Data repeatability for the Cosmed K4 b2 portable metabolic cart during non-steady state exercise

BACKGROUND:Portable metabolic carts are a popular tool to assess aerobic capacity and affirm many cardiorespiratory conditions. They may also measure strength training performance. Given their popularity and increased usage to assess strength training performance; their data accuracy and consistency are important to determine. OBJECTIVE:Measure Cosmed K4 b2 portable metabolic cart data repeatability from consecutive seated calf press workouts. METHODS: Fifteen women and twelve men did two workouts that began with a stationary cycling warm-up followed by calf presses. Gases were measured before the calf press portion of workouts to establish baseline VO2 and VCO2 values, as well as continually throughout and after the calf press protocol. Subjects were detached from the cart once gas values returned to baseline after workouts concluded. In addition to VO2 and VCO2, repeatability was quantified for: breaths per minute, tidal volume, ventilation, O2 uptake relative to body mass, expired O2 and CO2 fractions, percent fat and carbohydrate utilization, METS and total energy cost. Mean and peak values per variable were analyzed. Repeatability was assessed separately for male and female data, as well as with values pooled, by the following: intraclass correlation coefficients, eta squared, limits of agreement, coefficient of variation and smallest real difference percent. RESULTS: Per variable, repeatability values across workouts were low. Female intraclass correlation coefficient mean values were more repeatable for variables related to gas measurements, yet male data were generally more repeatable for those related to substrate usage. CONCLUSIONS: Results for some repeatability indices were influenced by measurement magnitude. Peak values were predictably less repeatable than those for mean values. Most smallest real differences percent scores are so high they were rendered irrelevant or meaningless to determine true differences among paired values. Results suggest low data repeatability that are likely appropriate and realistic for the exercise protocol, hardware and intensity examined.