
  15. 15. Carter J, Jeukendrup AE. Validity and reliability of three commercially available breath-by-breath respiratory systems. Eur J Appl Physiol. 2002; 86(5): 435–41. pmid:11882930
  16. 16. Rodriguez FA, Keskinen KL, Kusch M, Hoffmann U. Validity of a swimming snorkel for metabolic testing. Int J Sports Med. 2008; 29(2): 120–8. pmid:17960507
  17. 17. Beijst C, Schep G, Breda E, Wijn PF, Pul C. Accuracy and precision of CPET equipment: a comparison of breath-by-breath and mixing chamber systems. J Med Eng Technol. 2013; 37(1): 35–42. pmid:23110656
  18. 18. Macfarlane DJ. Automated metabolic gas analysis systems: a review. Sports Med. 2001; 31(12): 841–61. pmid:11665912
  19. 19. Hodges LD, Brodie DA, Bromley PD. Validity and reliability of selected commercially available metabolic analyzer systems. Scand J Med Sci Sports. 2005; 15(5): 271–9. pmid:16181250
  20. 20. Taylor JR. An Introduction to Error Analysis. 2nd ed. Sausalito: University Science Books; 1997.
  21. 21. Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, et al. Standardisation of spirometry. Eur Respir J. 2005; 26(2): 319–38. pmid:16055882
  22. 22. Bassett DR Jr., Howley ET, Thompson DL, King GA, Strath SJ, McLaughlin JE, et al. Validity of inspiratory and expiratory methods of measuring gas exchange with a computerized system. J Appl Physiol (1985). 2001; 91(1): 218–24.
  23. 23. Darter BJ, Rodriguez KM, Wilken JM. Test-retest reliability and minimum detectable change using the K4b2: oxygen consumption, gait efficiency, and heart rate for healthy adults during submaximal walking. Res Q Exerc Sport. 2013; 84(2): 223–31. pmid:23930548
  24. 24. Thomas SS, Buckon CE, Schwartz MH, Russman BS, Sussman MD, Aiona MD. Variability and minimum detectable change for walking energy efficiency variables in children with cerebral palsy. Dev Med Child Neurol. 2009; 51(8): 615–21. pmid:19627334
  25. 25. Babineau C, Leger L, Long A, Bosquet L. Variability of maximum oxygen consumption measurement in various metabolic systems. J Strength Cond Res. 1999; 13(4): 318–24.
  26. 26. Skinner JS, Wilmore KM, Jaskolska A, Jaskolski A, Daw EW, Rice T, et al. Reproducibility of maximal exercise test data in the HERITAGE family study. Med Sci Sports Exerc. 1999; 31(11): 1623–8. pmid:10589867
  27. 27. Meyer T, Davison RC, Kindermann W. Ambulatory gas exchange measurements—current status and future options. Int J Sports Med. 2005; 26 Suppl 1: S19–27.
  28. 28. Brisswalter J, Tartaruga MP. Comparison of COSMED'S FitMate and K4b2 metabolic systems reliability during graded cycling exercise. Scand J Clin Lab Invest. 2014; 74(8): 722–4. pmid:25375032
  29. 29. Meyer T, Georg T, Becker C, Kindermann W. Reliability of gas exchange measurements from two different spiroergometry systems. Int J Sports Med. 2001; 22(8): 593–7. pmid:11719895
  30. 30. Macfarlane DJ, Wong P. Validity, reliability and stability of the portable Cortex Metamax 3B gas analysis system. Eur J Appl Physiol. 2012; 112(7): 2539–47. pmid:22075643
  31. 31. Macfarlane DJ, Wu HL. Inter-unit variability in two ParvoMedics TrueOne 2400 automated metabolic gas analysis systems. Eur J Appl Physiol. 2013; 113(3): 753–62. pmid:22945269
  32. 32. Gore CJ. Physiological testing for elite athletes (Australian Sports Commission). Champaign: Human Kinetics; 2000.
  33. 33. Brehm MA, Knol DL, Harlaar J. Methodological considerations for improving the reproducibility of walking efficiency outcomes in clinical gait studies. Gait Posture. 2008; 27(2): 196–201. pmid:17467276