INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Obstructive sleep apnea (OSA) is characterized by recurrent episodes of pharyngeal obstruction which develop during sleep (1). To treat this disorder, methods have been devised to either relieve (2) or bypass (6) the site of pharyngeal airflow obstruction. Although tracheostomy remains the most effective method for bypassing the obstruction, it is seldom used owing to its attendant morbidity (7). Current literature, however, suggests that the morbidity of tracheostomy may be reduced when narrower transtracheal cannulae are employed (12). With such narrow cannulae, gas must be insufflated directly into the trachea for apneic patients to breathe. Nevertheless, early studies have demonstrated that insufflating a low flow of oxygen (2 to 6 L/min) only partially reduced both the frequency of apneic episodes and the severity of associated oxyhemoglobin desaturations (15).

The persistence of OSA during administration of low-flow transtracheal insufflation (TTI) may relate to the fact that patients relied solely on the TTI flow source for inspiratory airflow during periods of upper airway obstruction. In the presence of complete upper airway obstruction, low-flow (2 to 6 L/min) TTI may not have stabilized the breathing pattern during sleep because flow rates were insufficient to meet patients' inspiratory flow demand. If this were so, one might expect greater improvements in OSA when higher insufflation flow rates are used. Nevertheless, studies have not yet examined responses in apnea severity to higher flow rates.

In this study, we hypothesized that OSA would improve progressively as TTI flow was increased. To test this hypothesis, air was insufflated over a range from 0 to 15 L/min for a brief period of non-rapid eye movement (NREM) sleep while monitoring breathing patterns. In this first experiment, we established that higher TTI flow rates were required to abolish obstructive apneas, but were associated with repetitive laryngeal obstructions. A servo-controlled system was then developed for controlling the administration of high-flow TTI (hf-TTI) and implemented in a second experiment in which the polysomnographic responses to a prolonged period of hf-TTI administration were examined. Our findings indicate that this hf-TTI system may be effective in treating patients with OSA.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Tracheostomized patients referred to the Johns Hopkins Sleep Disorders Center were eligible for this study if they had clinically significant obstructive sleep apnea as defined by a NREM apnea-hypopnea index of greater than 20 episodes/h (with the tracheostomy capped) and were free of clinical and laboratory evidence of cardiorespiratory insufficiency (daytime hypercapnia or hypoxemia and evidence of right or left heart failure). Five such patients were recruited with baseline anthropometric and polysomnographic characteristics as summarized in Table 1. Tracheotomies had been placed because patients could not tolerate nasal continuous positive airway pressure (n = 4) or were refractory to treatment with 18 cm H₂O nasal continuous positive airway pressure (n = 1). Informed consent was obtained in each patient for a protocol that was approved by the Johns Hopkins Institutional Review Board.

Experimental Apparatus

Polysomnography. Standard polysomnographic monitoring was performed during all study protocols, and included monitoring of electroencephalograms (EEG) (C₃-A₂, C₃-O₁), left and right electro-oculograms (EOG), submental electromyogram (EMG), and electrocardiogram (ECG modified V2 lead). Oxygen saturation (Sa_O2) was also monitored (Biox 3700; Ohmeda Inc., Boulder, CO). Body position was monitored visually with infrared video cameras so that patients could be maintained supine throughout the protocol.

Respiratory monitoring. A Shiley tracheostomy tube was inserted before each transtracheal insufflation trial. An external cap was affixed to the tracheostomy tube, and sealed ports were made in the cap for inserting an insufflation cannula and a Luer stub adapter to monitor tracheal pressure. A standard Hyatt type esophageal balloon was passed perinasally into the midesophagus for esophageal pressure monitoring. In two patients, supraglottic pressure was also monitored with a fluid-filled catheter placed 16 cm from the nares. Catheter position in the supraglottic space was confirmed by nasopharyngoscopy. All pressures were monitored with Gould-Statham transducers.

Airflow through the upper airway was monitored either with a pneumotachometer (model 3700A; Hans Rudolph Inc., Kansas City, MO) affixed to a tight-fitting face mask or with an oronasal thermistor. The pneumotachometer was used in the first experiment to better elucidate breathing mechanics during short exposures to different levels of TTI airflow. An oronasal thermistor was employed in the second experiment to minimize any disturbance in sleep caused by wearing a full face mask, thereby permitting an assessment of polysomnographic responses to a prolonged period of TTI administration.

TTI administration. The apparatus for delivering compressed air, and controlling and monitoring respiratory responses to TTI administration is shown in Figure 1. An air compressor and flow regulator were used to insufflate air through an ~ 3 meter length of oxygen extension tubing (Baxter Inc., Valencia, CA) and a transtracheal catheter (SCOOP catheter, Transtracheal Systems Inc., Denver, CO). A mass flow meter (Matheson Inc., Secaucus, NJ) and a solenoid were connected in series to monitor the level of insufflated flow (IN), and direct flow to the patient or atmosphere. Tracheal pressure was continuously monitored, and the signal was digitized by a microcomputer. The computer controlled the direction of flow through the solenoid (Labview; National Instruments Inc., Austin, TX) based on the tracheal pressure level. When this level remained below a threshold, airflow was applied to the patient. When tracheal pressure exceeded this limit, TTI flow was diverted by the solenoid to atmosphere.

View larger version (13K): [in this window] [in a new window]   Figure 1.   Experimental setup during TTI. Air compressor generated flow which was applied to either the trachea via flow meter and transtracheal cannula or exhausted to atmosphere. Airflow through the upper airway ( UA) was monitored with a pneumotachometer and tight fitting facemask (Experiment 1) or a thermistor (Experiment 2). Tracheal pressure (Ptrach) is monitored by a microcomputer which controlled a solenoid valve based on the level of tracheal pressure (Experiment 2 only).

Data acquisition. All physiologic signals were amplified and recorded continuously on a polygraph recorder (Grass recorder; Astromed Inc., Warwick, RI). Signals from the analog amplifiers were also digitized at 100 Hz and stored on optical disk for off-line analysis (DI-200 A/D board and Windaq/200 software; Dataq Instruments, Akron, OH).

Experimental Protocols

Patients were monitored on three separate nights to determine responses to TTI titration (Experiment 1, n = 5) and to prolonged hf-TTI administration (Experiment 2, n = 4) as detailed subsequently.

Experiment 1Polysomnographic responses to TTI titration. On one night, TTI was administered at flows of 0, 5, 10, and 15 L/min in random order. For each TTI level, patients were first allowed to initiate at least 5 min of NREM sleep. Thereafter, TTI was applied for at least 10 min of NREM sleep. TTI was discontinued and the trial was aborted if the patient awoke for more than 3 min. Once the patient reinitiated sleep, TTI was restarted until a continuous period of 10 min non-REM sleep was obtained. Servo-control with tracheal pressure feedback for TTI delivery was not implemented in this protocol.

Experiment 2Polysomnographic responses to hf-TTI at 15 L/min. On two separate nights, patients breathed with the tracheostomy closed either on or off hf-TTI. They were allowed to initiate sleep between 10:00 and 11:00 P.M. During the hf-TTI night, TTI of 15 L/min was initiated after either three epochs of stage 1 or one epoch of stage 2 NREM sleep had occurred. Hf-TTI was only interrupted: (1) by the computer-controlled solenoid whenever tracheal pressure exceeded 20 cm H₂O, and (2) by the investigator for any awakening of three continuous minutes. Hf-TTI was then reinitiated whenever tracheal pressure fell below 20 cm H₂O or after sleep onset, respectively. The trial was maintained for approximately 3 h of sleep, and terminated thereafter. The 3-h maximum total sleep time (TST) was adopted to minimize drying of the tracheal mucosa.

Data Analysis

Sleep architecture. Standard polysomnographic scoring techniques were used to stage sleep (18) and to determine the frequency of respiratory and nonrespiratory arousal during NREM and rapid eye movement (REM) sleep. Arousal was classified by the changes in either EEG or EMG activity of more than 3 s according to American Sleep Disorders Association (ASDA) criteria (19).

Indices of sleep-disordered breathing. Standard polysomnographic techniques were used to determine indices of sleep-disordered breathing, as previously described (20). In brief, an apnea was defined by the complete cessation of airflow for more than 10 s. Hypopnea was defined as a greater than 50% reduction of airflow that was associated with either an arousal from sleep or greater than 4% oxyhemoglobin desaturation. The mean baseline and average low oxyhemoglobin saturations for the apneas and hypopneas were calculated. The tracheal pressure signal was used to assess respiratory effort and discriminate central, mixed, and obstructive apneas, as previously described (20). During TTI administration, discrete episodes of laryngeal obstruction were observed, as detailed in RESULTS (Experiment 1). Such episodes were tabulated and included in the calculation of the sleep-disordered breathing rate whenever they exceeded 5 s in duration, regardless whether they were associated with arousals or oxyhemoglobin desaturation.

Standard indices of sleep-disordered breathing were calculated as follows. In both experimental protocols, sleep-disordered episodes were counted for each trial period. A trial was defined as an approximately 10-min period of NREM sleep at each TTI level in Experiment 1, and for the approximately 3-h period of sleep in Experiment 2. The frequency of NREM sleep-disordered breathing episodes was then computed, as previously described (20). The cumulative duration of sleep-disordered breathing episodes was also calculated. This value was divided by the total sleep time to generate the time spent in disordered breathing (TDB) as percentage of total sleep time.

Respiratory physiology. Tidal fluctuations in tracheal pressure and airflow through the upper airway were examined during TTI administration. The tracheal pressure swings as well as the expiratory peak and inspiratory nadir in the tracheal pressure signal were measured. In addition, the upper airway critical pressure was taken to be the level of tracheal pressure below which airflow through the upper airway ceased in early inspiration.

Statistical analysis. Parameters were expressed as the mean for each patient in each experimental condition. Group data are reported as mean ± SD. To examine responses in breathing patterns to TTI (Experiment 1), two-way analyses of variance (ANOVA) were performed (Minitab Inc., State College, PA) with the patient number treated as a random factor and the TTI flow rate treated as a fixed factor. When the ANOVA revealed significant differences (p < 0.05), post hoc analysis was performed with paired t tests to determine which levels of TTI differed from baseline (no TTI). In Experiment 2 (polysomnographic responses to hf-TTI), paired two-tailed t tests (Minitab Inc.) were performed to compare sleep and breathing patterns on hf-TTI with baseline (no TTI). A p value of less than 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Experiment 1: Polysomnographic Responses to TTI Titration

The breathing patterns during the TTI titration experiment are illustrated for one patient in Figure 2. In the left panel, a typical obstructive apneic episode is seen when this patient was breathing through a closed tracheostomy without TTI. During these apneas, flow through the upper airways (UA) ceased while tracheal pressure (Ptrach) and esophageal pressure (Pes) swings increased progressively. A similar pattern predominated during low-flow TTI administration (5 L/min), and was characterized by marked reductions in UA and progressively increasing swings in tracheal and esophageal pressures (not shown in Figure 2). In both the "no" and "low" flow conditions, apneas and hypopneas were usually associated with oxyhemoglobin desaturation (Sa_O2) and arousal (EMG and EEG) from sleep, and terminated with the resumption of tidal airflow.

View larger version (32K): [in this window] [in a new window]   Figure 2.   Representative polysomnographic recording in one apneic patient breathing with tracheostomy closed and TTI at 0 L/min (left panel ) and TTI at 15 L/min (hf-TTI) in stage 2 sleep (middle panel ) and stage 1 sleep (right panel ). Distinct breathing patterns are noted in each panel consisting of a typical obstructive apnea (left panel ), stable breathing pattern (middle panel ), and two episodes of laryngeal obstruction (right panel ). The obstructive apnea (left panel ) was characterized by absent airflow ( UA) with negative Ptrach swings and decreasing SaO2. In contrast, the laryngeal obstruction (right panel ) was characterized by absent airflow and inspiratory efforts with progressively increasing Ptrach and no decrease in SaO2. Both episodes were terminated by microarousals from sleep. UA = airflow through the upper airways; Ptrach = tracheal pressure; Pes = esophageal pressure; SaO2 = oxyhemoglobin saturation.

In contrast to breathing patterns on 0 and 5 L/min TTI, breathing patterns changed when hf-TTI was applied, depending on sleep stage (Figure 2). During stage 2-4 NREM sleep, obstructive apneic episodes, oxyhemoglobin desaturations, and arousals from sleep were abolished, and breathing pattern stabilized (middle panel). During stage 1 NREM sleep (transitional sleep), however, the breathing pattern was characterized by recurrent episodes of absent airflow through the upper airway (right panel), indicating that the upper airway had occluded. During these periods of upper airway obstruction, we also observed that negative swings in tracheal pressure had ceased. Rather, tracheal pressure increased progressively during these episodes, often exceeding 30 cm H₂O. These findings suggested that the lungs were being passively inflated with insufflated air which could not vent to atmosphere through an occluded upper airway. It is important to note that such episodes were often terminated by arousal from sleep, but were not associated with oxyhemoglobin desaturations.

To determine the site of upper airway obstruction for events occurring in stage 1 NREM sleep, both the supraglottic and tracheal pressures were recorded in two patients during TTI administration. In Figure 3, a period of repetitive upper airway obstruction is illustrated from one patient during stage 1 sleep. In this figure, airflow through the upper airway ceased in each of four episodes. These episodes were associated with a steady increase in tracheal pressure without any change in supraglottic pressure, indicating that the upper airway had occluded below the supraglottic space, presumably in the larynx. For this reason, such episodes are referred to as laryngeal obstructions.

View larger version (9K): [in this window] [in a new window]   Figure 3.   Polysomnographic recording demonstrating a laryngeal site of collapse during repetitive episodes of upper airway obstruction in transitional NREM stage 1 sleep from one patient on TTI at 15 L/min. When airflow through the upper airway UA ceased, tracheal pressure (Ptrach) progressively increased while supraglottic pressure (Psg) fell to atmosphere.

In Figure 4, sleep-disordered breathing parameters are depicted for each TTI trial. Compared with periods without TTI (0 L/min), low-flow TTI (5 L/min) did not change the frequency of obstructive apneas and hypopneas (73.4 ± 16 versus 71.5 ± 21.6 episodes per hour NREM sleep, TTI of 0 versus 5 L/min). In contrast, we observed a marked reduction (p < 0.01) in obstructive apneas and hypopneas at higher TTI flow rates (8.6 ± 8.2 and 3.6 ± 4.2 episodes per hour NREM sleep, TTI 10 and 15 L/min, respectively). Nevertheless, frequent laryngeal obstructions were observed at 13.4 ± 8.4, 38.0 ± 39.2, and 40.1 ± 25.5 episodes/h for TTI at 5, 10, and 15 L/min, respectively. Laryngeal obstructions were present in two patients at 5 L/min, in three patients at 10 L/min, and in four patients at 15 L/min during periods of transitional sleep (Figure 4, middle panel). No mixed or central apneas were observed in trials either on or off TTI.

View larger version (24K): [in this window] [in a new window]   Figure 4.   Responses in breathing patterns to various levels of transtracheal airflow during NREM sleep. The frequency of episodes per hour NREM sleep is presented for obstructive apneas and hypopneas (left panel ), laryngeal obstructions (middle panel ), and sleep-disordered breathing (obstructive apneas/hypopneas and laryngeal obstructions, right panel ). Mean ± SD (closed symbols) and individual data (open symbols) are shown for each experimental condition.

Further analysis of the tracheal pressure waveforms in stable NREM sleep during TTI administration revealed that tidal swings in tracheal pressure fell from 39.7 ± 20.8 to 34.3 ± 13.8 to 24.9 ± 15.7 to 17.5 ± 10.0 cm H₂O at 0, 5, 10, and 15 L/min, respectively. This progressive decrease in tracheal pressure swings was attributable to a rise in the inspiratory nadir from 28.3 ± 18.7 to 19.2 ± 13.1 to 6.7 ± 5.7 to 2.1 ± 3.2 cm H₂O, respectively. The peak expiratory tracheal pressure increased less markedly from 11.4 ± 6.7 to 15.1 ± 11.3 to 18.3 ± 10.8 to 15.4 ± 9.8 cm H₂O, respectively.

With tidal fluctuations in tracheal pressure, airflow through the upper airway also varied, as illustrated in Figure 5. In this figure, one can see that as tracheal pressure fell during inspiration, UA decreased progressively. In fact, UA often ceased as tracheal pressure fell below a critical pressure, indicating that the upper airway had occluded. We found that the upper airway had closed during inspiration in all of our patients while breathing on TTI at 5 L/min, but that upper airway occlusion no longer occurred at TTI rates of 10 and 15 L/min in one and two of our patients, respectively. Moreover, upper airway closure occurred (UA ceased) when tracheal pressure fell below critical pressures of 4.2 ± 5.5, 1.6 ± 2.2, and 0.8 ± 2.7 cm H₂O for TTI flow rates of 5, 10, and 15 L/min which decreased 7.8 ± 7.4 cm H₂O at higher levels of TTI (p = 0.12).

View larger version (22K): [in this window] [in a new window]   Figure 5.   Inspiratory (upper left panel ) and expiratory (upper right panel ), breathing phases represented with corresponding airflow through the upper airway ( UA) and tracheal pressure (Ptrach) signals (lower panel ) when an insufflation flow ( IN) of 15 L/min was applied during stable sleep. During inspiration, VUA ceased (the upper airway occluded) as Ptrach fell (diagonal arrows, lower panel ) as insufflated air ( IN) filled the lungs. During expiration, exhaled and insufflated air vented out the upper airway as tracheal pressure rose. These findings indicate that the upper airway closed and opened in early and late inspiration, respectively, as shown at vertical dotted lines (lower panel ).

Experiment 2: Polysomnographic Responses to hf-TTI (15 L/min)

Breathing pattern. In Figure 6, sleep-disordered breathing indices are represented for each patient on the hf-TTI system. Compared with breathing with the tracheostomy closed (no TTI), hf-TTI abolished obstructive apneas and hypopneas during sleep (63.8 ± 21.8 versus 3.1 ± 1.1 episodes/h, no TTI versus hf-TTI, respectively; p < 0.015; Figure 6, far left panel). On hf-TTI, laryngeal obstructions occurred at the rate of 7.5 ± 8.4 per hour of sleep (Figure 6, middle left panel). In those patients who had laryngeal obstructions during hf-TTI (n = 3), the mean duration of laryngeal obstructions was 14.0 ± 3.5 s. Arousals occurred in 67.3% of laryngeal obstructions. Combining obstructive and laryngeal sleep-disordered breathing, we observed a marked decrease in the sleep-disordered breathing rate on the hf-TTI system (63.8 ± 21.8 versus 10.7 ± 9.1 episodes/h; no TTI versus hf-TTI, respectively; p < 0.03; Figure 6, middle right panel). As a result, the time spent in disordered breathing (DB) decreased markedly (55.5 ± 20.0 versus 5.8 ± 1.1% TST; baseline versus hf-TTI, respectively; p < 0.008, Figure 6, far right panel). Therefore, this hf-TTI system markedly diminished the severity of sleep-disordered breathing.

View larger version (23K): [in this window] [in a new window]   Figure 6.   Breathing patterns during 3-h trials off transtracheal insufflation (no TTI) with the tracheostomy closed (TC) and on hf-TTI. The frequency of NREM sleep-disordered breathing episodes (right middle panel ) and its components (obstructive apnea/hypopnea, far left panel; laryngeal obstructions, left middle panel ) are depicted. Also shown (far right panel ) is the time spent in disordered breathing as a percent of TST. Each patient's data are represented by open symbols and the mean ± SD is represented by closed symbols. p Values represent differences between no TTI and hf-TTI conditions.

Sleep architecture. In Tables 2 and 3, arousal and sleep architecture data are reported for 3-h trials on and off the hf-TTI system. We found the hf-TTI regimen was associated with a marked reduction in arousal frequency from 60.0 ± 26.0 to 8.3 ± 5.4/h in NREM sleep, and from 67.5 ± 3.5 to 0.0 ± 0.0/h in REM sleep (Table 2). Moreover, the distribution of arousals during hf-TTI in NREM sleep were nearly evenly comprised of spontaneous arousals and arousals caused by obstructive and laryngeal sleep-disordered breathing episodes (2.8 ± 0.5, 2.0 ± 1.4, and 3.5 ± 4.4/h, respectively). On hf-TTI, no significant differences in time in bed (TIB), TST, sleep efficiency (SE), sleep stage distribution, and latency to stage NREM and REM sleep were detected in our small sample of patients (Table 3). Nevertheless, each patient exhibited an improvement in sleep stage distribution with either a greater percentage of NREM time spent in stages 2-4 (Patients RD, DI, and RS) or a greater percentage of total sleep time spent in REM sleep (Patient SC).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The overall purpose of this study was to examine the effect of TTI through a small catheter on sleep-disordered breathing. Medically uncomplicated tracheostomized patients with severe OSA were studied because their tracheostomy allowed us to insufflate and monitor tracheal pressure at the same time. In the first experiment, TTI was titrated for brief periods from zero to 15 L/min. This experiment yielded two major findings: First, we confirmed previous studies of the persistence of obstructive apneas and hypopneas during low-flow TTI administration, and found that hf-TTI was required to abolish these episodes during stable sleep. Second, we found that hf-TTI induced repetitive laryngeal obstructions during the transition from wakefulness to sleep. During laryngeal obstructions, tracheal pressure rose progressively as the lungs filled passively with the insufflated air. The latter finding led us to believe it unsafe to deliver hf-TTI unless tracheal pressures were maintained below 20 cm H₂O. We therefore implemented a hf-TTI system that monitored tracheal pressure continuously and administered air whenever tracheal pressure remained below this limit, and examined the polysomnographic responses during an extended hf-TTI trial. In this experiment, we confirmed that hf-TTI both eliminated obstructive apneas and hypopneas, and largely prevented the development of laryngeal obstructions. Moreover, hf-TTI markedly reduced the frequency of microarousals during both NREM and REM sleep. Although consistent improvements in sleep architecture were not observed, patients demonstrated selective increases in sleep efficiency, stage 2-4 NREM sleep, and in the percentages of time spent in either NREM or REM sleep. Our findings demonstrated that hf-TTI effectively eliminated OSA during sleep, thereby stabilizing both the sleep and breathing patterns. The mechanisms for specific breathing patterns observed during hf-TTI administration are discussed next.

Our findings during the TTI titration experiment are consistent with those in previous reports. At a lower flow rate of 5 L/min, we confirmed that recurrent obstructive hypopneas persisted during sleep (15), and further demonstrated that these episodes were eliminated by hf-TTI during stable sleep. Such improvement can be attributed to the fact that higher flow rates better satisfied the patients' ventilatory requirements because tracheal pressure swings decreased under hf-TTI. In fact, our hf-TTI regimen provided our patients 250 ml/s (15 L/min) during inspiration, a level which approaches the normal range of peak inspiratory flows during sleep (21). Despite this higher flow, we found that tracheal pressure swings were still quite large (nearly 20 cm H₂O, see Figure 5), suggesting that even 15 L/min had not fully met our patients' peak inspiratory airflow demands. Nevertheless, we believe that hf-TTI was effective in stabilizing the breathing pattern because it provided a sufficient level of airflow for patients to inspire.

Although breathing stabilized with hf-TTI, airflow dynamics differed considerably from those observed during normal tidal breathing in two respects (Figure 5). First, we found that both the peaks and nadirs in the tracheal pressure swings rose, indicating that the lungs were hyperinflated during hf-TTI administration. Second, we found that air never flowed into the upper airway during tidal breathing on TTI. In fact, when tracheal pressure fell during inspiration, airflow out the upper airway gradually decreased and often ceased at a positive level of tracheal pressure (Figure 5, left panel), consistent with the fact that upper airway critical pressures are often positive in apneic patients (2). These findings indicate that upper airway obstruction persisted during TTI administration and prevented our patients from inspiring through the upper airway. Rather, they inspired solely from the TTI source, and exhaled through the upper airway only after tracheal pressure rose during expiration (Figure 5, right panel). The pressure-flow waveforms during hf-TTI, therefore, indicate that the upper airways opened and closed as tracheal pressure fluctuated throughout the respiratory cycle (Figure 5, bottom panel ). Our findings imply that insufflated air flowed either into the lungs or out the upper airway, depending on the level of tracheal pressure. Furthermore, the breathing route on hf-TTI differs from that during both nasal CPAP and tracheostomy breathing in that patients inspired solely from the TTI source and exhaled solely through their upper airway.

An unexpected finding was the development of repetitive upper airway obstructions during hf-TTI administration as patients transitioned from wakefulness to sleep. During these obstructions, insufflated air ceased flowing out the upper airway and tracheal pressure rose markedly. Indeed, tracheal pressures well exceeded those required to open the pharynx in sleeping apneic patients (2, 4), a finding that can be best explained by active closure of the vocal cords (22). Further evidence for laryngeal obstruction is provided by our finding that tracheal and supraglottic pressures diverged during these episodes (Figure 3). These findings suggest that laryngeal obstructions during hf-TTI were caused by reflex activation of glottic adductors muscles.

Which reflex mechanisms might account for the development of laryngeal obstructions in transitional sleep? It is possible that TTI stimulated flow, pressure, or temperature receptors within the trachea or larynx, which are known to trigger glottic closure and suppress diaphragmatic activity (25). In fact, we found evidence for diaphragmatic inhibition during laryngeal obstructions when spontaneous inspiratory efforts ceased and tracheal pressure rose progressively during these events. As tracheal pressure rose, diaphragmatic activity may have been further inhibited by a marked increase in lung volume (25, 30). Moreover, concomitant hypocapnia may have further activated glottic adductor muscles (31) and inhibited the diaphragm (32), thereby prolonging these episodes. Regardless of the precise reflex pathways involved, it seems clear that these reflex responses were exquisitely dependent on sleep-wake state because laryngeal obstructions only occurred during stage 1 sleep and were terminated by transitions to wakefulness and to other stages of NREM and REM sleep. Thus, it is likely that laryngeal obstructions were generated and sustained by a complex interaction of laryngeal, lung inflation, and CO₂ responses during transitional sleep. When our servo-controlled system discontinued TTI flow (Experiment 2), the larynx reopened and breathing resumed promptly, suggesting that TTI administration was responsible for both the initiation and propagation of these episodes.

There are several implications of our findings concerning TTI administration in patients with OSA. First, the present study provides strong evidence that high-flow rather than low-flow TTI is required to treat OSA. Second, our findings suggest that powerful reflex responses to hf-TTI influence the pattern of laryngeal adductor and diaphragmatic activity, particularly during transitional sleep. These responses account for the development of laryngeal obstructions during TTI administration and pose a risk of barotrauma and further sleep fragmentation. Third, we provide a method for preventing the development of laryngeal obstruction by interrupting TTI whenever tracheal pressure exceeds a threshold level. When the TTI flow was interrupted, laryngeal obstructions decreased in both duration and frequency as patients progressed more readily into deeper stages of NREM sleep. With this servo-controlled system, we demonstrated that hf-TTI could be safely and efficaciously administered for prolonged periods of sleep. Although we empirically set a 20 cm H₂O threshold for interrupting TTI delivery, we speculate that setting a lower threshold in selected patients may have further reduced the frequency of laryngeal obstructions. Further investigations of airflow dynamics and reflex responses to hf-TTI as well as more extensive clinical trials must be performed before this modality can be employed therapeutically in OSA.