Effect of interbreath fluctuations on characterizing exercise gas exchange kinetics

TLDR

Breathing irregularities produce breath‑to‑breath fluctuations that reduce the precision of nonsteady‑state gas‑exchange kinetics estimation during exercise. The study aimed to quantify how this noise affects confidence in estimating kinetic parameters and discriminating between models. Five subjects performed eight 0‑to‑100 W cycle‑ergometer transitions; breath‑by‑breath ventilation, CO₂ output, and O₂ uptake were interpolated, time‑aligned, averaged, and used to estimate first‑order kinetic parameters (time constant and delay) via linear least squares, nonlinear least squares, and maximum likelihood. Breath‑by‑breath noise behaved as an uncorrelated Gaussian process with a standard deviation largely independent of metabolic rate, and an expression was derived for the number of square‑wave repetitions needed to achieve a desired parameter confidence using nonlinear least squares and maximum likelihood, while linear least squares was less suitable.

Abstract

Breathing has inherent irregularities that produce breath-to-breath fluctuations (“noise”) in pulmonary gas exchange. These impair the precision of characterizing nonsteady-state gas exchange kinetics during exercise. We quantified the effects of this noise on the confidence of estimating kinetic parameters of the underlying physiological responses and hence of model discrimination. Five subjects each performed eight transitions from 0 to 100 W on a cycle ergometer. Ventilation, CO2 output, and O2 uptake were computed breath by breath. The eight responses were interpolated uniformly, time aligned, and averaged for each subject; and the kinetic parameters of a first-order model (i.e., the time constant and time delay) were then estimated using three methods: linear least squares, nonlinear least squares, and maximum likelihood. The breath-by-breath noise approximated an uncorrelated Gaussian stochastic process, with a standard deviation that was largely independent of metabolic rate. An expression has therefore been derived for the number of square-wave repetitions required for a specified parameter confidence using methods b and c; method a being less appropriate for parameter estimation of noisy gas exchange kinetics.