INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: apnea/hypopnea syndrome; airway closure; flow limitation

Obstructive sleep apnea/hypopnea syndrome (OSAHS) is a disorder characterized by episodes of partial or complete upper airway obstruction during sleep. This manifests as a reduction (hypopnea) in or complete cessation (apnea) of airflow, despite ongoing inspiratory efforts, and is terminated by a transient arousal from sleep and restoration of upper airway patency. OSAHS consequently lead to disturbances in blood gases and sleep structure and has been associated with a variety of neurobehavioral and cardiovascular complications (1-4).

Because physical obstruction plays an important role in these patients, it is expected that the cycle of apnea and arousal will be accompanied by large changes in the mechanical properties of the respiratory system. In particular, it seems obvious that the resistance to airflow between the mouth/nose and the alveoli will be markedly increased as the obstruction manifests itself. Consequently, most previous studies of the mechanics of breathing during sleep have focused on changes in resistance (5-8). However, there is reason to suspect that obstruction during sleep should influence other mechanical aspects of the respiratory system. If, for example, obstructive apneas were to be accompanied by changes in lung volume (9, 10), then lung elastic properties would likely be affected. We hypothesized, therefore, that sleep-disordered breathing caused by upper airway obstruction produces adverse changes in both upper airway and lower lung mechanical properties.

The primary goal of the present study was, thus, to investigate how upper airway and lower lung mechanics change in adult human subjects with OSAHS as they progress through the various stages of sleep and at arousal. However, measuring lung mechanics in this context is particularly challenging from a technical standpoint because when flow is dynamically limited, it becomes independent of driving pressure and the usual concept of resistance no longer applies. Indeed, Officer and colleagues (11) showed that the presence of expiratory flow limitation can lead to an overestimation of dynamic elastance when using conventional multiple linear regression to estimate mechanical parameters from transpulmonary pressure and flow. Our approach was, therefore, to use numerical methods that we have developed previously specifically for dealing with inspiratory flow limitation (IFL) during sleep (12). We have shown, both in simulated data and in data collected from obese sleeping pigs, that these techniques are capable of providing reasonable estimates of lung mechanics in the presence of IFL. However, they have not yet been applied to sleeping humans, so establishing their usefulness in this regard was a secondary goal of our study.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied seven adult men with untreated obstructive sleep apnea diagnosed in the Sleep Laboratory of The Royal Victoria Hospital in Montreal. Subject characteristics are shown in Table 1. The subjects were aged 39.3 ± 10.7 years with body mass index of 29.3 ± 7.0 kg/m² (mean ± SD). The experimental protocol was approved by the Research Ethics Board of the Royal Victoria Hospital, and informed consent was obtained from each subject.

Polysomnography

The subjects underwent full polysomnography, which included the monitoring of electroencephalographic (EEG) activity (C₄/A₁, C₃/A₂), eye movement, submental electromyographic (EMG) activity, oxygen saturation by pulse oxymetry, thoracoabdominal movements by inductive plethysmography, body position by direct observation, and snoring by microphone. All signals were recorded on a computerized system (Sandman; Mallinckrodt/Nellcor Puritan Bennett, Ottawa, Canada). Sleep staging was performed by a qualified sleep technologist based on standard criteria (13). For all subjects, the total sleep time, sleep efficiency (SE = total sleep time/time spent in bed), and time spent in the various sleep stages were determined. Six of the subjects were studied at night, and one underwent a daytime sleep study due to a regular nightshift work schedule (14, 15). One subject was excluded from our study because of inadequate sleep.

Respiratory events during sleep were scored according to standard criteria (16) and classified into apneas (central, mixed, and obstructive), hypopneas (central and obstructive), and increased upper airway resistance episodes. A hypopnea corresponded to a transient reduction in airflow of more than 10 seconds despite ongoing respiratory efforts. The reduction in airflow had to be either greater than 50% or associated with either a microarousal or a desaturation of more than 3%. An upper airway resistance event was defined as an episode of increased respiratory effort of more than 10 seconds leading to an arousal, but which did not meet the criteria for a hypopnea. These definitions were recommended by the American Academy of Sleep Medicine Task Force and were recently validated by Cracowski and colleagues (17). For purposes of our analysis, we wished to avoid complete cessation of airflow. Therefore, we asked the subjects who displayed obstructive apneas mostly in the supine position to switch to a lateral position, to favor only a partial collapse of the airway.

Respiratory Monitoring

Airflow (V') was measured with a pneumotachograph (Fleisch #2) connected to a differential pressure transducer (Model HCXPM002D6V, SensorTechnics; SCIREQ, Montreal, Canada). The pneumotachograph was attached to the opening of a tightly fitting full-face mask. The total dead space of the mask and pneumotachograph was less than 100 ml. Before each study, the pneumotachograph was calibrated against a rotameter. A gauge pressure transducer (FPM-02PG, Fujikura; Servoflo, Lexington, MA) was connected to a side port of the mask to measure mask pressure (P_m). Respiratory effort was assessed in terms of the magnitude of swings in esophageal pressure (P_es) measured with a small balloon-tipped catheter placed in the esophagus via the nasopharynx. Insertion of the catheter was aided by the application of a small amount (2-4 ml) of 4% viscous Xylocaine to the nasopharynx. P_es was recorded with a pressure transducer connected to the proximal end of the catheter and referenced to atmosphere. Transpulmonary pressure P_tp (equal to P_es  P_m) was used for the computation of lung mechanical properties. Figure 1 shows a typical recording of flow and esophageal pressure signals during wakefulness, during an episode of hypopnea at Stage 2 NREM with varying resistance and airflow, and during snoring in slow-wave sleep (SWS). These signals demonstrate the increase in amplitude of respiratory effort that occurs with the onset of airway obstruction.

View larger version (19K): [in this window] [in a new window]   Figure 1.   Example (subject number 1) of changes in the shape of transpulmonary pressure (Ptp) and flow signals across sleep-wake state. (A) Wakefulness, (B) during an episode of hypopnea with varying airway resistance in Stage 2 NREM sleep, and (C ) stable flow limitation during snoring in slow-wave sleep.

Pharyngeal pressure (P_phar) was also measured in four of our subjects (numbers 4-7) using a silicone catheter with a pressure transducer at its tip (2 mm ID; Gaeltec, Isle of Skye, Scotland). This catheter was passed through the same nostril as the esophageal catheter, under topical anesthesia, and advanced until its tip resided in the hypopharynx just below the base of the tongue. The catheter tip position was confirmed by visual inspection through the mouth (5). P_phar allowed us to partition the respiratory system into its upper airway component (P_UA = P_phar  P_m) and its lung (P_LL = P_es  P_phar) component. Both esophageal and pharyngeal pressure transducers were calibrated using a conventional water manometer.

As well as being captured with the polysomnographic recording system, V', P_es, and P_phar were recorded at a sampling rate of 64 Hz using LABDAT data acquisition software (RHT-InfoDat, Montreal, Quebec) on a personal computer mounted on a trolley at the bedside. The computer and all its associated electrical components were powered through a medical-grade isolation transformer connected to the hospital electrical mains. Data analysis was performed using the Matlab 5.3 mathematical software (The MathWorks, Natick, MA).

Data Processing

We restricted our analysis to NREM sleep because during REM sleep the data were very variable breath to breath and there was only a small amount of data available. We selected segments of data in different stages of NREM sleep and wakefulness in the following manner. After the sleep-wake states in the polysomnographic record were scored, time markers were provided for the LABDAT data record corresponding to wake, Stage 2, Stages 3 and 4, and arousal from sleep. Sections of data in the LABDAT record corresponding to a given sleep-wake state were then randomly selected over the course of the night. Appropriate segments of these data sections were then analyzed. Segments of quiet (unobstructed) breathing were analyzed during wakefulness. During Stage 2 NREM sleep we selected segments containing obstructive hypopneas and episodes of increased upper airway resistance. Within these segments, we analyzed the breaths from the second obstructed breath until arousal, and from the second arousal breath until obstruction. We also chose obstructed segments during Stages 3 and 4 (SWS) NREM sleep in those subjects who reached this sleep stage, which was characterized mainly by stable snoring. We analyzed a total of 3,028 breaths, averaging 79.3 (SD 63.7) breaths during wakefulness, 221.7 (SD 137.3) hypopneas, and episodes of IFL in Stage 2 sleep, 51.3 (SD 29.0) breaths for arousal in Stage 2 sleep, and 187.3 (SD 133.4) breaths for SWS, in each patient.

We used two numerical methods to assess respiratory mechanics information-weighted histograms (IWH) and a modified Mead and Whittenberger (MMW) method. We have previously described these techniques and demonstrated their application in data from sleeping obese pigs (12). They are summarized briefly in the next section.

Information-weighted histograms IWH are calculated by first using recursive multiple linear regression (18) to fit the single-compartment linear model

(1)

View larger version (23K): [in this window] [in a new window]   Figure 2.   Example (subject number 1) of information-weighted histograms for lung resistance RL and elastance EL during (top) normal and (bottom) obstructive breathing. The means ± SD of each histogram are indicated on the figures.

Modified Mead-Whittenberger method The MMW method is based on the classical Mead-Whittenberger method (20), in which Equation 1 is fit to an entire data segment to yield single values for R and E. A limitation of the original method is that it uses only those data points at the extremes of lung volume to estimate E, and so is particularly vulnerable to noise in the data. We modified the original method to use all the data points over the breath to determine E, while retaining the basic assumptions that E does not vary throughout the breath. We then calculate the resistive pressure (P_res) as

(2)

View larger version (16K): [in this window] [in a new window]   Figure 3.   Example of lung resistive pressure Pres versus flow V' curve for normal (A) and obstructed (B) breaths (subject number 1). E was obtained using the modified Mead-Whittenberger method. R was the obtained by regressing Pres against V'.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 2 gives the sleep characteristics of the subjects we studied, together with a summary of their obstructed events during sleep. All subjects had moderate to severe obstructive sleep-disordered breathing, but only four of them (numbers 1, 2, 6, and 7) managed to enter SWS.

Figure 4 gives R_L (lung resistance) and E_L (lung elastance) over the respiratory cycle for all subjects, using both IWH and MMW methods. As sleep progressed, both methods showed a significant increase (t test, p < 0.05) in parameters, compared with the values during wakefulness. With the onset of sleep (Stages 1 and 2 NREM), the upper airway became occluded, resulting in cessation or reduction of airflow. This persisted despite continuing respiratory efforts, until there was a brief arousal from sleep, restoration of upper airway patency, and resumption of airflow. The IWH and MMW techniques both showed a marked reduction in mechanical parameter values following arousal, returning to levels similar to those during wakefulness.

View larger version (23K): [in this window] [in a new window]   Figure 4.   Lung elastance EL (A) and resistance RL (B) over the respiratory cycle for all subjects, estimated during wakefulness, Stage 2 sleep, at arousal and during slow-wave sleep, using both IWH (gray , mean values of histograms) and MMW (white). Data points are means estimated from both methods for each patient. *Represents p < 0.05 and **p < 0.005 for comparison between sleep stages; represents p < 0.05 for comparison between IWH and MMW.

Figure 5 gives a typical example of the P_res-V' curves obtained with the MMW method for the whole lung together with its lower lung and upper airway components. Curves are shown for a normal breath during wakefulness and for an IFL breath with snoring. These curves show that the increase in R_L occurring during sleep and snoring was caused almost entirely by an increase in R_UA. The same was true of all subjects as determined by both analysis methods (Figure 6B). We performed an analysis of variance on the mean values of the IWH for R_UA for all subjects, after log transformation, and showed that R_UA changed significantly with sleep state (p < 0.05)

View larger version (16K): [in this window] [in a new window]   Figure 5.   Pres-V' curves during wakefulness (A) and during IFL in Stage 2 NREM sleep (B) in a representative subject (number 4). Lung resistive pressure, solid thick line; lower lung resistive pressure, solid thin line; upper airway resistive pressure, dashed line.

View larger version (21K): [in this window] [in a new window]   Figure 6.   (A) Lower lung elastance estimated with the IWH (gray , mean values of histograms) and the MMW method (white) during wakefulness, Stage 2 sleep, and at arousal. (B) Corresponding estimates of resistance for both the lower lung (LL) and the upper airways (UA).

A novel finding in this study, as indicated in Figure 4, was that the onset of IFL correlated in all subjects with a significant increase in E_L (analysis of variance performed in each patient separately using repeated measurements, p < 0.05). In those subjects in whom we measured P_phar, we calculated the lower lung component of E_L. We found that the increase in E_L during IFL was largely accounted for by an increase in lower lung elastance (E_LL) (Figure 6A), which was reversed at the end of the obstructed period. E_LL was significantly affected by sleep state (analysis of variance on means of log-transformed IWH, p < 0.05), although lower lung resistance (R_LL) was not significantly affected (p = 0.052).

The widths of the IWH were quantitated in terms of their variances about the mean. These variances for the various sleep stages during inspiration and expiration are given in Table 3. We found statistically significant differences (F test, p < 0.05) for all the subjects between sleep and arousal. Also, during episodes of IFL, the variances during inspiration were significantly larger than during expiration for both R_L and E_L.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our study provides an evaluation of the mechanics of the whole lung, from mouth to pleural space, as well as its upper airway and lower lung components, in patients with OSAHS through the various stages of sleep. A major finding in our study is that the onset of IFL is accompanied by a substantial increase in E_L. The partitioning of the mechanics of the whole lung into its lower lung and upper airway components confirms that this change in E_L occurs in the lower lung. We also found, as expected, that our patients had a normal R_L while awake, but a markedly increased R_L when IFL was present in the various stages of sleep. This increase was due almost entirely to an increase in upper airway resistance (Figure 5B), resulting from collapse of the upper airway. Lower lung resistance changed very little (Figure 6B).

One possible explanation for our finding of an increased E_L during obstructed breathing in sleep is that sleep induces a loss in the mechanical coupling between airways and lung parenchyma. This, in turn, would reduce the outward tethering forces exerted by the parenchyma, thereby allowing the airways to narrow more easily (21). Alternatively, lung volume may have become reduced during sleep due to relaxation of the chest wall. This would affect both the lower (9) and upper (22) mechanical components of the lung. For example, a decrease in lung volume would not only decrease the parenchymal attachments but also increase the forces arising from surface tension, which would lead to an increase in elastance. Appelberg and colleagues (23) recently reported that expiratory reserve volume (ERV) is significantly lower in subjects with sleep-related breathing disorders compared with nonsnoring subjects. They also found a significant independent association between ERV and nocturnal obstructive apnea and oxygen desaturation frequency. Another study (10) showed that the severity of apnea-induced desaturation was highly correlated with the supine ERV as well as with the difference between supine ERV and seated closing volume. This suggests that there is closure of some respiratory units during obstruction, which is reversed at the end of the obstruction. This explanation matches our findings because the increase in E_L observed during IFL in Stage 2 NREM was reversed at the end of the obstructed period.

Another possibility is suggested by several studies that have reported changes in pulmonary hemodynamics following repetitive apneas. Fletcher and colleagues (24) proposed that a large number of obstructive apnea episodes might lead to the development of pulmonary edema, which in turn, would lead to a worsening of gas exchange. These conclusions were drawn from measurements of oxygen saturations, blood pressures, and ventilation-perfusion inequality, the latter also being significantly correlated with the degree of small airway closure (1). The extreme negative intrathoracic pressures generated during airway obstruction would also promote the development of edema, which might have contributed to the increased E_L that we observed. However, the rapidity with which the increase in E_L returned to baseline upon arousal suggests that edema did not play a significant role in our observations.

Large inspiratory efforts, or sighs, would also be expected to have a substantial effect in reducing E_L. Deep inspirations occur during wakefulness and Stage 2 NREM sleep (arousal) and cause dilation of the airways and reinflation of atelectatic zones (25). This is exemplified in Figure 4, which shows that E_L returns to near its wakefulness value after arousal. Although the arousal and wakefulness data are not significantly different, the variability of parameters is bigger for arousal than for wakefulness (Table 3), possibly due to variable degrees of airspace recruitment in the former. Arousals and deep inspirations are absent in SWS, and so presumably allow the atelectasis to accrue progressively. This may have contributed to our finding (Figure 4) that E_L was substantially increased during SWS compared with Stage 2 NREM and wakefulness.

Our estimates of R_L during various sleep stages behaved as expected and are in agreement with previous studies (6, 7). Specifically, there was a significant increase in R_L (Figure 4) as the patients went from a nonobstructed (Wakefulness, Arousal) to an obstructed (Stage 2, SWS) breathing pattern. The bigger scatter (Figure 4) during Stage 2 NREM can be explained by the marked changes in respiratory effort, upper airway caliber, and tidal volume occurring over the cycle of hypopneas. In contrast, during SWS the breathing pattern was much more stable, explaining the small variation in R_L during SWS. The intersubject variability was also smaller during normal compared with obstructed breathing (Figure 4), presumably due to differences in the degree of upper airway obstruction and inspiratory effort during the latter.

It has been suggested that airway obstruction is both an inspiratory and an expiratory event during sleep apnea (26, 27). As both the IWH and MMW methods are able to treat inspiratory and expiratory resistances separately, we calculated resistance over both inspiration and expiration and found an increase in the expiratory R_L during obstructed breathing in Stage 2 and SWS compared with wakefulness and arousal values, which supports the notion that both inspiratory and expiratory events are involved in the stability of the upper airway during wakefulness and sleep in patients with OSAHS.

The pharyngeal catheter allowed us to confirm that the increase in R_L during Stage 2 NREM sleep was indeed due to an increase in upper airway resistance (Figure 6B). The corresponding P_res-V' curve for the upper airway (Figure 5) shows some hysteresis, presumably due to the pharynx being narrower at end inspiration compared with early inspiration (28, 29). This is believed to be influenced by the viscoelastic properties of the upper airway walls and the surface tension forces within the upper airway (30). Such hysteresis was only observed during inspiratory flow limited breaths, which implies that a substantial narrowing of the upper airway is necessary for the hysteresis to occur.

Our study also provided an assessment of the IWH and MMW methods for monitoring flow limitation in humans. As both methods use all data points within the breathing cycle to compute their respective parameter estimates, they should be robust in the presence of noise. This is in contrast to the original MMW method, which has been used recently for the assessment of R_L during sleep (31) and uses only a small number of data points to estimate elastance. Both IWH and MMW methods gave similar results (Figure 4), despite the fact that the MMW method assumes a single value of E_L for the entire breath, while the IWH lets E_L vary continuously throughout the breath. Of course, the two methods are not without their limitations. Despite the fact that they allow mechanics parameters to vary over the breath, both are based on the single-compartment model of the lung, and so do not take the known frequency-dependent natures of resistance and elastance into account. During regular breathing this is not too severe a limitation because most of the power in the breathing signal is at the breathing frequency itself. However, if breathing frequency changes markedly, or if the breathing pattern contains significant high-frequency components, then variations in resistance and elastance reported by either the IWH or MMW methods might be due in part to the frequency-dependent nature of these parameters. Also, in the present study, we used the pressure recorded by an esophageal balloon as a surrogate for pleural pressure. This relies on the perhaps questionable assumption that P_es is an accurate reflection of average pleural pressure in the recumbent position. Nevertheless, the IWH and MMW methods both seem able to provide estimates of overall lung mechanics during upper airway obstruction that reflect the presence of flow limitation, supporting their potential value as monitoring tools.

In summary, we have shown that E_L increased during IFL in sleeping subjects. This observation is most likely explained by a reduction in lung volume resulting in air space closure, as the increase in E_L was rapidly reversed after the end of the obstructed period. We found that most of the increase in R_L during IFL was due to the collapse of the upper airway, which was also reversed at the end of obstruction. We have also demonstrated that the MMW method and IWH can both be used in the setting of obstructive sleep apnea for continuous estimation of lung mechanical parameters and particularly for the detection of flow limitation. Furthermore, the width of the IWH gives an indication of the degree of airway obstruction.