End-tidal CO₂ responses and ventilatory correlates during voluntary hyperventilation: a repeated-measures pilot study

Controlled voluntary hyperventilation is widely used to study the ventilatory control of end-tidal CO₂ (PetCO₂), but the breath-by-breath determinants of PetCO₂ during an externally paced maneuver remain incompletely characterized. This repeated-measures pilot study aimed to describe PetCO₂ changes and to quantify associations with ventilatory timing and expired gas output during paced voluntary hyperventilation.

Thirty healthy males (age range 19.2-39.8) performed a metronome-paced voluntary hyperventilation maneuver at 30 breaths/min intended to last up to 3 min. Breath-by-breath data were recorded with a COSMED K5 (COSMED, Rome, Italy). After excluding breaths with implausible respiratory frequency, 2,298 breaths from 30 participants were analyzed with a linear mixed model (LMM) using a cubic B-spline for elapsed time, participant-specific random intercepts and random slopes, and fixed effects of minute ventilation (VE), expired CO₂ output (VCO₂), and expiratory time (Te). A generalized estimating equation (GEE) with AR(1) working correlation was a sensitivity analysis, and participant-level cluster-bootstrap CIs provided the primary inferential intervals.

Participant-level mean PetCO₂ during the maneuver was 17.1 ± 1.8 mmHg (range 13.3-21.2 mmHg) and declined over time. In the primary LMM, VE β = -0.144, VCO₂ β = 0.009, and Te β = 4.18. Because residual diagnostics showed non-normality, heteroscedasticity, and strong serial autocorrelation (GEE AR(1) alpha approximately 0.79 for the spline model and 0.86 for the linear-time model), cluster-bootstrap CIs are reported as the primary inferential intervals: VE -0.174 to -0.110, VCO2 0.005 to 0.012, Te 2.330 to 6.164, and a predicted PetCO2 change from 0 to 180 s of -8.58 mmHg (95% CI -10.30 to -6.92 mmHg). Marginal R² = 0.775 and conditional R² = 0.828; leave-one-participant-out cross-validation yielded R² = 0.737 and RMSE = 2.52 mmHg.

PetCO₂ during paced hyperventilation was associated with elapsed time, VE, VCO₂, and Te. Because VE and VCO₂ are algebraically coupled, and VCO₂ and PetCO₂ also share information from the expired-CO₂ signal, they should not be interpreted as independent physiological determinants; the principal contribution is the within-participant expiratory-time association. The findings are exploratory and require confirmation in clinical populations.