INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The sleep apnea-hypopnea syndrome (SAHS) is characterized by recurrent episodes of upper airway (UA) closure. UA closure occurs when the UA collapsing forces are not counterbalanced by adequate dilating forces. Because there is no bony or cartilaginous structure to support the pharyngeal airway, the only available dilating forces originate from the adapted contraction of UA dilators muscles. UA dilators activity is indeed a major determinant of both UA patency (UA resistance) and stability (UA collapsibility) (1, 2). This dual role of the UA dilators contraction is illustrated by the phasic decrease in UA pressure that is observed in isolated UA (3) and by the inverse relationship that exists between the pressure at which UA close (critical pressure) and the level of muscle activity (4, 5).

The role of UA dilator muscles in maintaining UA patency and stabilizing UA is particularly important in SAHS patients whose UA resistance and collapsibility are increased both during wakefulness and sleep (6). Several mechanical factors such as differences in UA shape and dimension (9), UA edema (10), abnormalities in the UA activation pattern with decreasing UA pressure (11), and/or abnormalities in the mechanical coupling between UA muscles and perimuscular soft tissues (12) can account for the UA instability observed in these patients during sleep.

The detrimental influence of sleep on UA patency depends on additional specific mechanisms. First, skeletal muscles tonic and phasic activities both decrease during sleep (13) leading to a corresponding decrease in UA volume. Second, sleep may alter the pattern of UA muscles activation. The normal activation of these muscles is known to precede that of inspiratory muscles with a higher rate of rise of electromyogram (EMG) activity (14) that reaches its peak before that of the diaphragm. In sleeping SAHS patients, this normal pattern is observed only during the postapneic ventilatory period. During obstructed breaths, the phasic activation and peak EMG activity of UA muscles frequently follow the onset and peak in the EMG activity of the inspiratory chest wall muscles (15). The critical importance of these abnormalities in UA activation pattern is illustrated by the occurrence of obstructive breathing disorders during sleep in patients with artificially paced diaphragms, in whom by nature UA muscles activity is dissociated from diaphragm activity and no longer precedes or even accompanies it (16). The loss of the coordination between UA muscles and the diaphragm may thus represent a crucial determinant of the pathophysiology of obstructive sleep apneas, but it is a difficult phenomenon to investigate because it is not observed during wakefulness.

To alleviate this difficulty, we reasoned that phrenic nerve stimulation, producing negative intrathoracic pressure through diaphragm contractions could represent a practical way to explore the effects of the dissociation between UA dilators and inspiratory muscles during wakefulness. Furthermore, it could provide a relatively simple means to determine the critical pressure for UA in SAHS patients without resorting to complex sleep studies, as well as to assess and monitor the adequacy of treatments. As a first step to address this hypothesis, we studied UA resistance during diaphragm twitches and the corresponding flow pattern in normal awake subjects. To assess the influence of rib cage distortion on these parameters we compared the effects of two different methods of bilateral phrenic stimulation (electrical stimulation in the neck [ES] and cervical magnetic stimulation [CMS]). We also studied the influence of nasal airway collapse by use of a nasal stent in some of the experiments.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was divided in two protocols. Nine subjects (7 male, 2 female) participated in the first one and seven others (5 male, 2 female) in the second one. Their age range was 22 to 41 yr with a body mass index of 22.8 ± 2.3 kg/m² (mean ± SD). None of them was taking medication or had symptoms suggestive of SAHS. The study was conducted according to local ethical and legal rules, and each subject signed an informed consent form.

Surface recordings of the right and left costal diaphragmatic EMG activity were obtained using silver cup electrodes placed on the axillary line in the seventh to eighth right and left intercostal spaces and connected to an electromyograph (Nihon Kohden Neuropack Sigma/ Nihon Kohden, Tokyo, Japan [Protocol 1] or Biopac system/Biopac, Santa Barbara, CA [Protocol 2]). An esophageal balloon was inserted through one nostril after local anesthesia (xylocaine 2% spray) into the lower third of the esophagus as assessed by the occlusion technique (17). A nasal continuous positive airway pressure (CPAP) mask was placed on the nose, its airtightness being assessed during a maximal inspiratory effort against occlusion. The mask was opened to room air through a pneumotachograph (Hans Rudolph, Kansas City, MO). The diaphragm EMG response (M wave), esophageal pressure (Pes) referenced to mask pressure, mask pressure, and instantaneous flow during 1 s after stimulations were recorded on computer and hard-disk stored after digitization at 50 kHz, for subsequent analysis.

Protocols

Protocol 1: Description and comparison of ES- and CMS-related inspiratory flow patterns. Subjects were seated in an armchair. Their head was maintained in the neural position by headrests. No contraption was used to fix the head to the headrests, but the subjects were asked to carefully maintain contact with them without changing position during the different maneuvers. This was continuously checked by the investigator (F.S.) who performed the stimulations.

All measurements were done with the subject breathing exclusively by the nose. ES and CMS were performed in all subjects. ES was achieved with the electromyograph built-in stimulators according to the technique described in the literature (18). The phrenic nerve was stimulated at the posterior border of the sternocleidomastoid muscle, at the level of the cricoid cartilage, using two bipolar electrodes with saline-soaked felt tips 5 mm in diameter with 2 cm between electrodes. The stimulator was set to deliver a square wave pulse of 0.1 ms duration. The right and left phrenic nerves were spotted by using a low-intensity ES stimulation train at 1 Hz, and a recruitment curve was established to determine the supramaximal stimulation level corresponding to a plateau in the amplitude of the diaphragmatic M waves. ES intensity was then further increased by 10 to 20% to ascertain supramaximality. CMS was performed with a Magstim 200 stimulator using a circular 90-mm coil (S/N 540113 B, maximum output 2.5 Tesla, pulse duration 0.1 ms; Magstim, Whiteland, Dyfed, UK) according to the previously reported technique (19) slightly modified regarding the position of the neck that was kept "natural" (20) rather than flexed. Therefore, the position of the head was the same during ES and CMS. The coil was centered over the spinous process of the seventh cervical vertebra (C7) and the stimulation intensity was set to 100% (Imax). A drawing of the internal circumference of the hole centering the coil was made on the skin to ascertain the exact position of the coil and the absence of neck movement. The adequate positioning of the electrodes and magnetic coil were checked and if necessary readjusted to obtain a maximal M wave. Stimulations were always performed at end expiration during quiet breathing.

Ten stimulations were obtained at the supramaximal level with each technique. Then the intensity of ES or CMS was progressively decreased (5-mA decrement steps for ES and 5% Imax decrement steps for CMS) until M waves disappeared. At least five stimulations were performed at each intermediate level. ES and CMS were done in random order. In four subjects, measurements were repeated 1 to 2 wk apart.

Protocol 2: Effect of nasal patency on ES-related inspiratory flow pattern. In this protocol, only ES was used. The subjects were studied supine with the head supported by a premolded pillow. Electrical pulses were delivered through a stimulus isolation unit (Grass SIU 5A) from a Grass stimulator (S88; Grass Instruments, Quincy, MA). The same experimental procedure as in Protocol 1 was used for phrenic nerve localization, determination of supramaximal stimulation intensity, and stepwise decrease in stimulation intensity. Measurements were done with and without an internal nasal stent (Nozovent; WPM International AB, Göteborg, Sweden) in a random order.

In two subjects genioglossal (GG) EMG activity was recorded with two electrodes mounted on a mouthpiece and in contact with the mouth floor (21). GG EMG signal was amplified (CP 122; Grass Instruments), filtered (10 Hz to 10 KHz), rectified, and integrated with a moving time averager with a time constant of 100 ms (MA 1000; CWE, Ardmore, PA). Nasal pressure was measured in one of these subjects using a catheter tip transducer (S8b; Gaeltec Ltd, Skye, Scotland) that was introduced through one naris after it had been sprayed with xylometazoline hydrochloride 0.05% and anesthetized as described previously. The catheter was positioned in the nasopharynx at 8 cm from the nares.

Data and Statistical Analysis

Resistance of the respiratory system was determined at peak twitch flow and at peak twitch pressure by the ratio driving pressure (esophageal  mask pressure)/twitch flow. Nasopharyngeal (NP) resistance was calculated by the ratio NP pressure-mask pressure/instantaneous twitch flow. Phrenic stimulation-induced inspirations were classified as flow-limited when twitch flow plateaued or decreased in spite of increasing magnitude of the corresponding negative intrathoracic pressure. For each stimulation in each protocol, maximal Pes, peak twitch flow, total pulmonary resistance measured at peak twitch flow and peak Pes were analyzed, and maximal twitch flow of flow-limited twitches was obtained. The values of these variables obtained with ES and CMS (Protocol 1) or with and without nasal stent (Protocol 2) were compared with a multiple analysis of variance. A Bland and Altman plot (22) was used to evaluate the reproducibility over time of measurements (peak twitch flow resistance of all stimulated breaths and maximal twitch flow of flow-limited breaths) obtained at maximal stimulation intensity for each stimulation technique.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Protocol 1: Description and Comparison of ES- and CMS-related Inspiratory Flow Patterns

ES and CMS were successfully achieved in eight subjects; in the ninth one ES could not be performed owing to the impossibility of obtaining a reliable maximal diaphragmatic EMG signal. Stimulation intensity was supramaximal with the two stimulation methods according to the M wave amplitude recruitment curve. ES and CMS induced typical twitch shaped Pes swings and twitch flows  0. 2 s duration (Figure 1). The peak Pes progressively decreased with decreasing intensity of phrenic nerve stimulation with a positive relationship between these variables (individual correlation coefficient range: 0.66 to 0.97, p < 0.05 in all cases). The individual values of the maximal Pes levels reached with ES and CMS are detailed in Table 1. For the whole group, the twitch peak Pes reached at maximal stimulation intensity was significantly higher with CMS than with ES (11.5 ± 3.9 and 9.6 ± 2.6 cm H₂O, respectively, mean ± SD, p = 0.04). A progressive decrease in peak twitch flow hence in twitch total respiratory resistance with decreasing electrical and/or magnetic stimulation intensity was observed in eight subjects, with a positive relationship between these variables (Figure 2). Interestingly, in three males subjects, this increase in resistance reached a plateau at the highest CMS intensities (Figure 2). No difference was found in the total respiratory resistance at peak twitch flow or peak twitch pressure reached with ES and CMS (Table 1). The day-to-day reproducibility of twitch maximal peak flow resistance values obtained with ES and CMS is illustrated in Figure 3A. For each phrenic stimulation method, there was a good agreement between the resistance values obtained at the two visits.

View larger version (6K): [in this window] [in a new window]   Figure 1.   Recording of the Pes and instantaneous flow tracings in Subject 5 during ES. At the lowest stimulation intensity, a parallel increase in flow and driving pressure (Pes less nasal pressure) is observed (A); at the highest stimulation intensity, a typical flow limitation pattern is observed with a decrease in twitch flow in spite of increasing driving pressure (B).

View larger version (18K): [in this window] [in a new window]   Figure 2.   Changes in total respiratory peak flow resistance with increasing stimulation intensity during CMS and ES in a representative subject (No. 5). In both conditions an increase in resistance is observed with increasing stimulation intensity but during CMS resistance plateaued with stimulation intensity higher than 80%.

View larger version (13K): [in this window] [in a new window]   Figure 3.   Bland and Altman plot of total respiratory resistance measured at peak flow and maximal twitch flow of flow-limited breaths at maximal stimulation intensity with both stimulation methods. Horizontal line represents the mean of the measurements obtained at each visit, and dashed lines the 2 SD interval. There was a good agreement between the resistance or VImax values obtained at the two visits with each respective method.

A pattern of inspiratory flow limitation (IFL) was observed whatever the stimulation technique in all but one subject (No. 7) where IFL was only observed with ES but not with CMS. As illustrated in Figure 1, flow-limited breaths typically demonstrated two successive flow profiles: (1) a rapid increase in instantaneous flow that paralleled the rise in Pes, followed by a plateauing and decrease in flow despite increasing intrathoracic pressure; (2) a reincrease in instantaneous flow with decreasing intrathoracic pressure (this second part of the twitch being obviously not flow-limited). IFL was systematically observed at the highest stimulation intensities. The percentage of all twitches that demonstrated a flow-limited pattern was 71.5 ± 27.2 during ES and 48.9 ± 30.0 during CMS (p = 0.07). The effect of increasing stimulation intensity on the maximal twitch flow (VImax) of each flow-limited twitch was analyzed by looking at the individual relationship between VImax and the corresponding Pes. Except in one subject (No. 8), VImax was not correlated with Pes values during ES (Figure 4). During CMS, a positive correlation was observed between VImax and the corresponding Pes in seven of eight subjects who demonstrated IFL during CMS (Figure 4), the severity of flow limitation decreasing with increasing CMS intensity. The average VImax value obtained at maximal stimulation intensity was significantly higher during CMS than ES (1.18 ± 0.29 L · s¹ and 0.94 ± 0.16 L · s¹, respectively, p = 0.04). The average drop in twitch flow after the reach of VImax was 0.41 ± 0.20 L · s¹ with CMS and 0.60 ± 0.19 L · s¹ with ES (p = 10⁴). Figure 3B represents the Bland and Altman plot obtained with the VImax value of IFL breaths measured with each stimulating method. As for resistance values, there was a good agreement between the VImax values obtained with ES or CMS at the two visits.

View larger version (12K): [in this window] [in a new window]   Figure 4.   Relationship between VImax of flow-limited twitches obtained in a representative subject (No. 1) and the corresponding Pes when increasing stimulation intensity during CMS and ES. No correlation was observed between these variables during ES but a positive relationship was noted during CMS.

Protocol 2: Effect of Nasal Patency on ES-related Inspiratory Flow Pattern

Changes in Pes and instantaneous flow with decreasing ES intensity could be assessed in six subjects; in the last one (Subject no. 7), only maximal ES with and without nasal stent were recorded owing to the difficulty of obtaining stable M-wave amplitudes at the different stimulating intensities. For the whole group, no difference was found in the twitch peak Pes reached at maximal stimulation intensity (Table 2). A flow-limited pattern similar to the one described in Protocol 1 was observed in six of seven subjects. In these subjects the percentage of twitch flow classified as flow-limited was 63.8 ± 37.2% without and 56.2 ± 36.9% with stent (p = 0.03). Without a stent, a plateauing of the VImax/Pes relationship was observed in four of six subjects, a positive correlation betweeen these variables being observed in the two others. With a nasal stent, peak twitch flow increased significantly with increasing peak flow esophageal pressure in the four subjects in whom a plateau was observed without stent (Figure 5). The difference in VImax measured at maximal stimulation intensity without and with nasal stent was not significant (0.88 ± 0.13 L · s¹ and 0.90 ± 0.17 L · s¹, respectively, p = 0.3).

View larger version (15K): [in this window] [in a new window]   Figure 5.   Relationship between VImax of flow-limited twitches obtained in a representative subject (No. 1) and the corresponding Pes when increasing stimulation intensity without and with nasal stent. No correlation was observed between these variables without stent but a positive relationship was noted with the stent in place.

In two subjects (No. 3 and 7), GG EMG activity was recorded. As illustrated in Figure 6, UA can be considered as passive during the initial part of the twitch flow, a rise in EMG activity being observed only once a minimal value of driving pressure had been reached; this was followed by a rapid increase in GG EMG activity that generally corresponded to the time the highest driving pressure had been reached and when twitch flow reincreased.

View larger version (11K): [in this window] [in a new window]   Figure 6.   Examples of two different tracings (A, B) of driving pressure, instantaneous flow, average GG EMG activity, and computed nasal resistance obtained at maximal ES intensity without nasal stent in Subject 7.

In Subject 7, NP pressure was measured during the experiment. We found that the average nasal resistance values progressively increased during twitch flow until the nadir of the twitch flow had been reached (NP resistance at 500 ml · s¹: 9.7 ± 1.4 cm H₂O · L¹ · s; at peak flow: 11.1 ± 2.2 cm H₂O · L¹ · s; at peak Pes: 15.2 ± 4.2 cm H₂O · L¹ · s, mean ± SD). As illustrated in Figure 6, the increase in resistance during twitch may vary from one stimulation to another. This Figure brings interesting information: (1) the higher the increase in NP, the lower the VImax even if the driving pressure and the amplitude of the GG EMG response are higher; (2) the decrease in twitch flow after the initial VImax could be seen whether or not NP resistance increased during the decrease in flow. This increase in NP resistance during the course of the twitch was observed with and without nasal stenting. However, in this subject, NP resistance measured with nasal stent at the different flow levels was significantly less than that obtained without stent (NP resistance at 500 mL · s¹: 5.8 ± 0.8 cm H₂O · L¹ · s; at peak flow: 6.6 ± 1.2 cm H₂O · L¹ · s; at peak Pes: 7.6 ± 2.0 cm H₂O · L¹ · s).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We found that in normal awaked subjects, the twitch flow induced by phrenic nerve stimulation is associated with high total respiratory resistance and is often accompanied by inspiratory flow limitation. This is the case in the majority of ES twitches, and in many CMS twitches, although there are differences in the ES- and CMS-related flow patterns.

Mechanisms of Inspiratory Flow Limitation During Diaphragm Twitches

The deleterious effects of phrenic nerve stimulation-induced diaphragm contractions on UA patency have been reported in dogs (23), but species difference and major technical differences (animals anesthetized, tongue secured to the jaw, mouth kept opened, tetanic phrenic stimulation) obviously preclude any direct comparison between the results of the two studies. The animal study showed UA resistance to be dramatically higher when the phrenic nerve stimulation was applied during expiration than following the onset of UA muscle activity, and also demonstrated that in these conditions, phrenic nerve stimulation can induce flow limitation.

The flow limitation pattern that we observed in our normal awaked subjects is markedly different from the one generally reported during sleep in apneic and nonapneic snorers (24, 25) or in normals breathing at subatmospheric pressure levels (22). During sleep, flow limitation results in a tendency for inspiratory flow to reach a plateau as the corresponding inspiratory effort increases. At times, a smooth decrease or fluttering is observed instead of a true plateau. The flow pattern observed during diaphragm twiches in our awaked subjects was different. We observed a first rise in the twitch-associated flow, up to a maximal value. After this peak, flow dropped despite the continuing rise of the twitch inspiratory driving pressure. This was followed by a reincrease in flow contemporaneous with the decreasing phase of the twitch Pes. Assimilating the UA to a Starling resistor, the drop in flow can be accounted for by a continuing increase in upstream resistance when pharyngeal pressure has reached the critical pressure level. It should be noted that a typical flow limitation pattern associated with diaphragm twitches occurred even when nasopharyngeal resistance plateaued (Figure 6) and that the same flow pattern was observed with and without stenting. This suggests a contribution from other factors than increased upstream resistance. Among them is possibly the fact that phrenic nerve stimulation induces a paradoxical inward displacement of the upper rib cage. The magnitude of this displacement is, for a given level of diaphragm activation, accounted for by rib cage distortability (26). Such an inward upper rib cage displacement could interfere with the inspiratory lengthening of the UA walls, a factor known to enhance UA collapsibility (27). This could explain why the nadir of twitch flow usually corresponded to the peak negative intrathoracic pressure and why CMS and ES produced slightly different flow twitch patterns (see below).

It must be noted that even if a characteristic inspiratory flow-limited pattern was observed in the majority of our twitches, the VImax reached during these twitches was usually largely higher than the one observed in normals during sleep at slighly subatmospheric pressure (28). Because no phasic GG EMG activity occurred during the first part of the twitch flow, the difference in VImax characteristics obtained during sleep and in the present study could be accounted for by the sleep-induced decrease in tonic activity of skeletal muscles. Any comparison with previous results published in the literature is difficult because situations associated with a decrease (29) or increase (30) in UA muscles EMG modify both tonic and phasic activities. However, this hypothesis is supported by (1) the decrease in the end-expiratory volume of UA observed during sleep (4), reflecting the importance of tonic EMG activity on UA patency, and (2) the increase in UA resistance after tetanic phrenic nerve stimulation when it is realized after barbiturate administration (21).

Differences in Inspiratory Flow Pattern Between ES and CMS

Several hypotheses can be drawn to explain the observed differences in airflow dynamics between the two methods. The first is that, as mentioned previously, the "pure" nature of the diaphragm contraction after selective stimulation of the phrenic nerve during ES induces a paradoxical movement of the upper rib cage. Such paradoxical movement does not occur, or occurs to a lesser extent, during CMS which has been shown to stabilize the upper part of the rib cage through costimulation of several muscle groups acting on it (e.g., muscles innervated by the spinal nerve or cervical roots giving birth to the brachial plexus) (31, 32). Upper rib cage stabilization is the main factor responsible for the sometimes higher Pes observed in response to supramaximal CMS than to supramaximal ES (of note, the twitch Pes values with ES and CMS that appear in Tables 1 and 2 are markedly lower than published values in comparable subjects because we recorded twitches with the airway open rather than closed as is usually the case). In our study, the drop in inspiratory flow following the reach of VImax was less marked with CMS than with ES, at comparable levels of pressure. This lends support to the hypothesis that differences in rib cage distortability play a role in differences in flow dynamics. It also suggests that using CMS to determine UA critical pressure would provide more realistic information than ES. Indeed, no alteration in the coordination between the diaphragm and extradiaphragmatic muscles normally acting on the upper rib cage to limit its distortability has been reported in non-REM sleep. From a speculative point of view, according to the differences in respiratory muscles' electrophysiological characteristics prevailing during non-REM and REM sleep (dramatic reduction in skeletal muscle tone with preservation in diaphragmatic activity in REM sleep), our results may account for the higher frequency of obstructive breathing disorders observed in REM sleep compared with non-REM sleep.

A second explanation of the CMS-ES differences could be that the stimulation of neck muscles during CMS may unload extrathoracic UA structures. This may in turn stabilize the UA and decrease UA resistance (33) and critical pressure (34).

A third explanation could lie in the direct stimulation by CMS of nerves commanding UA dilator muscles. From previous studies, it is likely that during CMS the magnetic field can reach the anterior part of the thorax, where it stimulates the intrathoracic component of the phrenic nerve hence making phrenic nerve conduction times shorter with CMS than with ES (20). Therefore, the magnetic field could also stimulate the hypoglossal nerve and thus provoke a certain degree of airway dilation. In this regard, it is interesting to note that in six of seven subjects where it could be assessed, twitch flow limitation was observed during CMS, suggesting that whatever the mechanisms underlying the apparent lesser UA collapsibility during CMS, they were not sufficient to prevent some degree of UA collapse and therefore to fully maintain flow.

Site of Inspiratory Flow Limitation during Diaphragm Twitches

Preventing alae nasi collapse significantly decreased the percentage of flow-limited events. However, in the six subjects who exhibited a flow-limited pattern, nasal stenting did not abolish flow limitation. It modified the VImax/ES intensity relationship, from a plateau to a positive correlation in four of six cases and from a negative relationship to a plateau in one case. This suggests that nasal stenting unmasked another site of collapse, and that the increase in nasal resistance only partly contributed to the build-up of maximal instantaneous flow during stimulated breaths, and that the UA segment responsible for flow limitation as seen during twitches is situated lower in the UA.

The present study was not designed to determine the site of airway collapse during phrenic nerve stimulation. Several investigators reported that the transmission of phrenic nerve stimulation-related negative intrathoracic pressure to the mouth could be imperfect, with ES (35, 36) or CMS (37, 38), in normals (35), and in patients with chronic obstructive pulmonary disease (36). This was generally assigned to "partial or complete glottis closure," but the site of UA collapse was never precisely documented. A case report by Scharf and coworkers suggested that phrenic nerve stimulation can induce UA closure below the pharynx: the most important increase in UA resistance observed during phrenic nerve pacing in a quadriplegic patient occurred at the level of the vocal cords with no increase in supralaryngeal resistance (39).

Importantly, breathing route was consistently different between previous reports suggesting problems with pressure equilibration through the airway during phrenic nerve stimulation and our study. In these studies, phrenic stimuli were delivered with the mouth opened, a mouthpiece in place, and the nose clipped (namely, conditions close to mouth breathing). Our study is the only one in which diaphragm contractions were studied in conditions matching nose breathing. Mouth breathing can influence the site of UA collapse by bypassing the most compliant and most resistive part of UA structures (velopharynx). It can reasonably be hypothetized that the collapse observed in our study occurred at the level where UA are the most compliant, namely the nasopharynx. Isono and coworkers (40) have shown that during sleep the nasopharynx is the most compliant UA structure in 86% of subjects with no difference between normals and patients with obstructive sleep apnea. It is possible that this was also the case in our normal awake subjects but the influence of the differences in UA resting tone between wakefulness and sleep on segmental UA compliance is unknown.

Of note, an improvement in pressure equilibration between the pleura and the mouth has consistently been noted when phrenic stimulation was performed during quasi-static inspiratory efforts (36, 37). This suggests that the adaptation of UA muscle tone in response to resistive loading promotes the UA stability.

Perspectives

Inspiratory flow limitation in normal subjects can be observed during dynamic inspiratory efforts associated with important transmural pressure gradients. This is the case during sniffing where alae nasi dilating force is insufficient to maintain nasal patency. The occurrence of flow limitation in healthy young nonobese subjects (devoid of any obvious UA anatomical abnormalities and free of symptoms suggesting UA functional abnormalities) during diaphragm twitches, that are far closer to tidal breaths than sniffs, can be considered a surprising finding. We rather take it as further evidence that the contraction of UA dilators and the precise timing of this contraction are mandatory for normal inspiration to take place. We conclude that (1) phrenic nerve stimulation is a useful tool to evaluate airflow dynamics of the passive upper airways during wakefulness, (2) airflow dynamics may differ between ES and CMS, (3) nasal collapse is partially implicated in the characteristics of twitch airflow during ES. The usefulness of this approach in the diagnosis and management of patients with suspected or documented SAHS remains to be determined.