Reversible Effect of Falling Ventilatory Drive in Drive-dependent OSA

Overview

Obstructive sleep apnea (OSA) is a highly prevalent disorder that has major consequences for cardiovascular health, neurocognitive function, risk of traffic accidents, daytime sleepiness, and quality of life. For years, a "classic" model of OSA has been used to describe the disorder, which fails to capture it's complexity. Recently, a model for OSA called drive-dependent OSA was discovered be more prevalent in the OSA population. The drive-dependent subgroup benefits exclusively from increased ventilation, increased dilator muscle activity, and reduced event risk when drive spontaneously rises. This study seeks to provide direct evidence that reducing the loss of drive prevents the loss of ventilation, pharyngeal muscle activity, and thus the onset of OSA respiratory events, specifically in "drive-dependent" but not "classic" OSA. This will be achieved using CO2 delivered at precise times during breaths in sleep to prevent loss of overall ventilatory drive.

This is a detailed physiological study, with gold-standard measurements of ventilatory drive and dilator muscle activity, we aim to mitigate falling drive with carefully-timed inspired CO2 administration in patients with (N=18) and without (N=18) drive-dependent OSA. We will test the hypothesis that OSA events are preventable by mitigating falling drive exclusively in drive-dependent OSA. Subjects will attend a virtual Screening and Consent visit to assess eligibility for enrollment. Participants will take part in a video call with the consenting doctor to obtain consent (Zoom). After consent, patients will first attend a baseline routine sleep study to confirm eligibility and establish baseline characteristics. Patients will subsequently attend an overnight physiology study with gold standard instrumentation (ventilation and ventilatory drive, see below) to establish OSA phenotype (presence/absence of drive-dependent OSA). Finally, an overnight physiological intervention study dedicated to mitigating ventilatory drive decline with carefully-timed inspired CO2 stimulation will be performed. At the Dynamic CO2 Study, we will also record upper airway dilator muscle activity (multiunit genioglossus electromyography, EMG). Once respiratory events begin, an investigator will switch the 4-way tap to deliver 2% inspired CO2 during the first 3-4 recovery breaths following a respiratory event. Consequently, end-tidal CO2 will no longer drop as abruptly to lower levels, and thus on the return to sleep, ventilatory drive will not decline as seen during the prior event. Sham interventions (medical air) will also be performed (1:1) to ensure that event risk is not lowered trivially due to spontaneous event resolution or deepening sleep with time. If necessary to precisely control end-tidal CO2 and ventilatory drive, we will 1) raise the magnitude to a maximum of 6%CO2 (in 14% O2) for a single breath intervention, or 2) lower the magnitude to 1.0% or 1.5% via dilution. The first 30 mins of each study will be dedicated to identifying the optimal dose, before formal interventions with alternating shams proceed. Data analysis Apneas, hypopneas, sleep stages and arousals from sleep will be scored using current AASM guidelines (hypopneas defined by at least a 30% reduction in airflow in conjunction with either 3% desaturation or arousal) by a technician blinded to the study condition. For each physiology night, breath-by-breath tables will be made describing ventilation, ventilatory drive, and genioglossus activity. Ensemble average analysis of breath-by-breath signals (ventilatory drive, ventilation, genioglossus EMG) will be used to examine the extent to which ventilation (∆Flow) and genioglossus EMG (∆EMGgg) rise on average during the subsequent "event period" with the ventilatory drive stimulation (∆Drive) compared to the prior reference event (red shading). The reduction in probability of a (clinically-scored) respiratory event will be recorded. The same variables will be calculated from using sham interventions. Statistical Analysis The quantitative primary outcome variable will be the continuous probability of a respiratory event after the active inspired CO2 intervention; the difference in this outcome variable between intervention and sham (i.e. sham-corrected effect) will be evaluated using mixed model logistic regression analysis (logit link function). The primary comparison will be whether the sham-corrected reduction in event likelihood is greater in drive-dependent OSA vs. classic OSA subgroups (per intervention × subgroup interaction, fixed effects). Subject will be included as a random effect. Results will be expressed as an odds ratio with 95%CI. P\<0.05 will indicate statistical significance. Secondary analysis will replace the binary intervention term (independent variable) with a continuous term describing the average increase in ventilatory drive with the intervention (∆Drive), allowing odds of event resolution to be expressed per unit increase in drive (per 100% eupnea) for each phenotype. This magnitude of effect will be contrasted with the magnitude observed in the prior observational analyses (odds ratio = 6.8 in drive-dependent OSA); similarity of magnitude will be taken as support for reversibility in the observational studies. Additional secondary analysis will replace the binary event likelihood term (dependent variable) with the increase in ventilation (∆Flow, linear model) and the increase in genioglossus EMG (∆EMGgg, linear model), which will enable us to quantify how much ventilation and dilator muscle activity is saved when drive decline is mitigated (for each OSA subgroup). Sensitivity analysis will examine whether arousal threshold modifies the efficacy of the drive intervention (intervention × arousal threshold interaction term). Power analysis Sample size is based on the primary outcome model: 36 patients will provide 90% power to detect a odds ratio of 3.0 for response to intervention in drive-dependent OSA vs. classic OSA subgroups (assuming sham-corrected probability of response in drive-dependent OSA = 0.63 \[0.33-0.60\], average \[95% prediction interval\] and 0.36 \[0.14-0.67\] in non-drive dependent OSA; N=1000 simulations, probabilities are conservative estimates based on the observational data). Power is also \>90% to detect a 20%eupnea greater increase in ventilation in drive-dependent OSA vs. non drive dependent OSA (uncertainty based on preliminary data). Power is \>80% to detect a 20% baseline greater increase in genioglossus EMG in drive-dependent OSA vs. non drive dependent OSA (uncertainty based on preliminary data).