Data collection, handling, and fitting strategies to optimize accuracy and precision of oxygen uptake kinetics estimation from breath-by-breath measurements

Abstract

Phase 2 pulmonary oxygen uptake kinetics (Φ2 tVo2P) reflect muscle oxygen consumption dynamics and are sensitive to changes in state of training or health. This study identified an unbiased method for data collection, handling, and fitting to optimize VO kinetics estimation. A validated computational model of Vo2P kinetics and a Monte Carlo approach simulated 2 × 10 moderate-intensity transitions using a distribution of metabolic and circulatory parameters spanning normal health. Effects of averaging (interpolation, binning, stacking, or separate fitting of up to 10 transitions) and fitting procedures (biexponential fitting, or 2 isolation by time removal, statistical, or derivative methods followed by monoexponential fitting) on accuracy and precision of Vo2P kinetics estimation were assessed. The optimal strategy to maximize accuracy and precision of tVo2P estimation was 1-s interpolation of 4 bouts, ensemble averaged, with the first 20 s of exercise data removed. Contradictory to previous advice, we found optimal fitting procedures removed no more than 20 s of Φ1 data. Averaging method was less critical: interpolation, binning, and stacking gave similar results, each with greater accuracy compared with analyzing repeated bouts separately. The optimal procedure resulted in Φ2 tVo2P estimates for transitions from an unloaded or loaded baseline that averaged 1.97 ± 2.08 and 1.04 ± 2.30 s from true, but were within 2 s of true in only 47-62% of simulations. Optimized 95% confidence intervals for tVo2P ranged from 4.08 to 4.51 s, suggesting a minimally important difference of ~5 s to determine significant changes in tVo2P during interventional and comparative studies.