Methods Used to Process Data from Indirect Calorimetry and Their Application to VO2MAX
Computing, Health and Science
Biomedical and Sports Science, Centre for Exercise and Sports Science Research
Our purpose was to provide objective evidence in support of recommendations for how to process data in gas exchange indirect calorimetry (GEIC). A computer generated data set, devoid of biological variability and measurement error, was used to assess the error in data processing methods of a) 0.25, 0.5 and 1.0 min time averages, b) a 7 breath average, and c) a 11 breath smoothing (moving average) function. We aligned averaged data to the end and center of the time interval, and also used a breath-by-breath data set from an incremental exercise test to volitional fatigue to show the results of data processing using real data. Aligning time averaged data to the end of the interval under-estimated oxygen consumption (VO2) (mean±SD, mL/min) by 29.99±5.21 (0.25 min end), 65.09±6.40 (0.5 min end), and 134.20±3.48 (1 min end). Aligning time averaged data to the center of the interval resulted in slight over-estimation of VO2, with mean±SD (mL/min) errors of -12.61±5.10 (1.0 min), - 5.94±3.43 9 (0.5 min), and -3.41±4.01 (0.25 min). Results from the 7 breath averaged data revealed monoexponential decreases in error with increases in VO2, ranging from (rest to peak VO2, mL/min) 67.93 to 11.8 and 0.47 to 0.01, for end and center aligned data, respectively. Smoothing (11 breath) also resulted in monoexponential decreases in error with increases in VO2, with a range of residuals from (rest to peak VO2) 1.08 to 0.03 mL/min. Similar procedures were applied to examples of breath-by-breath data obtained from research subjects, and therefore data containing biological variability. In addition, data processing involving interpolation and filtering was applied to the human subjects acquired data. Averaging VO2 data from GEIC across 0.5 and 1.0 min time intervals adds to the total error of measurement in GEIC. Where possible, data averaging should occur by breath averaging (~7 breaths) aligned to the center of the interval. Data processing of breath-by-breath data using interpolation and filtering has potential, and required further research to assess validity.