Comparison of methods of chemical loop gain measurement during tidal ventilation in awake healthy subjects

The loop gain (LG) is defined as the ratio of a ventilatory response over the perturbation in ventilation, and it is used to analyze ventilatory control stability. The LG can be derived from minute ventilation (V̇e), end-tidal Pco 2 ([Formula: see text]), and end-tidal Po 2 ([Formula: see text]) values. Several methods of LG assessment have been developed, which have never been compared. We evaluated the computability, the short-term repeatability, and the agreement of six published (or slightly modified) models for LG determination. These models included three unconstrained autoregressive models, univariate (V̇e), bivariate (V̇e, [Formula: see text]), and trivariate (V̇e, [Formula: see text], and [Formula: see text]), and three analytical transfer function constrained models based on V̇e, V̇e and CO 2 -sensitivity, and V̇e and central and peripheral CO 2 sensitivities, respectively. The models were tested with tidal breathing data in 37 awake healthy subjects (median age 35 yr; 23 women, 14 men). Modeling failed in 11, 0, and 0 subjects for the three unconstrained models, respectively, and 4, 1, and 9 subjects for the three constrained models, respectively. Bland and Altman analyses of the LG values in the medium frequency range of two separate recordings demonstrated good repeatability for four models, excluding univariate and trivariate unconstrained models. The four repeatable models gave LG values that were in agreement (medium frequency LG, median 0.100–0.210), although the constrained model based on V̇e systematically overestimated LG values. The variances explained by these models were ∼20%. In conclusion, model-based analyses of tidal breathing were performed with different approaches that gave comparable results for chemical LG and explained variance. NEW & NOTEWORTHY Several methods of chemical loop gain measurement have been published but never compared. We show that a better repeatability is obtained with analytical constrained models compared with autoregressive unconstrained models and that the repeatable models gave comparable results of loop gain, even if the calculation based on ventilation-only recording gave higher values than those obtained with both ventilation and end-tidal Pco 2 recording. The explained variance of ventilation was similar whatever the model.