Abstract Rationale Obstructive sleep apnea (OSA) is characterized by patterns of respiratory events followed by hyper-ventilatory breaths, which frequently coincide with arousals. In literature, post-event hyperventilation is explained by a combination of the respiratory-event severity and consequent accumulated respiratory stimuli, arousal presence, and the magnitude/severity of arousal (when present). We hypothesize that the contributions to post-event ventilation of arousal presence and magnitude vary between REM and NREM sleep. Methods We retrospectively analyzed polysomnography data from 2055 participants of the Multi-Ethic Study of Atherosclerosis (MESA) study. Ventilation was quantified from nasal pressure and expressed in %Eupnea, where the eupneic ventilation was estimated as the mean ventilation in the 7-minute analysis window. For each obstructive respiratory event, post-event ventilation (VEPost-Event) was quantified as the mean of the first two breaths immediately after arousal onset or respiratory event termination. Respiratory-event severity was quantified as the event-specific ventilatory burden (VB=event duration × Eupneic Ventilation-Ventilation During Event) and standardized to have a zero mean and a standard deviation of one. Arousal was quantified by its presence (ArP, 0=absent, 1=present) and magnitude with Wake-Sleep-Balance (WSB), which was a weighted combination of the power in frequency bands of EEG and converted into deciles to provide an integer value between 0 and 9. Sleep states were represented as a binary variable (REM=1 and NREM=0).To investigate the modulatory effect of REM vs. NREM sleep on arousal-mediated ventilatory response, we employed the linear mixed-effects model with VEPost-Event ∼ArP+ArP*WSB+VB + (1|Participant) as the base model. We step-wise added fixed-effect sleep-state (SleepState) interaction terms: SleepState*ArP, SleepState*ArP*WSB, and SleepState*VB. Model fit was assessed using Akaike Information Criterion (AIC). Results 55255 respiratory events in REM and 160070 events in NREM were analyzed from 1781 participants (after excluding participants with AHI5 or inadequate signal quality). The best model (based on AIC, see Fig.1) was VEPost-Event ∼ArP+SleepState*ArP+ArP*WSB+SleepState*ArP*WSB+VB+SleepState*VB + (1|Participant). All terms were statistically significant. REM sleep was found to be associated with increased arousal-mediated ventilation: arousal presence contributed an additional 21.795%CI: 19.8, 23.6 %Eupnea in REM beyond NREM, and each decile increase in arousal magnitude added 1.00.7, 1.4 %Eupnea. Conversely, REM sleep attenuated the effect of VB (β=-7.5-8.1, -7.0 %Eupnea), consistent with lower chemosensitivity during REM sleep. Conlcusions Confirming our hypothesis, arousal contributions to post-event ventilation were found significantly greater in REM versus NREM sleep, independent of respiratory event severity. This could have clinical significance as therapies targeting arousal threshold may yield greater improvements in destabilizing post-event ventilation in individuals with REM-predominant OSA. This abstract is funded by: National Health and Medical Research Council (NHMRC)
Zhang et al. (2026) studied this question.