INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Obstructive sleep apnea is characterized by occlusion of the pharyngeal airway, which produces a cessation of oronasal airflow. Previous studies have concluded that the pharyngeal collapse occurs during inspiration as a result of increasing negative intraluminal pressure acting on a hypotonic airway (1). However, the patency of a collapsible tube is determined by the transmural pressure and the compliance of the tube wall. In animals the role of both collapsing intraluminal (2) and extraluminal pressures (3) have been demonstrated. In addition, there is substantial evidence that the pharyngeal airway during sleep behaves as a Starling resistor, that is, pharyngeal occlusion occurs once the intraluminal pressure decreases below the surrounding pressure (4, 5). In patients with obstructive sleep apnea, pharyngeal occlusion occurs at a pressure above atmosphere (6, 7). Likewise, direct measurements of pharyngeal lumen cross-sectional area (CSA), made using fiberoptic imaging, have shown that negative intraluminal pressure is not required to produce narrowing of the airway during sleep (8, 9). Taken together, these findings suggest that extraluminal pressure may be an important determinant of upper airway patency.

The potential for extraluminal pressure to produce pharyngeal collapse is suggested by the findings of several studies showing pharyngeal narrowing during expiration. In humans in 1983, Sanders and Moore (10) found that, from the penultimate to the final breath before occlusion, pharyngeal resistance increased during both inspiration and expiration. More recently, we have shown that in normal sleeping subjects the pharyngeal airway exhibits both inspiratory- and expiratory-related changes in CSA. These changes were more pronounced in obese people and in patients with obstructive sleep apnea (11). Pharyngeal narrowing during expiration is likely to be due to the surrounding extraluminal pressure and may be an important factor in the pathogenesis of obstructive apnea.

The aim of the present study was to test the hypothesis that expiratory-related pharyngeal narrowing occurs prior to the development of negative intraluminal pressure. We studied this by comparing changes in inspiratory and expiratory pharyngeal CSA, for consecutive breaths preceding an obstructive sleep apnea. We used fiberoptic imaging techniques and simultaneous measurement of breathing in patients with obstructive sleep apnea/hypopnea syndrome.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

We studied eight patients with obstructive sleep apnea, all of whom had previously undergone full polysomnography to determine the severity of any obstructive sleep apnea. The anthropometric, polysomnographic, and clinical data are summarized in Table 1. The apnea/ hypopnea index (AHI) was calculated both from the overnight sleep study and from the experimental study when the patients were in the supine position and in NREM light sleep. Four of the eight patients were studied with esophageal pressure (Pes) measurements. The experimental protocol was approved by the Human Subjects Committee at the University of Wisconsin Medical School.

Measurements

Screening sleep study with full polysomnography. Electroencephalograms (EEG), electrooculograms (EOG), and chin electromyograms (EMG) were recorded using the international 10-20 system of electrode placement (EEG: C3-A2 and C4-A1; EOG: F7-A2 and F8-A2). An index of airflow was obtained from temperature changes measured at the nose and mouth (Breathsensor; Edentec, Eden Prairie, MN). Snoring was monitored using a tracheal microphone. The presence or absence of respiratory effort was determined from rib cage and abdominal movements, monitored using respiratory inductance plethysmography (RIP) (Respitrace; Ambulatory Monitoring, Arsley, NY) and surface recording of diaphragmatic EMG (Rochester Electro-Medical, Rochester, NY). An estimate of arterial oxygen saturation (Sa_O2) was obtained using a pulse oximeter (Biox 3700; Ohmeda, Boulder, CO).The presence of any limb movements were calculated from surface recording of tibia EMG.

Sleep study with measurement of retropalatal CSA. EEG, EOG, and chin EMG were measured as described above. Airflow () was measured using a pneumotachometer (No. 1; Hans Rudolph, Kansas City, MO) attached to a nasal mask. Pressure within the mask (Pmask) was measured continuously (MP-45, ± 5 cm H₂O differential pressure transducer; Validyne Engineering Corp., Northridge, CA). Tidal volume (VT) was obtained from the integrated airflow signal. Sa_O2 was monitored using a pulse oximeter (Ohmeda). Rib cage and abdominal movements were monitored using DC-coupled respiratory inductance plethysmography (Respitrace; Ambulatory Monitoring). The relative gain of the rib cage and abdominal movements was determined using an isovolume procedure.The relative amplitude of the sum signal was calculated by comparing it with the integrated airflow.

Esophageal pressure. In four patients, measurements of esophageal pressure (Pes) were used to provide an index of pleural pressure (12). After topical nasal anaesthesia with 10% lidocaine hydrochloride solution (Astra Pharmaceutical, Lund, Sweden), a catheter tipped with a pressure transducer (Model TL-500; Millar Instruments, Galveston, TX) was passed through the nose and positioned in the middle third of the esophagus. The tip of the catheter was approximately 40 cm from the nares. A check that Pes reflected pleural pressure was made by observing the pressure changes with respect to atmospheric pressure during voluntary sniffing and coughing.

Pharyngeal lumen visualization. The retropalatal airway lumen was visualized using a pediatric fiberoptic endoscope (3.2 mm OD, model 3C-10; Olympus, London, UK, or model FB-10X; Pentax). After topical upper airway anaesthesia, the scope was passed through the opposite nostril to the Pes catheter and positioned approximately 2 to 3 cm above the soft palate. A continuous image of the retropalatal lumen was obtained from a closed-circuit video camera connected to the scope (Endovision 3000, Pentax; Precision Instrument Co., Marlboro, MA). The video image was relayed to a frame grabber board (DT3851; Data Translation) on a 486/66 compatible computer. The synchronized respiratory signals (up to six channels) were relayed to an A/D board (LabMaster DMA; Scientific Solutions, Solon, OH) on the computer. The video frames and respiratory signals were digitized using specially developed software at five frames per second and 25 Hz, respectively (13). The video images and the modulated airflow signal (FM-1 mod/demod; Wolf Industries, San Marino, CA) were also recorded onto videotape for data storage.

Protocol

Each patient reported to the laboratory 1 h before his or her normal sleep time. Firstly, sleep-staging electrodes and the respiratory inductance plethysmography bands were attached. The Pes catheter was passed as described above. The patient then lay supine on a bed and performed an isovolume procedure, breathing slowly against an occluded airway to change the shape of the rib cage and abdominal compartments (14). This procedure was carried out to determine the relative contributions of the rib cage and abdominal signals to the sum of the two signals. Next the fiberoptic scope was passed into the retropalatal airway. The nasal mask was secured and the exact position of the scope was adjusted. The mask was carefully sealed, and a check for air leakage around the mask was made by occluding the airflow during an attempted inspiration and expiration. The remaining transducers were attached and further fine adjustments to the orientation of the scope were made. The patient was then instructed to take several deep breaths; during this procedure, the position of the scope on the longitudinal pharyngeal axis was confirmed. At this time the patient was allowed to go to sleep. If the patient reported that he or she would normally breathe through the mouth while asleep, it was taped shut. During the study, any overt movement of the mouth, opening of the lips, or dropping of the jaw was noted by an investigator who was in the room with the patient. Data collected during these periods were not analyzed. A continuous on-line integration of the airflow signal was used to assess whether less obvious air leaks (such as those associated with loss of facial muscle tone) occurred during the study; these data were also discarded.

All variables were monitored continuously throughout the study. The fiberoptic images and the analogue signals were accessed to the on-line computer during selected periods of repeated apnea. This allowed the collection of images prior to an apneic period, but excluded the need to continuously record all data.

Analysis

For four breaths preceding an obstructive apnea, the inspired minute ventilation (I), inspired tidal volume (VT), inspiration time (TI), expiration time (TE), and total breath duration (Ttot) were calculated. Estimates of changes in end-expiratory lung volume (EELV) were calculated from the sum of the rib cage and the abdominal movements. The retropalatal CSA was obtained for each digitized frame (five frames per second) by manually outlining the retropalatal lumen using computer software (Sigma Scan; Jandel Scientific, Corte Madera, CA) as shown in Figure 1. During this process the investigator was blinded to the phase of respiration.

View larger version (59K): [in this window] [in a new window]   Figure 1.   An example of a fiberoptic image of the retropalatal airway with anatomic landmarks indicated. The outline of the pharyngeal lumen (used to calculate CSA) is shown by the white line.

For each of the four breaths preceding the apnea, the CSA of the retropalatal lumen was calculated at five fixed percentages within inspiration (0, 25, 50, 75, and 100%) and six fixed percentages within expiration (0, 20, 40, 60, 80, and 100%). For breaths when the video frame did not coincide with the fixed percentage at which we wished to make our measurement, the frame nearest in time to the desired phase in the respiratory cycle (i.e., 0, 20, 40%, etc.) was used. For breaths at the transition between each breath (i.e., end-expiration to the start of the subsequent inspiration), the frame during which flow crossed zero was defined as end-expiration (100%), and the subsequent digitized frame (i.e., 0.2 s later) was used as the start of inspiration (0%). For the breath immediately preceding the apnea, end-expiration was defined as the frame during which expiratory flow returned to, or crossed, zero. The largest CSA that occurred during the four breaths was termed 100% and the within-breath changes of luminal CSA were referenced to this measurement. The four breaths preceding the obstructive apnea were numbered breath-4 through breath-1, with breath-1 being the breath immediately preceding the apnea.

For each patient, we analyzed all obstructive apneic periods during which the airway lumen was visible and no movement or respiratory artefact (e.g., a swallow) occurred. An obstructive apnea was defined as a cessation of airflow for greater than 10 s accompanied by respiratory effort. To exclude any so-called mixed apneas (a cessation of respiratory drive followed by a respiratory effort associated with an occluded airway), we required the first apneic inspiratory effort to occur within 5 s of the end of inspiration in breath-4. A "breath" was defined as a respiratory effort producing a VT greater than 25 ml.

In the four patients in whom Pes was recorded, a separate analysis was performed of the CSA, , and Pes during the breath immediately preceding the apnea (breath-1) and during the first occluded effort during the apnea. The CSA and the corresponding Pes were analyzed at six points: (1) the nadir inspiratory Pes of breath-1, (2) the maximum expiratory Pes of breath-1, (3) end-expiration of breath-1, (4) the point at which occlusion of the airway occurred during the subsequent apnea, (5) the nadir inspiratory Pes of the first inspiratory effort during the apnea, and (6) the point of reopening during the obstructive apnea (not all patients showed expiratory related reopening during the apnea).

Statistical Analysis

For each individual, measurements from the four breaths preceding the apnea were averaged. For the patients who were studied twice, the data from the two nights were combined. Comparison of the mean values of respiratory variables VT, TI, TE, Ttot, and I, for each of the four breaths preceding the apnea was carried out using a one-factor analysis of variance with repeated measures; post hoc comparisons of the changes in each respiratory variable for each breath compared with the subsequent breath and breath-4 versus breath-1 were performed using Fisher's least significant difference statistic. For all comparisons statistical significance was taken as p < 0.05.

Analysis of changes in CSA was performed within each breath and also between each breath and the subsequent breath. For each breath, within-breath measurements of CSA were compared using a one-factor analysis of variance with repeated measures; post hoc comparisons of the CSA were performed using Fisher's least significant difference for the start of inspiration (0%) versus midinspiration (50%), midinspiration versus early expiration (20%), and early expiration versus end-expiration (100%). Between-breath measurements of CSA were compared using a one-factor analysis of variance with repeated measures; post hoc comparisons of the CSA were performed using Fisher's least significant difference statistic for midinspiration (50%), early expiration (20%), and end-expiration (100%), for each breath compared with the subsequent breath and for breath-1 versus breath-4. For the within-breath and between-breath comparisons, the period in the respiratory cycle at which the comparisons were made (i.e., 20, 50%, etc.) was selected from a preliminary analysis. They represent the point at which the maximum and minimum lumen CSAs most often occurred in each patient. This approach may have produced an underestimation of the relative changes in some patients. However, it reduced the likelihood of producing a significant change in the CSA because of a data selection bias.

A separate analysis of the breath immediately preceding the apnea (breath-1) and the subsequent occluded respiratory effort was carried out using paired t tests, for the Pes at which the lumen closed compared with the nadir Pes of the previous unoccluded breath and the Pes at which the lumen reopened.

Analysis of the changes in EELV was carried out using paired t tests. The difference in EELV (EELV) between each breath and the preceding breath, e.g., Breath-1 = (EELV breath-2 minus EELV breath-1) was compared with the difference in the EELV for the subsequent breath.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Each sleep study lasted approximately 4 h. All patients had obstructive apneas during which clear pictures of the retropalatal airway lumen were obtained. Studies were most often terminated either because secretions obscured a clear view of the airway lumen or because the patient woke up and was unable to get back to sleep.

Respiratory Variables of Four Breaths Immediately Preceding an Obstructive Apnea

The group mean values (± SD; n = 8; 3 to 22 preapneic periods per patient) for each respiratory variable are shown in Table 2. Over the four-breath period there was a significant reduction in I (breath-4 versus breath-1, p = 0.05), but not when each breath was compared with the subsequent breath (breath-4 versus breath-3, p = 0.06; breath-3 versus breath-2, p = 0.44; breath-2 versus breath-1, p = 0.06). The reduction in I was associated with a significant reduction in VT over the four-breath period (breath-4 versus breath-1, p = 0.02) and between successive breaths (breath-4 versus breath-3, p = 0.04; breath-2 versus breath-1, p = 0.03). The prolongation of each of the respiratory timing variables was not significant (breath-4 versus breath-1, Ttot, p = 0.18; TE, p = 0.30; TI, p = 0.89). During breath-4 the relatively large I suggests a degree of hyperventilation, possibly associated with the termination of a previous apnea. However, for the breaths selected it is difficult to judge the extent of any post-apnea hyperventilation due to the absence of a reliable measure of PCO₂.

Pharyngeal Patency during Breaths Preceding an Obstructive Apnea

An example of the four breaths preceding an obstructive apnea together with the corresponding changes in retropalatal CSA are shown for one patient in Figure 2. Note that the nadir inspiratory retropalatal CSA, and the end-expiratory CSA were reduced with each successive breath. The group mean (± SEM) CSA, measured at fixed percentages within each breath, and the corresponding are shown in Figure 3.

View larger version (19K): [in this window] [in a new window]   View larger version (72K): [in this window] [in a new window]   Figure 2.   (Top panel ) An original recording (Patient 7) of airflow (Flow; inspiration positive), esophageal pressure (Pes), and rib cage and abdominal movements. Tracings show four breaths leading to an obstructive apnea (breaths-4, -3, -2, -1). During the apnea, respiratory effort is indicated by the negative swings in the esophageal pressure and paradoxical rib cage and abdominal movements. (Bottom panel ) Fiberoptic images of the retropalatal airway during the four breaths shown in the top panel where breath-1 is the breath immediately preceding the apnea and breath-4 is the breath farthest away from the apnea. Within each breath the images selected correspond to the smallest cross-sectional area that occurred during inspiration (Nadir Insp), the largest CSA during expiration (Peak Exp), and the CSA at end-expiration (End Exp). Within each image the dark area is the airway lumen, the lighter horseshoe shape is the epiglottis, and the white triangular shape in the bottom left corner is the esophageal pressure catheter. Note that in this example, for breath-2 and breath-1, the airway lumen at peak expiration is a similar size; this was not representative of the group mean results.

View larger version (20K): [in this window] [in a new window]   Figure 3.   Group mean (± SEM; n = 8) airflow (; inspiration positive) and cross-sectional area of the retropalatal airway (CSA) measured at fixed percentages (0 to 100%) within inspiration (I) and expiration (E) during the four breaths preceding apnea (Breath-4 to Breath-1). The group mean CSA was calculated from the individual CSAs expressed as a percentage of the maximum area that occurred at any time during the four breaths preceding each apnea. The asterisks indicate statistically significant changes in CSA. The comparisons within each breath are indicated by the brackets above the CSA measurements, and the comparisons between each breath are indicated by the brackets below.

Within-breath changes in CSA. Within each breath, the inspiratory-related decrease in the retropalatal CSA was significant only during the last breath before the apnea (Breath-1, CSA at the start of the breath versus CSA at 50% inspiration, p < 0.05). Each of the four preapneic breaths was associated with a significant increase in CSA during the transition from midinspiration to early expiration (CSA at 50% inspiration versus CSA at 20% expiration, p < 0.05), followed by a significant narrowing of the airway lumen during expiration (CSA at 20% expiration versus CSA at end-expiration, p < 0.05).

Between-breath changes in CSA. During the four preapneic breaths there was a significant reduction in the CSA at end-expiration (breath-4 versus breath-1, p < 0.05); however, the reduction was not significant when each successive breath was compared with the subsequent breath. Between the penultimate breath and the breath immediately preceding the apnea there was a significant decrease in both the nadir inspiratory CSA and the peak expiratory CSA (breath-2 versus breath-1: CSA at 50% inspiration, p < 0.05; CSA at 20% expiration, p < 0.05).

Occlusion of the Retropalatal Airway during an Obstructive Apnea

An example of the changes in retropalatal CSA, , and Pes in one patient during the breath immediately preceding an apnea and the subsequent apnea is shown in Figure 4. Note that narrowing occurred during the expiratory phase of the breath preceding the apnea and that the subsequent airway occlusion occurred at a pressure that was less than that of the previous unoccluded breath. The group mean (± SEM) changes in retropalatal CSA, , and Pes are shown in Figure 5. During the apnea airway occlusion occurred at 8.2 ± 1.9 cm H₂O, this pressure was significantly less negative than the pressure generated during the previous unoccluded breath, 10.8 ± 2.9 cm H₂O (p = 0.04). In three patients the airway reopened during the apnea (i.e., during expiration with minimal airflow); in these patients the reopening pressure was significantly less negative than the closing pressure (n = 3, closing pressure = 9.9 cm H₂O; reopening pressure = 1.9 cm H₂O, p = 0.007).

View larger version (27K): [in this window] [in a new window]   Figure 4.   An original recording (Patient 6) of airflow (Flow; inspiration positive) and esophageal pressure. Tracings show three breaths leading to an obstructive apnea. Fiberoptic images are shown for the breath immediately preceding the apnea (Breath-1) and the apneic period. During Breath-1 the images selected correspond to the smallest cross-sectional area (CSA) that occurred during inspiration (Nadir Inspiration), the largest CSA during expiration (Maximum Expiration), and the CSA at end-expiration (End Expiration). During the apnea the images correspond to the time at which airway occlusion occurred (Occlusion), the time at which the maximum esophageal pressure was generated (Nadir Inspiration), and the time at which the airway reopened (Reopening). Within each image the dark area is the airway lumen, the lighter horseshoe shape is the epiglottis, and the white triangular shape on the right is the esophageal pressure catheter.

View larger version (17K): [in this window] [in a new window]   Figure 5.   Group mean (± SEM; n = 4) cross-sectional area of the retropalatal airway (CSA) (closed circles) and esophageal pressure (Pes) (open circles) measured at defined times during the breath immediately preceding the apnea and during the apnea. The group mean CSA was calculated from the individual CSAs expressed as a percentage of the maximum area that occurred at any time during the four breaths preceding each apnea.

Changes in the Pressure/Flow Relationships during Breaths Preceding an Obstructive Apnea

The changes in , Pes, and corresponding retropalatal CSA for the three breaths immediately preceding an obstructive apnea, for each individual in whom esophageal pressure was measured (n = 4), are shown in Figure 6. Note that the subjects in whom the airway narrowed most during inspiration were not always the ones in whom the swings in Pes were largest. Interestingly, during each expiratory phase all subjects showed a similar pattern of airway narrowing. Three of the 61 apneas analyzed showed complete airway occlusion at end- expiration before the generation of inspiratory-related negative pressure.

View larger version (30K): [in this window] [in a new window]   Figure 6.   For each patient (Patient 3 [open circles], n = 17; Patient 6 [closed squares], n = 22; Patient 7 [open squares], n = 14; Patient 8 [closed circles], n = 21) mean airflow (), esophageal pressure ( Pes), and retropalatal cross-sectional area (CSA) was measured at fixed percentages (0 to 100%) within inspiration (I) and expiration (E) during the three breaths preceding apnea (Breath-3 to Breath-1). The CSA is expressed as a percentage of the maximum area that occurred at any time during the four breaths preceding each apnea.

We observed that not all patients achieved a reduction in CSA in the same way. In some patients the pharyngeal lumen narrowed because of movement of the lateral walls (Figure 2) and in some the anterior/posterior walls moved (Figure 4). This heterogeneity in the direction of narrowing suggests that not all patients achieve airway collapse in the same way.

Changes in the End-Expiratory Lung Volume during Breaths Preceding an Obstructive Apnea

Changes in individual and in group mean levels of EELV are shown for each of the breaths preceding the apnea in Figure 7. Over the preapneic breaths there was a significant reduction in EELV (breath-3 versus breath-1, p = 0.03). The reduction in EELV between successive breaths were not significant (breath-3 versus breath-2, p=0.15; breath-2 versus breath-1, p = 0.06).

View larger version (15K): [in this window] [in a new window]   Figure 7.   Individual (n = 8) (closed squares) and group (open circles) mean changes in end-expiratory lung volume (EELV) for three consecutive breaths preceding apnea (Breath-3 to Breath-1). EELV was calculated as the difference in EELV between the breath shown and the preceding breath.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have tested the hypothesis that a reduction in pharyngeal CSA occurs during expiration prior to the development of negative intraluminal pressure. The main findings of our study were that: (1) for each successive breath prior to an obstructive apnea there was a progressive reduction in the end-expiratory CSA, (2) there was significant airway narrowing during both inspiration and expiration in the breath immediately preceding an obstructive apnea, (3) during the apnea, occlusion of the pharyngeal lumen occurred at an esophageal pressure that was significantly less negative than the previous unoccluded breath.

Limitations of the Techniques and Methodology

The findings of this study should be interpreted bearing in mind the following limitations. Our patients were selected according to the AHI measured during a previous screening sleep study. This may have led to an underestimation of the airway collapsibility that occurred when the patient slept supine. Therefore, we also calculated the apnea index (AI) that occurred under our experimental conditions (light sleep/supine position); in some cases there was a large difference (Table 1). In these patients we are likely to have overestimated the degree of airway narrowing that would have occurred if the patient had slept in the lateral position.

Our group contained both male and female subjects of varying ages and BMI. It is possible that these factors might have influenced the intraluminal and extraluminal upper airway collapsing forces. We did not find any relationship between any of these variables (sex, age, BMI) and the magnitude of the changes in the inspiratory or expiratory CSA, although with such small sample sizes our conclusions must be interpreted with caution.

Our measurements of CSA could have been influenced by any rostrocaudal movement of the fiberoptic scope. To prevent this, we analyzed only periods in which there was no change in the relationship between the fiberoptic image and anatomic landmarks at different planes (e.g., soft palate, epiglottis, vocal cords). In our patients with severe obstructive apnea, movement of the head associated with arousal at the termination of an apnea frequently occurred. Many apneic periods were rejected because of such movements.

The methods we used to analyze the fiberoptic images of the pharyngeal airway meant that we were unable to measure absolute changes in airway size. In this case, for the patients in whom the absolute CSA was relatively small, a minimal decrease in percentage CSA would have large implications for airflow. However, our approach of examining relative changes in CSA over a four-breath period is likely to have reduced this potential limitation. Furthermore, in a study that compared the pattern of airway narrowing measured using fiberoptic techniques with those obtained with magnetic resonance imaging (that latter providing absolute measurements of airway caliber), the magnitude of airway narrowing was similar (15).

In the present study, we measured retropalatal CSA. Previous studies have shown that narrowing of the pharyngeal airway can occur at multiple levels (16, 17). However, retropalatal level is the primary site of narrowing in the majority of patients with sleep apnea (8, 18, 19). In three patients, during the apnea, we observed reopening of the retropalatal airway in the absence of airflow (Figure 4). It is likely that this occurred because of reopening of the retropalatal but not the retroglossal airway.

Previous studies have shown that sleep apnea can be exacerbated in patients with nasal blockage and that the route of breathing may also influence upper airway dilator muscle activity (20, 21). We avoided this confounding effect by analyzing only periods when the patients were breathing through their noses. This means it may not be appropriate to extrapolate our findings to patients who breathed through their mouths during sleep.

Progressive Expiratory Pharyngeal Narrowing

We have observed narrowing of the end-expiratory pharyngeal CSA during breaths preceding an obstructive apnea and on a number of occasions the airway occluded before the subsequent inspiratory effort. The expiratory narrowing we observed may have been associated with a decay of the positive intraluminal pressure generated during early expiration. Previous studies have shown that pharyngeal compliance increases progressively as CSA decreases (22). This would favor further narrowing, as would the finding that the pharyngeal airway is more compliant in patients with obstructive sleep apnea compared with nonobese normal subjects (23, 24). However, a passive return of the CSA to baseline is unlikely to account for the progressive reduction in airway caliber seen across successive preapneic breaths. Active neuromuscular constriction could produce progressive expiratory-related pharyngeal collapse; however, recent evidence in sleeping humans, suggests that these muscles are unlikely to contribute to airway narrowing either during eupneic breathing or central or obstructive sleep apnea (9, 25, 26).

In the present study we found that airway occlusion was preceded by a small decrease in the end-expiratory lung volume (EELV). Changes in lung volume could be associated with changes in pharyngeal transmural pressure as a result of caudal traction on the lateral neck. In dogs, caudal traction has been shown to produce changes in pharyngeal patency, independent of changes in upper airway muscle activity (27).We suggest that in our study, the expiratory-related reduction in lung volume was associated with a decrease in transmural pressure that resulted in narrowing of the pharyngeal lumen. The changes in lung volume that we observed were relatively small compared with a previous study in normal awake humans (28). These workers found a reduction in the minimum velopharyngeal CSA measured at residual volume, compared with that measured at functional residual capacity. It is likely that during sleep, the pharyngeal lumen is more susceptible to changes in lung volume. Begle and coworkers (29) showed that during sleep relatively small changes in FRC (mean change in FRC, 290 ml; range, 80 to 490 ml) were associated with significant changes in pulmonary resistance, even in the face of a reduction in upper airway dilator muscle activity. The susceptibility of the airway to collapse is also likely to be increased when lung volume is reduced by moving from the standing to the lying position, especially in obese patients. Indeed, for comparable changes in lung volume, patients with obstructive sleep apnea have larger changes in pharyngeal CSA than do normal subjects (30).

Finally, the expiratory-related narrowing of the pharyngeal lumen would have led to an increase in airflow velocity and therefore a decrease in intraluminal static pressure (Bernoulli principle). This effect would have caused the intraluminal pressure to become more negative during the subsequent inspiration and thus contribute towards airway collapse.

In the majority of apneas we analyzed, inspiratory effort was required to produce pharyngeal occlusion, underlining the importance of this factor. It is interesting that the level of negative intraluminal pressure required to produce the occlusion was negligible and often less than that which occurred during the previous unoccluded breath (Figure 5). In keeping with other studies, we found that the intraluminal pressure did not become progressively more negative across the breaths leading into the apnea (31). This finding may suggest that pharyngeal compliance increases progressively for the breaths preceding an apnea; as mentioned earlier compliance changes dynamically in relation to airway volume. In addition, the volume/compliance relationship may be influenced by the sleep-related loss of upper airway dilator muscle activity, such that the airway becomes more compliant during sleep (31, 32).

Implications of Progressive Expiratory Narrowing for the Pathogenesis of Obstructive Sleep Apnea

In our study, the occurrence of pharyngeal narrowing during expiration, when intraluminal pressure was positive, suggests that the surrounding extraluminal pressure was also positive and higher than intraluminal pressure. Thus, our findings are consistent with studies in anesthetized rabbits and dogs that have demonstrated that surrounding extraluminal pressure, applied by mass loading of the anterior neck and negative pressure around the neck, respectively, can significantly influence pharyngeal caliber (3, 33). They are also consistent with studies in humans that have noted airway occlusion at atmospheric pressure during either spontaneously occurring central apneas (8) or central apneas induced by pharmacologic muscle paralysis in patients with obstructive sleep apnea (19). It should be noted that the relative contribution of the extraluminal pressure versus intraluminal pressure to pharyngeal occlusion can not be determined from our data.

In summary, the findings of our study support the idea that pharyngeal occlusion is due to a combination of inspiratory and expiratory narrowing, with the latter rendering the lumen vulnerable to complete collapse during the subsequent inspiratory effort.