INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Nasal continuous positive airway pressure (nCPAP) has considerably improved the treatment and prognosis of patients with obstructive sleep apnea syndrome (OSAS). The optimal nCPAP level is generally determined during an overall night of polysomnographic investigation, but the nCPAP titration procedure, and more particularly the criteria characterizing the optimal nCPAP level, differ among studies. It is now commonly admitted that the optimal level of nCPAP must not only abolish apneas and hypopneas, but also normalize the upper airway resistance value (1). Whereas apneas and/or hypopneas can be easily detected in flow signals sensed by external transducers such as thermistors or pneumotachographs, the detection of residual intermittent upper airway obstruction (UAO) theoretically requires measurement of esophageal pressure (Pes). However, the presence of a nasopharyngeal catheter may not only modify the patient's UAO (4), but also disturbs the patient's sleep (1, 5). In this connection, Chervin and Aldrich showed only minimal effects of monitoring of Pes on sleep architecture, but these authors did not analyze the arousal index (6). Therefore, noninvasive markers of UAO, based on analysis of the inspiratory flow contour and the detection of flow limitation, have been proposed (1, 2). Currently, the forced-flow oscillation (FO) technique, a noninvasive method for quantitative measurement of respiratory resistance, is available.

The first aim of the present study was to demonstrate the applicability of the FO technique to sleeping patients receiving nCPAP, and its reliability in assessing the UAO level. For this purpose, the possible influences of FO on sleep and lung mechanics were investigated, and respiratory resistive impedance was compared with inspiratory lung resistance (RLI) (i.e., airway plus tissue resistance, as determined from flow and Pes signals). Additionally, resistive impedance was compared with two other parameters also used as indices of UAO for nCPAP titration: esophageal pressure amplitude (Pes) (3, 7), and the flow score (FS) derived from analysis of the inspiratory flow contour (2).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Eleven OSAS patients without a history of pulmonary obstruction from asthma or chronic obstructive pulmonary disease (COPD) were studied during a night of polysomnography-controlled nCPAP titration. The patients' characteristics are given in Table 1. The diagnosis of OSAS had been established previously on the basis of nocturnal polysomnography including electroencephalography (C4-A1, C3-A2), electrooculography, chin electromyography, electromyography of the tibialis anterior muscle of both legs, assessment of oronasal airflow with thermistors, rib cage and abdominal movements (Multi-Parameter Analysis recorder 2/Medilog 9200; Oxford Medical Instrument, Abingdon, UK), and arterial pulse oximetry (Model BS; Nellcor, Hayward, CA). Sleep stages were scored according to the international criteria proposed by Rechtschaffen and Kales (8). Arousals were scored according to the American Sleep Disorders Association recommendations, as abrupt shifts in electroencephalographic frequency for at least 3 s, irrespective of any change in the submental electroencephalogram (EEG) during non-rapid eye movement (NREM) sleep but accompanied by a concurrent 3-s increase in chin electromyographic amplitude during rapid eye movement (REM) sleep (9). According to the clinical criteria commonly used for thermistor signals, an abnormal breathing event during sleep was defined as either a complete cessation of airflow lasting at least 10 s (apnea) or a reduction in oronasal airflow lasting at least 10 s and associated with a decline in Sa_O2 of at least 3% (hypopnea) (10). OSAS was diagnosed on the basis of an apnea-hypopnea index (AHI) value of more than 10 episodes per hour of sleep.

During the nCPAP titration night, sleep polygraphic investigation was modified as follows: nasal flow () was measured via a tight-fitting, low dead-space nasal mask with a Fleisch No. 2 pneumotachograph (Fleisch, Lausanne, Switzerland) connected to a differential pressure transducer (Sensym SCX 01D, ± 70 cm H₂O; Sunnyvale, CA); airway opening pressure (Pao) was measured via a short tube connected to a port in the nasal mask by a similar transducer; Pes was measured with a pressure transducer (Gaeltec; Dunvegan, Isle of Skye, UK) connected to a catheter placed transnasally, and carefully positioned in the lower third of the esophagus to minimize cardiogenic artifact; and oral airflow was evaluated qualitatively with an oral thermistor. All signals were sampled at 128 Hz with an analog-to-digital system (MP100; Biopac System, Goleta, CA), and were fed into a computer for further analysis.

Experimental Protocol

Nasal CPAP was increased in 1 cm H₂O increments from the lowest value preventing rebreathing (4 cm H₂O) to the level at which the three following conditions were fulfilled: (1) apneas, hypopneas, and snoring disappeared; (2) no inspiratory flow limitation occurred (i.e., no plateau was observed in the nasal inspiratory flow signal; (3) the amplitude of swings in Pes did not exceed twice its value during quiet breathing in the awake state (3). The level of nCPAP was decreased at the awakening patient's request. Each nCPAP step lasted at least 10 min, and FO were applied for a 5-min period before each change in the nCPAP level.

Forced Oscillations

The pseudorandom forced flow used in the study was composed of 13 harmonics (4 to 16 Hz) of the fundamental frequency (1 Hz), with enhanced amplitudes at the lower frequencies to limit the influence of spontaneous breathing. The phases were calculated in order to minimize the peak-to-peak amplitude of the excitation signal. The forced flow signal, generated by a digital-to-analog converter, excited, through a power amplifier, a 50-W loudspeaker (Audax HM 130 XO) enclosed in a 2.5-L rigid chamber and placed in parallel with the nCPAP device, as in the experimental setup described by Peslin and colleagues (11). The peak-to-peak amplitude of the resulting flow was about 0.2 L/s.

Respiratory Resistive Impedance

Airway opening pressure (Pao) and data were collected over 16-s periods and high-pass filtered (third order, cutoff frequency = 3.5 Hz) to eliminate the low harmonics of breathing noise. A fast Fourier transform (FFT) algorithm was applied to adjacent 4-s periods. Impedance data were calculated from the auto- and cross-spectra, and were retained for analysis when they corresponded to a coherence value greater than 0.9 (12). The real part of respiratory impedance (ZR) was submitted to linear regression analysis against frequency, and the index adopted to evaluate respiratory resistance was the ZR value calculated at 16 Hz (respiratory resistive impedance at 16 Hz [RFO]). Determination of RFO was impossible during obstructive apneas because of the low coherence value of the impedance data, but pertinent RFO values were obtained during the few central apneas that occurred when FO were applied.

Lung Resistance

Lung mechanics were derived from the same 16-s periods as respiratory impedance. Pes, transpulmonary pressure (PTP = Pao  Pes), and data were low-pass filtered (third order, cutoff frequency = 1.5 Hz) to eliminate the high harmonics resulting from pseudorandom noise. Pes, PTP and data were analyzed cycle by cycle. The change in amplitude of esophageal pressure (Pes) was calculated. Lung elastance (EL) was first determined by linear regression analysis of PTP versus volume (V) over the entire ventilatory cycle, according to the equation PTP = EL × V + P₀, where P₀ is a constant that reflects transpulmonary pressure at the end-expiratory lung volume. Inspiratory lung resistance (RLI) was then determined by linear regression analysis of resistive pressure (PTP  (EL × V)  P₀) versus during inspiration. During the same cycles we also determined mean lung resistance (RL), by using multiple linear regression analysis of PTP versus and its time integral, V, over the entire ventilatory cycle. Pes, RLI, EL, and RL were taken as the averages of their respective values.

To determine whether lung mechanics were affected by the application of FO we also calculated RLI, RL, and elastance during the 32-s period just before and just after FO application, before increasing the nCPAP level.

Flow Score

Inspiratory flow contour was analyzed qualitatively by two observers according to the criteria proposed by Montserrat and coworkers (2). Disagreements in interpretation, which occurred for less than 5% of the data sets, were resolved by consensus. With this, an FS was calculated and expressed as an integer variable varying from 1 (not limited) to 3 (severely limited). We did not analyze the ventilatory cycles corresponding to FS = 4 (no flow) because in this circumstance RLI and RL values could no longer be calculated.

Data Analysis

Lung and respiratory mechanical parameters were calculated for from one to four 16-s data sets derived from the FO period, irrespective of sleep stage (i.e., in both NREM and REM sleep). To ensure coherent values for respiratory impedance and lung resistances, we discarded from analysis: (1) data sets obtained when the optimal conditions for nCPAP titration were not fulfilled, (i.e., when mouth opening or air leakage at the mask were detected by the occurrence of a shift in the baseline flow signal) (2) data sets showing the presence of esophageal spasms or central or obstructive apneas; and (3) data sets characterized by a highly unstable pressure or flow amplitude. It is worth noting that the occurrence of esophageal spasms or of central apneas did not affect resistive impedance values, and that simultaneous analysis of the flow signal and resistive impedance allowed us to distinguish between central and obstructive apneas.

Statistical analysis of data was done for each patient individually. Comparisons between respiratory and lung parameters were made through correlation analysis. Comparisons with FS were made through analysis of variance (ANOVA). The effect of application or cessation of FO on RLI and RL was assessed through ANOVA. A value of p < 0.05 was considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Whatever the nCPAP level in our study, the mean positive airway pressure was not affected by application of FO. Neither changes in sleep stage nor in chin electromyographic activity nor arousal occurrence were observed after application or cessation of FO (Figure 1). Similarly, no changes were detected in or Pes profiles (Figure 1), and no significant change in either RLI, RL, or EL was observed after application or cessation of FO.

View larger version (41K): [in this window] [in a new window]   Figure 1.   Polygraph recording from a representative patient during a period of severe upper airway obstruction in NREM sleep, and associated plots of resistive pressure (Pres) versus flow over the two ventilatory cycles delimited by the arrows. EEG = electroencephalogram; EOG = electrooculogram; EMG = chin electromyogram; Paw = airway pressure; = nasal flow; Pes = esophageal pressure; RC = rib cage movements; Ab = Abdominal movements. Note that application of FO did not affect sleep, respiratory signals, or the resistive pressure-flow relationship.

The coefficients of the correlations between the different indices of airway patency obtained for each of the 11 patients individually, and their associated p values, are given in Table 2. As illustrated by the r values, RFO correlated strongly with RLI in all patients (Table 2). A typical correlation for RFO with RLI in a representative patient is given in Figure 2. RFO correlated with Pes in only nine of the 11 patients, and in the two patients in whom it did not, no correlation was found between RLI and Pes (Table 2). The p values obtained individually for each of the 11 patients with ANOVA showed that RFO was significantly related to FS in only six patients (Table 2). In four of the five patients in whom RFO was not significantly related to FS, RLI was also not significantly related to FS (Table 2). Similar results were obtained when RLI was replaced by RL, or when data analysis was limited to NREM sleep. Typical examples of changes observed in the conventional indices of airway obstruction, Pes, and RLI that were independent of the inspiratory flow profile are shown in Figures 3 and 4.

View larger version (11K): [in this window] [in a new window]   Figure 2.   Typical relationship of resistive impedance at 16 Hz (RFO) as a function of RLI in a representative patient.

View larger version (24K): [in this window] [in a new window]   Figure 3.   Continuous tracing of nasal flow and Pes, illustrating the occurrence and gradual increase in upper airway obstruction without associated change in amplitude of Pes. RLI = inspiratory lung resistance.

View larger version (22K): [in this window] [in a new window]   Figure 4.   Nasal flow and Pes recordings obtained in the same patient at different moments during the night, and illustrating the occurrence of changes in amplitude of Pes and RLI without associated changes in inspiratory flow contour. RLI = inspiratory lung resistance.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The techniques available for direct measurement of airway resistance, such as plethysmography or airflow interruption, cannot be used to monitor airway patency during titration of nCPAP. Therefore, in the present study, RLI (i.e., airway plus tissue resistance) was chosen as the reference index of airway patency. Our results show that RFO, determined at 16 Hz with the FO technique, is a good noninvasive and quantitative index of RLI and that this index appears to predict upper airway obstruction better than the amplitude of Pes (3) or inspiratory flow contour analysis (1, 2).

Respiratory events during sleep are known to be responsible for arousals, associated sleep fragmentation, and daytime sleepiness. Respiratory obstructive events were initially believed to be confined to apneas and hypopneas, but it was recently observed that arousals could also result from abnormal breathing efforts caused by residual UAO (13). The optimal nCPAP level was then proposed to be characterized by the disappearance not only of apneas and hypopneas, but also of residual UAO (14). Residual UAO is generally detected by analysis of Pes (3) or by the monitoring of RLI (1), both of which require the placement of an esophageal catheter. Therefore noninvasive indices of UAO, based on analysis of the inspiratory flow pattern (1, 2), have been proposed. Condos and associates observed that the behavior of the airway was more closely monitored by analysis of the inspiratory flow contour than with their own estimate of RLI, taken as the ratio of PTP to maximum inspiratory flow (1). Montserrat and coworkers suggested that their FS index, which well predicted the amplitude of Pes, could be used as an alternative in evaluating the therapeutic CPAP level (2). However, objective and quantitative noninvasive indices of upper airway resistance (UAR) might allow better detection of UAO than would changes in the inspiratory flow contour.

The FO technique allows noninvasive measurements of respiratory resistive impedance, whose changes have been shown to fairly reflect those in airway resistance (15). This technique was recently applied to an airway analogue mimicking upper airway collapsibility and submitted to increasing levels of CPAP and it was shown that the amplitude of the analogue impedance was a suitable index for detecting obstruction (16). The present study was therefore designed to demonstrate the reliability of the FO technique in assessing UAO in sleeping patients receiving nCPAP. Toward that end, we compared RFO with RLI. Additionally, we also compared RFO with other indices of airway patency, namely Pes and FS (2).

To evaluate the possible influences of application of FO on sleep and lung resistance without running the risk of failing in nCPAP titration, we had to define a protocol in which FO was applied only once and for a brief period at each pressure level. Neither changes in sleep stage nor the occurrence of arousals was observed when application of FO was started or stopped, probably because of the relatively small amplitude of the forced flows compared with the amplitude of tidal and nCPAP flows. These preliminary results were the requirement for the applicability of the FO technique to titration of nCPAP. Indeed, they are of clinical importance not only with regard to hypnogram validity, but also because changes in sleep stage or arousals may modify upper airway muscle function and thereby affect airway resistance (17). Furthermore, that , Pes, RLI, RL, and EL were unaffected by the application of FO proves that FO per se had no influence on airway muscle tone, which is in accordance with what was previously observed in anesthetized rabbits (18). Our observations might appear at variance with those of other authors (19, 20), who found changes in ventilatory pattern and an increase in genioglossus muscle activity following the application of oscillating pressures. However these authors used pressure waves at a frequency of 30 Hz (i.e., in the frequency range of human snoring), whereas we used a pseudorandom oscillatory flow over the frequency range of 4 to 16 HZ (i.e., well below the snoring range). Moreover, Plowman and Lauff (19) observed that the pressure stimulus was a potentially powerful arousal-promoting stimulus, and Henke and Sullivan (20) did not totally exclude the possibility that the responses they observed could have been partly due to brief arousals. Besides, studies performed in awake normal subjects have found that CPAP decreased peak genioglossus muscle activity (21), and that there was no correlation between genioglossus electromyographic activity and pharyngeal resistance (22).

We chose RFO because it appeared to be a good index of upper airway resistance. Indeed, resistive impedance is roughly constant from 4 to 32 Hz in patients with normal lungs, but exhibits a negative frequency dependence, illustrating gas redistribution, in patients with pulmonary inhomogeneities (23). Most of this frequency dependence occurs below 16 Hz, because respiratory resistive impedance then reflects not only Newtonian resistance, but also the delayed resistance resulting from gas redistribution (24, 25). As the nCPAP level rises, this delayed resistance is likely to decrease as a result of the increase in lung volume, which tends to lessen inhomogeneities. Consequently, resistive impedance at a frequency below 16 Hz might be unsuitable for an accurate evaluation of the effect of nCPAP on upper airway resistance in OSAS patients with diseased lungs. By contrast, and as expected in our patients, who had no history of airway obstruction from asthma or COPD, resistive impedance extrapolated to 0 Hz provided results comparable to those obtained with resistive impedance at 16 Hz. In this connection, the hypoxemia observed in some of our patients can be explained by pulmonary vascular remodeling resulting from OSAS-induced pulmonary hypertension (26) and/or by a slight distal obstruction (i.e., an obstruction resulting in a decrease in distal forced expiratory flow [FEF] rates below 20%) (27).

RLI and RL were determined from resistive pressure only, which was likely to provide a more reliable estimate than that derived from total PTP (1). RFO was assessed over both inspiration and expiration, which might appear debatable, since a study based on a mechanical model suggested that average resistive impedance over the ventilatory cycle might be misleading in clinical applications (28). However, it is worth noting that that our mean lung resistance, RL, increased or decreased in the same way as RLI, which is in accordance with the possibility of upper airway obstruction being both an inspiratory and an expiratory event, as previously suggested (29).

RFO strongly correlated with RLI in all of our patients, which proves that resistive respiratory impedance determined at 16 Hz is a good index of RLI (i.e., airway plus tissue resistance). However, and surprisingly, RFO was a linear function of RLI up to an RLI value of about 25 cm H₂O · L¹ · s, but appeared to asymptotically tend toward a limit beyond this value of RLI (Figure 2). The most plausible explanation for this phenomenon might be the effect observed by Farré and associates in their collapsible airway analogue, that a reduction in tracheal pressure was accompanied by an increase in and plateauing of resistive impedance, whereas flow limitation went on developing (16). Indeed, when flow limitation occurs, impedance data mainly reflect the properties of the flow-limiting segment (28), and a constant elastance-resistance model fails to describe lung mechanical behavior. However, this might not be an impediment to the use of resistive impedance as an index of airway patency during nCPAP titration, since an increasing nCPAP level is accompanied by decreasing flow limitation and a tendency toward linearization of the relationship of RFO with RLI.

RFO correlated with Pes in nine of our 11 patients, and in most of these patients, the quality of the correlation was less than that observed with RLI (Table 2). The shape of the relationship of RFO versus Pes was similar to that of RFO versus RLI. Interestingly, in the patients in whom RFO did not correlate with Pes, RLI also did not correlate with Pes. This tends to demonstrate that although Pes is often used to determine the optimal nCPAP level (2, 3, 7), it is not always a good index of airway resistance (Figure 3) because it reflects not only resistive pressure but also elastic pressure, which may vary with sleep stage and lung volume. Our results, which remained unchanged when we limited our data analysis to NREM sleep, suggest that the changes in elastic pressure would mainly result from those in lung volume.

RFO values could not be predicted from FS in five of our patients, and in four of these five, RLI values also could not be predicted from FS (Table 2). This tends to demonstrate that FS is a less sensitive index of upper airway patency than are pulmonary resistance and swings in Pes (Figure 4). These results are at variance with those suggesting that the inspiratory flow contour could be routinely used as a marker of upper airway patency during nCPAP titration (2), but are in agreement with the observation that progressively more negative esophageal pressures could be associated with constant inspiratory flow (30). Moreover, flow limitation has been observed even in healthy nonsnoring subjects (31).

These latter results therefore suggest that quantitative indices of airway resistance, such as RFO and RLI, yield a more precise evaluation of upper airway patency than does Pes or FS during nCPAP titration.

The small number of patients in our study, and more particularly the small sample of analyzed data for some of them, may be considered the major limit of the study. The wide range in the number of data points (Table 2) is explained by our discarding from analysis data characterized by a highly unstable pressure or flow amplitude, or showing the presence of esophageal spasms, apneas, mouth opening, or air leakage at the mask. This data selection, associated with the small number and short duration of FO application, resulted in failure of the data retained for most patients to be representative of all of the CPAP levels used in the study, which made it difficult to define the resistive impedance criteria characterizing the optimal nCPAP level.

In conclusion, this study demonstrated the applicability of the FO technique to sleeping patients receiving nCPAP, and its reliability in assessing upper airway patency. Indeed, application of FO influences neither sleep nor lung mechanics, permits detection of air leaks or mouth opening, and allows noninvasive measurement of quantitative respiratory resistive parameters that appear to be good indices of lung resistance. Our results strongly suggest that the FO technique may be suitable for nCPAP titration. Further investigations, including monitoring of resistive impedance throughout the night of nCPAP titration, may allow determination of the resistive impedance criteria characterizing the therapeutic level of CPAP.