American Journal of Respiratory and Critical Care Medicine

Arousal is believed to be needed for upper airway opening in obstructive hypopneas–apneas, without compelling evidence to support this notion. The association may be incidental. I studied the temporal relation between arousal and opening and impact of arousal on flow response at opening in 82 patients (apnea–hypopnea index, 46 ± 35/hour). Obstructive apneas–hypopneas were induced by dial-down of continuous positive airway pressure. Obstructions and hypopneas occurred in 44 and 56% of dial-downs, respectively. When arousal occurred (83% of dial-downs), the temporal relation between arousal and opening was inconsistent between and within patients. Frequency of opening without or before arousal increased with milder obstructions (p < 10−9) and with delta power of EEG (p < 10−6). Time of opening was unaffected by whether arousal occurred before or after opening (18.0 ± 9.8 vs. 18.1 ± 10.5 seconds). Flow response was already excessive when opening occurred without or before arousal (180 ± 148% of initial flow decline) and was considerably higher when arousal occurred (267 ± 154%, p < 10−10). Flow undershoot after first ventilatory response was greater if arousal occurred (p < 0.01). It is concluded that arousals are incidental events that occur when thresholds for arousal and for arousal-independent opening are close. They are not needed to initiate opening or to obtain adequate flow and they likely increase the severity of the disorder by promoting greater ventilatory instability.

Arousal from sleep is believed to be an important, if not essential, mechanism for reestablishing airway patency in obstructive sleep apnea (OSA). Originally proposed by Remmers and coworkers (1), this notion was bolstered over the years by the everyday observation that obstructive apneas and hypopneas end abruptly and almost invariably with arousal. The suddenness of increase in flow suggests a discontinuous mechanism, unlike chemical and mechanoreceptor feedback. The concurrent arousal provides a perfect explanation for this discontinuous behavior. This notion is currently so strong that when arousals are not seen at event termination it is presumed that they exist but we failed to detect them.

The preeminence of arousal is based on temporal association between arousal and upper airway (UA) opening. The association may be incidental or even causal in an opposite sense; arousals causing OSA. An incidental association may occur if thresholds for arousal and arousal-independent UA opening are similar. An opposite causal association may result if arousals occur too soon, preempting an orderly compensation by reflex mechanisms. In this case, recurrent cycling (OSA) would occur when arousal threshold is low and would disappear when threshold rises. These alternate possibilities are not without support. (1) Some obstructive events terminate without obvious cortical arousal (28). These have generally been attributed to insensitive arousal identification criteria, occurrence of arousal in unmonitored cortical areas, or to unconventional presentation of arousal (delta, subcortical, autonomic, etc.) (see Berry and Gleeson [4] for review). Perhaps these arousal-free events represent one end of a loose incidental association. (2) Patients frequently alternate between recurrent OSA and stable breathing under conditions where such changes cannot be explained by differences in UA collapsibility (9). (3) Stable breathing tends to occur in delta sleep, when arousal threshold is high (1012). (4) Progressive recruitment of UA dilators before arousal (i.e., without arousal) is well documented in patients with OSA (1, 3, 1316) and level of activity may exceed waking levels (3). (5) Arousal is triggered when a critical level of chemo/mechanoreceptor input is reached (4, 1722). Arousal-independent recruitment of UA dilators is sensitive to the same inputs (16, 2329). Clearly, situations may occur when the input required for triggering arousal and for activating UA muscles enough to open the airway is similar.

In this study, rapid dial-downs of continuous positive airway pressure (CPAP) from a stable baseline were used to induce obstructive apneas–hypopneas in patients with OSA. The role of arousals was examined by: (1) documenting the temporal relation between UA opening and arousal. If arousal is required for UA opening, there should be a predictable relation between the two events under different conditions. Conversely, if the relation is incidental, this relation should be variable. (2) Comparing the magnitude of increase in flow with and without arousal at UA opening. This would establish whether arousals play a helpful or harmful role. A helpful role may be surmised if flow response without arousal is inadequate. Conversely, an adequate or excessive response without arousal that is further augmented by arousal would suggest a harmful role because an excessive ventilatory response helps perpetuate cycling (3033). (3) Determining whether severity of the next hypopnea is greater if arousal occurred at end of the first hypopnea–apnea. The advantages of dial-downs from CPAP over observations on spontaneous obstructive events are outlined in the online supplement.

This is the second report from this study. Earlier (9), I described the effect of passive abnormalities, as reflected by the lowest flow during dial-downs, on severity of OSA. A preliminary report of the current findings was published (34).

A more detailed account of Methods is given in the online supplement.

Patients, protocol, and general methods are identical to those reported earlier (9). Briefly, 82 patients referred for possible OSA were studied using standard polysomnography. CPAP was titrated after obtaining enough information for clinical evaluation. Flow, airway pressure, and polysomnography signals were recorded. Pressure was dialed-down to 1 cm H2O during stable sleep. Dial-down was maintained until arousal or 60 seconds. In approximately one third of interventions, dial-downs were maintained for 2–3 minutes to document the response beyond the first ventilatory phase.


  • Polysomnography before dial-downs: sleep, arousals, and respiratory events were scored using standard criteria (3537). Time spent with stable breathing was defined as sleep periods without respiratory events in excess of 3 minutes (9).

  • Dial-downs. Flow signal was offset by the leak. Mean inspiratory flow (Vt/TI) and tidal volume (Vt) during baseline (60 seconds preceding dial-down) were calculated. Flow and Vt were measured breath by breath after dial-down. Onset of flow response (Tflow) was identified from the point at which there was a clear increase in flow (Figures 1 and 2)

    . The lowest flow observed before Tflow was noted (V̇min). V̇min reflects passive collapsibility (9). Dial-downs were classified according to the relation between Tflow and Tarousal:

Type 1 (Figure 2A): no arousal.

  • Type 2 (Figure 2B): Tflow occurred first but an arousal occurred later between 0.3 and 15.0 seconds.

  • Type 3 (Figure 2C): an arousal occurred before, or within 0.3 seconds after, Tflow.

Definition of Arousals and their Onset (Tarousal)

A more detailed account of Definition of Arousals and their Onset (TAROUSAL) is given in the online supplement.

A combination of EEG inspection and Fourier analysis was used. First, three EEG leads were inspected for a high-frequency shift greater than 3 seconds. If so, Tarousal was the earliest point where the change occurred in any lead. A central EEG lead was analyzed by discrete Fourier transform in 3-second, nonoverlapping epochs spanning baseline (20 epochs) and dial-down. EEG power in the delta, theta, alpha, sigma, and beta ranges was computed. Confidence interval of 95% in baseline epochs was calculated for each frequency band. An arousal was deemed present at Tflow if a visible arousal that meets conventional criteria (36) was present or if total high-frequency power (7.3–35.0 Hz) in the Tflow epoch was greater than the upper bound of the 95% confidence interval.

In several cases, total high-frequency power at Tflow was within baseline range but there was a significant increase in one or more frequency ranges. Classification of dial-down as type 1 to 3 was not altered. However, to assess the impact of such isolated frequency changes, analysis of type 1 and 2 responses was done with and without inclusion of these dial-downs.

Additional Measurements

Latencies to Tflow and Tarousal: interval between onset of dial-down and Tflow and Tarousal.

  • Heart rate at Tflow: heart rate was measured from the beat straddling Tflow and at the corresponding time in inspiration (to account for sinus arrhythmia) in the two preceding breaths.

  • Arousal intensity at Tflow was rated by a single observer, on a scale of 0 to 4 (0 = no arousal), using a subjective scale (see Figure E2 in the online supplement).

  • Flow response at Tflow (ΔV̇1) was calculated from [V̇Br1 − V̇min] (Figure 1). (ΔV̇1) was expressed in liters per second and as the ratio of initial reduction in flow [flow response ratio = ((ΔV̇1/(baseline Vt/TI − V̇min))). This ratio reflects effectiveness of Breath 1 response in restoring flow to normal. The impact of arousal intensity on flow response was assessed by analysis of variance. Similar calculations were done for the next breath (ΔV̇Br2) in type 1 and type 2 observations.

  • The tendency for self-perpetuation was assessed as follows (Figure 1). The lowest flow beyond the first ventilatory response in long dial-downs was measured (V̇min2). Severity of the second hypopnea (baseline Vt/TI − V̇min2) was expressed as a fraction of severity of first hypopnea (baseline Vt/TI − V̇min) (severity ratio). Ratios of 0 and 1.0 indicate, respectively, no tendency and maximum tendency for self-perpetuation.

Results are expressed primarily as those of individual dial-downs. Where appropriate, average of results in all dial-downs in individual patients was computed (e.g., average response type, average flow response ratio, etc.). All values are mean ± SD except where indicated.

Patients were predominantly middle-aged, obese men (Table 1)

TABLE 1. Patient demographics and results of polysomnography

Mean ± SD

Age, yr49.1 ± 11.425–74
Sex66 men, 16 women
BMI, kg/m233.9 ± 5.924–55
Neck circumference, cm44.0 ± 4.134–58
Average O2sat during polysomnography, %95.1 ± 2.489–99
Minimum O2sat (%)79.9 ± 12.650–96
AHI, per hr46.0 ± 35.02–146
PETCO2 during sleep on CPAP, mm Hg49.0 ± 2.937.4–53.3
O2 saturation during sleep on CPAP, %
97.0 ± 2.0

Definition of abbreviations: AHI = apnea–hypopnea index; BMI = body mass index; CPAP = continuous positive airway pressure; PETCO2 = end-tidal carbon dioxide pressure.

n = 82.

. Overall apnea–hypopnea index was 46.0 ± 35.0/hour. Although apnea–hypopnea index varied widely (2–146, Table 1), the type of respiratory disturbance was obstructive in all patients. Other polysomnographic findings are shown in Table 1 (see online supplement for a more detailed account of Results).

A total of 525 dial-downs were examined. Of these, 138 (26%) were in REM sleep. V̇min percentage varied considerably (Table 2)

TABLE 2. Some characteristics of dial-downs (n = 525)


% of Dial-downs
min, % baseline*
> 75214
Response type
Type 18917
Type 211622
Type 3

*min (% baseline) = minimum flow reached during dial-down.

0% = complete obstruction.

. On average, it was 20.2 ± 25.2%. Latency to Tflow was 18.1 ± 10.6 seconds, and latency to Tarousal in type 2 and 3 observations was 17.7 ± 11.0 seconds.

Frequency of Different Response Types

The frequency of types 1, 2, and 3 was 17, 22, and 61%, respectively (Table 2). Figure 2 shows examples of the three response types. Additional examples are shown in Figures E3–E5 in the online supplement. Figure 3

shows the frequency distribution according to time difference (ΔT) between Tflow and Tarousal. There was a peak (32%) at zero (±0.3 seconds). In a substantial percentage of dial-downs (29%) arousal did not trigger an increase in flow immediately (Figure 3, e.g., Figure 2C and Figure E5 in the online supplement).

The relation between Tflow and Tarousal was inconsistent in individual patients. A total of 64 patients (78%) had different types at different times (see Figure E6 in the online supplement). Only 18 patients (22%) had exclusively type 3 responses. These patients were not different from the other 64 patients regarding age, sex, body mass index, or apnea–hypopnea index. However, they had a more collapsible airway (V̇min%: 7.6 ± 17.9 vs. 27.4 ± 29.4, p < 0.005).

Determinants of Response Type

Figure 4A

shows the probability of different response types as a function of severity of obstruction (V̇min). With complete obstruction in NREM, very few type 1 responses were seen (8%). As hypopnea severity decreased, the proportion of type 3 decreased and that of type 1 increased (p < 10−9, χ2 test).

Figure 4B shows the impact of delta power at baseline on the proportion of different types. At the lowest delta power in NREM (13 dB, comparable to Stage 1) there were few type 1 responses. The proportion of type 3 responses decreased, whereas that of type 1 increased as delta power increased (p < 10−6, χ test). At 26 dB (deep Stage 4), the frequencies of types 1 and 3 were comparable. The effect of delta power was continuous throughout the range of NREM sleep, with most of the change occurring in the range of Stage 2 (∼ 16–22 dB). There was no correlation between delta power and V̇min% (r = 0.015), indicating that the observed changes were not due to differences in mechanical severity.

At the same V̇min, there were significantly more type 3 responses and less type 1 responses in REM than in NREM sleep (p < 0.001; open symbols, Figure 4A). With complete obstruction, virtually no type 1 responses were seen. However, there was no difference in type distribution between REM and NREM when the data were plotted as a function of delta power (open symbols, Figure 4B).

Stepwise regression was performed to identify factors that correlate with probability of different types. Delta power, V̇min, and body mass index had significant independent effects (p < 10−12, p < 10−21, and p < 0.002, respectively). REM sleep had no effect after allowing for delta power and V̇min. Age, sex, body position, clock time, and baseline values of heart rate, V̇E, and O2 saturation also had no effect. The regression equation (r = 0.49, p < 10−30) was:

An almost identical regression was obtained when the average values of response type, V̇min%, and delta power obtained in each patient were used (p < 10−7):
Thus, a nonobese patient in REM or Stage 1 sleep (delta = 14 dB) who develops a complete obstruction (V̇min = 0) would have a response type of 3.0 (all events are type 3). In contrast, an obese patient (body mass index = 40), in deep Stage 2 or delta sleep (delta = 23 dB) who develops a mild hypopnea (e.g., V̇min = 70%) would have a response type of 1.5 (i.e., 50% type 1, 50% type 2 and no type 3, or 75% type 1, 25% type 3, etc.).

Impact of Arousal on Tflow

For all data, latency to Tflow was not different among types (17.7 ± 9.7, 18.6 ± 11.1, and 18.0 ± 10.6 second, p = 0.82 by analysis of variance). In contrast, latency to arousal was longer in type 2 than in type 3 (21.6 ± 11.7 vs. 16.4 ± 10.3 second, p < 10−5). These findings suggest that UA opening would occur at the same time, regardless of when arousal occurs.

This potentially important finding was tested further such that each patient was his/her own control. Patients who had type 3 responses and either type 1 or type 2 responses were identified. The difference between average latency to Tflow in either type 1 or type 2 and latency to Tflow in type 3 observations was computed in each patient. The same was done for latency to Tarousal in types 2 and 3. In type 2, Tarousal occurred 5.7 ± 7.7 seconds later than in type 3 (p < 10−5, n = 45 patients, Figure 5)

, whereas Tflow was not different (0.4 ± 7.1 second, p = 0.4). V̇min was comparable. Tflow was marginally delayed in type 1 relative to type 3 observations (2.8 ± 11.3 seconds, p = 0.07, n = 35 patients). However, V̇min was, on average, higher by 19 ± 12.4% (p < 10−5), indicating milder hypopneas. Multiple regression with Tflow as the dependent variable versus type, delta power and V̇min revealed no effect of type (p = 0.9) or, importantly, of delta power (p = 0.17) on Tflow. The effect of V̇min was significant (p < 0.0002, coefficient = 0.15 seconds/% baseline flow). Thus, opening occurred at the same time regardless of delta power or whether or when arousal occurred.

Impact of Arousal on Magnitude of Flow Response

First breath (see online supplement). Figure 6

shows the effect of arousal intensity at Tflow on flow response in the first breath showing a response. The response is given in absolute units (L/second) or as the ratio of initial flow decline (right axis). A ratio greater than 1.0 indicates overcompensation. Without arousal (types 1 and 2), flow increased 0.30 ± 0.24 L/second, corresponding to a ratio of 1.80 ± 1.48 (p < 10−11 for difference from 1.0). There was a progressive increase in flow response and flow response ratio as arousal intensity increased (p < 10−25 and p < 10−13, respectively, by analysis of variance). Response ratio approached four with the most intense arousals.

Second breath in types 1 and 2 (see online supplement). Flow response in Breath 2 varied depending on magnitude of first breath response and whether an arousal developed by Breath 2 (Figure 7)

. When there was still no arousal, flow generally increased when Breath 1 response was incomplete (flow response ratio = 0.94 ± 0.55 vs. 0.67 ± 0.22, p < 0.001, Figure 7, lower dashed line) and did not change when Breath 1 response was excessive (Figure 7, upper dashed line). In contrast, when arousal developed by Breath 2, flow was higher regardless of whether Breath 1 response was incomplete or excessive (solid lines, Figure 7). Flow response ratio of Breath 1 was no different whether an arousal was imminent (Figure 7, solid symbols) or not.

Impact of Arousal on Response Beyond the First Ventilatory Phase

Figure 8

shows the spectrum of responses in long dial-downs (n = 182). In the top tracing, V̇min2 was not lower than baseline mean inspiratory flow (severity ratio = 0). In the middle tracing, a second hypopnea developed, but it was less severe than the first (severity ratio 0.5). In the bottom tracing, the second hypopnea was as severe. The entire spectrum of ratios (0–1.0) was seen with and without an intervening arousal, but the average ratio was greater after an arousal (0.76 ± 0.38 vs. 0.60 ± 0.39; p < 0.01). Severity ratio correlated with intensity of arousal (r = 0.25, p < 0.001). With Level 4 arousal, severity ratio was 0.93 ± 0.19.

Heart Rate at Tflow in Types 1 and 2

Rather than being higher, heart rate at Tflow in types 1 and 2 (63.2 ± 9.0/minute) was lower than in the preceding two breaths (63.9 ± 10.1 in the immediately preceding breath and 64.3 ± 10.7/minute two breaths back; p < 0.05 by analysis of variance for repeated measures).

EEG and EMG Changes at Tflow in Types 1 and 2

(See online supplement for a more detailed account of EEG and EMG Changes at TFLOW in Types 1 and 2)

In 30% of type 1 and 2 observations (i.e., no high-frequency arousal), delta power was significantly higher at Tflow than at baseline. In an additional 15% there were significant increases in one or more higher frequency bands but not enough to increase total high-frequency power (see Table E1 in the online supplement). The impact of these changes was assessed by comparing flow response ratio and heart rate in observations with (45%) and without such changes. Flow response ratio was not different (1.88 ± 1.59 vs. 1.73 ± 1.39; p = 0.45, see also Figure E7A in the online supplement). Heart rate was lower at Tflow than in the preceding two breaths in both cases, and the magnitude of reduction was similar (see Figure E9 in the online supplement).

Changes in EMG were infrequent in type 1 and 2 observations (see online supplement).

The main findings are: (1) While arousal and UA opening occur, on average, at the same time, the temporal relation between the two events is inconsistent within and between patients. (2) UA opening occurs at the same time regardless of whether arousal occurs before or after UA opening or does not occur at all. (3) The increase in flow at UA opening is already excessive in the absence of arousal. Thus, arousal is not required for an adequate flow response. (4) The effect of arousal on UA function is not manifest until there is a clear high-frequency change in EEG. (5) The occurrence of cortical arousal at UA opening is associated with a greater flow overshoot and a greater subsequent undershoot.

Were Arousals Really Absent at Tflow in Types 1 and 2?

Several previous reports noted the occurrence of UA opening without overt arousal (28). A number of explanations were proposed that would still be consistent with arousal-mediated opening (see Berry and coworkers [4] for review). What follows is a discussion of these possibilities and how they were addressed in this study.

  1. An increase in EEG high frequency may be present but is equivocal or it lasts less than 3 seconds. This was addressed by comparing EEG power spectrum at UA opening with 20 baseline epochs. Arousal was called when high-frequency content was significantly higher than baseline even if the change did not meet conventional criteria. This analysis would detect arousals even if the increase in high frequency involved a fraction of the epoch, so long as the change is outside the range observed in baseline.

  2. Arousal may take the form of an increase in slow wave activity. Some investigators noted an increase in delta power near the end of obstructive events (8, 38). The significance of this phenomenon is unknown. Some believe it is an early or mild form of arousal (see Sforza and coworkers [39] for a review). Others believe it represents accelerated progression to deep sleep in sleep-deprived patients (38). An increase in delta was found in 30% of observations classified as type 1 or 2 (see Table E1 in the online supplement). There was, however, no difference in flow response whether such changes were present or not (see Figure E7A in the online supplement). Thus, even if delta increases are considered as arousals, they cannot be responsible for UA opening.

  3. Arousals may be present without any cortical changes (subcortical or autonomic arousals). This possibility was excluded by demonstrating that heart rate was not higher at UA opening (in fact it was lower) than in the preceding breaths. The evidence for subcortical arousals stems from observations that external stimuli (e.g., acoustic) can increase heart rate and blood pressure even without cortical arousals (40, 41). Although heart rate and blood pressure responses to external stimuli may vary in individual subjects, the average responses invariably involve an increase in both variables (4144). Thus, lack of an average increase in heart rate at UA opening in types 1 and 2 effectively rules out autonomic arousals at UA opening (an increase in blood pressure may still occur at UA opening, however, because of the increase in intrathoracic pressure [decreased cardiac afterload] and reduction in pulmonary vascular resistance as a result of improvement in alveolar gas tensions). Because respiratory muscle responses with subcortical arousals are considerably less than the autonomic response (compare ventilatory response to subcortical arousal in Badr and coworkers [45] with autonomic response in the same subjects in Morgan and coworkers [41]), lack of an autonomic response makes it extremely unlikely that UA opening in these cases was produced by subcortical arousals. Two observations from this study further support this conclusion:

  4. First, flow response in Breath 1 was similar whether cortical arousal was imminent (arousal in Breath 2) or not (Figure 7). To the extent that activity in subcortical mechanisms must have been higher when arousal was imminent, this lack of difference suggests that subcortical arousal mechanisms do not exert an important effect on UA dilators.

  5. Second, flow response in types 1 and 2 was, on average, 52% of the response in type 3 (data from Figure 6). It is difficult to explain this magnitude by arousals that could not be seen, could not be detected by Fourier analysis, and that were not strong enough to mount a pressor response. Basner and coworkers (46) found that acoustic stimuli applied during spontaneous obstructive apneas decreased their duration by 2 seconds when no overt (> 3 seconds) cortical arousal resulted. This was only 20% of the reduction produced by stimuli with overt arousal. This value of 2 seconds overestimates the potency of subcortical arousals; in most cases classified as “no arousal,” there were cortical changes that did not meet standard criteria (see discussion in Basner and coworkers [46]). Thus, the potency, if any, of subcortical mechanisms in opening UA is miniscule.

  6. Cortical arousals may be present in unmonitored areas. See online supplement.

  7. In a minority of type 1 and 2 observations, there were changes in one or more high-frequency bands, but they were not sufficient to increase total high-frequency power above baseline (see Table E1 in online supplement). These are almost certainly analytical or technical artifacts (see online supplement). Furthermore, neither flow response (see Figure E7A in the online supplement) nor heart rate response (see Figure E9 in the online supplement) was different in these cases from responses where there were no EEG changes at all.

Frequency of Different Response Types

(A more detailed account of Frequency of Different Response Types is given in the online supplement.)

The frequency of type 1 response (17%, Figure 3) is well within previously reported incidence of obstructive events ending without arousal (2, 3, 5, 7, 8). Dingli and coworkers (7) noted that spontaneous obstructive events occurring in delta sleep are less frequently associated with visible arousals. Current results extend this finding by showing that frequency of type 1 response increases continuously with delta power, with most of the change occurring in the delta range of Stage 2 sleep (16–22 dB, Figure 4B). Thus, in some patients, a small change in delta power within Stage 2 may cause arousals to disappear.

There are no formal reports of frequency of type 2 responses, although the occurrence of arousal after UA opening is a common finding in sleep records and is evident in some published records (e.g., figures 1 and 2 in Stradling and coworkers [47]). Frequency of type 2 responses showed no consistent relation with delta power or obstructive severity (Figure 4), suggesting that this type is transitory. As delta power increases, or obstructive severity decreases, some type 3 responses become type 2, whereas some type 2 responses become type 1, leaving frequency of type 2 responses unchanged.

Evidence for Nonessentiality of Arousal

This study produced much evidence that arousal is not essential for UA opening:

  1. The relation between Tflow and Tarousal was quite inconsistent (Figure 3). If arousal causes opening, why does it occur sometimes before and sometimes after opening, even in the same patient? Most patients (78%) displayed different types at different times and, even within the same response type in the same patient, the difference between Tflow and Tarousal varied considerably (see Figure E6 in the online supplement). This strongly suggests that the association is incidental.

  2. Type 3 was much less frequent when obstruction was milder (Figure 4A). If arousal is required for UA opening with a severe hypopnea, why is it not required with a milder one? That UA opening can occur without arousal at any obstruction severity indicates that the relation between UA dilators' activation and opposing forces is such that the dilators prevail, in due course, without arousal. If so, it is difficult to envision a scenario whereby UA dilators can prevail without arousal at one level of severity but not at a higher one (Figure 9)


  3. Frequency of UA opening without arousal increased with delta power (Figure 4B). Why should arousal be needed in light but not deep sleep? Obstruction was no less severe. Latency to UA opening was also not affected by delta power. Thus, dynamics of compensatory responses were not different. Arousal threshold increases with delta power (1012). The most reasonable interpretation is that arousal-independent compensatory mechanisms are capable of opening the UA, but when arousal threshold is low, arousal occurs first or concurrently.

  4. Dingli and coworkers (7) and Rees and coworkers (2) reported that duration of spontaneously occurring apneas–hypopneas was not different whether or not an arousal occurred (i.e., type 1 vs. types 2 and 3). Perhaps, it may be argued, that in a minority of events (i.e., type 1) arousal-free compensation is possible but in the majority (types 2 and 3) arousal is needed. However, in this study latency to opening was the same whether arousal occurred before (type 3) or after (type 2) opening (Figure 5). What was different was latency to arousal (Figure 5). The inescapable conclusion is that arousal is irrelevant to when UA opening occurs.

It follows that if a patient opens UA without arousal even once, he/she will have demonstrated an ability to open without arousal under other conditions if arousal does not occur first. That 78% of patients developed type 1 or type 2 response in at least one dial-down (see Figure E6 in the online supplement) implies that the great majority do not need arousal to open the UA.

Only 18 patients (22%) had exclusively type 3 responses. Ten of these, however, developed lengthy periods of stable breathing (> 50% of sleep time) during polysomnography in body positions where dial-downs showed complete obstructions (n = 8) or moderate hypopneas (n = 2). This is unequivocal evidence that a patient can mount effective compensation without arousal. Thus, only eight patients (10%) failed to open the UA without arousal at any time. It is possible that a small minority needs arousal. It is, however, also possible that in such patients arousal threshold was consistently low or that the mechanical abnormalities were sufficiently severe that it was difficult to mount the necessary dilator recruitment without reaching arousal threshold first.

In summary, the present findings indicate that in the vast majority, if not in all, of the patients, arousal is not required for UA opening.

An Alternate Scheme for Relation Between Arousal and UA Motor Control

(A more detailed account of An Alternate Scheme for Relation Between Arousal and UA Motor Control is given in the online supplement.)

Figure 10

presents a schema that accounts for current and previous findings. The arousal mechanism (4, 1722) and UA motoneuron pool (16, 2329) receive inputs from mechanoreceptors and chemoreceptors. The excitatory input required for arousal (respiratory arousal threshold) varies from time to time and among patients (4, 12, 17, 19, 48, 49) (Figure 10, dashed lines). The excitatory input to UA motor pool required to open the airway (UA opening threshold) varies with severity of mechanical abnormalities (see online supplement). In patients with OSA, the two thresholds are, on average, similar. This accounts for the fact that latency to UA opening is, on average, the same with or without arousal (Figures 3 and 5). However, because the two thresholds are subject to different influences, opportunity exists for UA opening to occur first, and vice versa. By increasing UA opening threshold, more severe obstructions increase probability of arousal occurring first (Figure 4A). Conversely, by increasing arousal threshold, deeper sleep increases probability of UA opening occurring first (Figure 4B).

To the extent that the range of the two thresholds may follow a normal distribution around their average values, the above schema would result in a normal distribution of the relation between Tflow and Tarousal. The distribution of [Tflow − Tarousal] fits this description with two exceptions (Figure 3). First, a second peak exists related to type 1 response. This is readily explained by the fact that when UA opening occurs first mechanoreceptor feedback should immediately decrease even though chemoreceptor input continues to rise for several seconds (circulatory delays). Net excitatory input to arousal mechanism may decrease, or not increase sufficiently, beyond UA opening, thereby averting arousal. Arousal still occurs if the reciprocal changes in the two inputs soon after UA opening result in sufficient increase in net excitatory input to reach arousal threshold (see also Discussion re Input B, below).

Second, the percentage of observations in which Tarousal and Tflow occurred within ± 0.3 seconds of each other (32%, Figure 3) was excessive, suggesting the presence of additional features that help synchronize the two events. Two such features are proposed. Input A (Figure 10): when arousal occurs first, an additional excitatory input is conveyed to the UA motoneuron pool (Figure 10, arrow A). This helps bring this pool to threshold. Such excitatory input cannot be large, however. If it were, UA opening would occur whenever arousal occurred first regardless of how close to threshold UA pool is. This was not the case (Figure 3, bars to right of zero). The current findings (Figure 3) can be explained if Input “A” is proportional to arousal intensity, as supported by data shown in Figure 6. Here, a mild arousal would bring the UA pool to threshold only if excitability of the pool is near threshold. Synchrony is achieved, but there is not much reduction in latency to UA opening. If UA pool were far from threshold, only an intense arousal would succeed in triggering UA opening. Input B (Figure 10): when UA opens before arousal throat vibrations, snoring sound or sudden respiratory unloading provide extra inputs to the arousal mechanism that may bring it to threshold (e.g., Figure 2B).

The main difference between the proposed schema (Figure 10) and current concepts is in the emphasis placed on Input A. At present it is believed that Input A is essential in the vast majority of events. In the current schema, reflex mechanisms are independently capable of opening UA. Input A is only a relatively weak synchronizing input. UA opening would occur at approximately the same time regardless of whether and when arousal occurred.

Pattern of Flow Response at UA Opening

(A more detailed account of Pattern of Flow Response at UA Opening is given in the online supplement.)

One of the most distinctive features of obstructive events is the abruptness with which they end. This clearly suggests a discrete event at UA opening. When coupled with arousal, the impression that arousal caused UA opening is almost unavoidable. In this study, flow increased abruptly whether or not arousal was present. The increase in flow in Breath 1 without arousal was 0.30 ± 0.24 L/second (Figure 6), larger than total flow demand (116 ± 26% of Vt/Ti on CPAP), with little further increase unless arousal occurred (Figure 7). Thus, an abrupt increase in flow need not reflect a discrete neurological event such as arousal.

The mechanism of abrupt increase in flow in the absence of arousal is not clear. It could be neurogenic, mechanical, or both. A neurogenic mechanism would entail a reflexly mediated disproportionate (versus time) increase in dilator activity at UA opening. A mechanical mechanism would produce this effect if the relation between rising dilator activity and UA area were discontinuous, for example, because of viscid secretions between apposed surfaces. Further studies are needed to clarify the mechanism (see online supplement for additional discussion).

Impact of Arousals on Magnitude of Flow Response

Flow response was higher when arousals occurred (Figures 6 and 7). Possible mechanisms include: (1) Arousal provides an excitatory input to UA motoneurons (Input A, Figure 10) (50, 51). (2) Diaphragm activity increases during transient arousal (32, 52, 53). This would have no effect if flow were limited to a very low level. However, if maximum flow were to increase sufficiently through the first mechanism, the arousal-induced increase in diaphragm activity would increase flow further. The magnitude of this effect was directly related to arousal intensity (Figure 6). Thus, arousal intensity is an important determinant of magnitude of ventilatory overshoot and, by extension, likelihood of recurrent cycling. The ideal response at UA opening is one in which flow rises to a level just below baseline flow but remains limited. In this fashion, some increase in chemical drive (i.e., negative intrathoracic pressure) and high UA resistance are maintained. This combination is required to generate a negative pharyngeal pressure sufficient to maintain UA dilator activity at an adequate level, thereby making it possible to attain a steady state. A flow overshoot at UA opening reduces the chance that Pco2 at the end of the ventilatory phase will be higher than unloaded steady-state Pco2, and it may in fact be lower. The lower the chemical stimulus at the end of the ventilatory phase the higher the likelihood that stimulus for dilator activation will become inadequate causing recurrence of obstruction.

Is the Arousal-related Increase in Flow Helpful or Harmful?

It may be argued that if arousal is not necessary for UA opening it is necessary to obtain adequate flow. Current results show that this is not true. Without arousal, flow response during Breath 1 was, on average, 180% of the initial decline (Figure 6, see also Figure E8B in the online supplement). Thus, without arousal there was already an overshoot corresponding to 80% of the initial decline. With arousal, the overshoot was substantially larger (Figures 6 and 7). This cannot be helpful (3033). (See “Impact of Arousals on Magnitude of Flow Response,” above.) That this is so is supported by the observation that the second hypopnea was more severe when arousal occurred during the first ventilatory response.

Do these observations mean that if arousals did not occur there would be no OSA? The answer is clearly no. Large overshoots and subsequent undershoots also occurred without arousal. The current data, however, indicate that the likelihood for self-perpetuating cycling should be reduced if arousals did not occur. The extent to which arousals are responsible for cycling is difficult to ascertain. However, indirect evidence suggests that they play an important role in most patients. The occurrence of periods of stable breathing in the majority of patients with OSA (9) indicates that most patients can mount stable compensation without arousal. Periods of stable breathing tend to occur as sleep deepens. As indicated earlier, deeper sleep neither reduces UA collapsibility nor alters the dynamics of the compensatory response. Thus, the most reasonable explanation is that when arousal threshold is low, arousals preempt orderly compensation and, by augmenting the overshoot, perpetuate cycling. As arousal threshold rises, arousal is delayed, permitting an orderly compensation by reflex mechanisms.

Sensitivity of Mechanisms Responsible for UA Opening

Latency to UA opening, with or without arousal, was 18.1 ± 10.6 seconds. Because dial-downs began from a stable baseline on CPAP, it is possible to estimate the changes in chemical drive, relative to unloaded levels, required for UA opening. An increase in chemical drive underlies the increase in both the chemo- and mechanoreceptor stimuli because mechanoreceptor feedback cannot increase unless pump muscle pressure increases. Latency to Tflow was within one circulation time (∼ 50–60 seconds) in all but a few (∼ 1%) dial-downs. Thus, the maximum increase in PaCO2 at Tflow would be the arteriovenous Pco2 difference (4–6 mm Hg). Chemoreceptors would be exposed to a fraction of that because of transport and diffusion delays. Thus, the increase in Pco2 at the peripheral chemoreceptors at Tflow was likely no more than 2 to 3 mm Hg, and the change at central chemoreceptors was even less. O2 saturation decreased, on average, by 3.1 ± 3.5%. These estimates suggest that the mechanisms responsible for UA opening are remarkably potent in most patients, with or without arousal.

REM Sleep

(A more detailed account of REM Sleep is given in the online supplement.)

When compared at the same V̇min, type 3 responses were more common in REM sleep (Figure 4A). With occlusions, there were virtually no type 1 responses (Figure 4A). This suggests that the balance between arousal threshold and UA opening threshold (Figure 10) is altered in favor of earlier arousal. The frequency of different response types was similar in REM and NREM sleep when compared at the same delta power (Figure 4B). Thus, in REM sleep the balance between the two thresholds is comparable to stage 1 NREM sleep. The unfavorable balance may be due a lower arousal threshold to airway obstruction (54, 55) and/or to inhibitory inputs to UA motoneuron pool during REM sleep (15, 51, 56, 57) (Figure 10, open arrow). The latter would increase the amount of excitatory input required to reach UA opening threshold, effectively increasing this threshold.

The greater likelihood of arousals in REM may explain why some patients have OSA only in REM sleep (58, 59). If arousal threshold is lower than UA opening threshold by a small amount in REM, the balance between the two thresholds may be reversed in NREM sleep because of generally higher delta power and lack of REM-related inhibitory inputs to UA motoneurons. Cycling precipitated by arousals would occur in REM but not in NREM.


That arousal is not necessary for UA opening and likely contributes to OSA severity has some implications.

Risk factors. OSA is a disease of middle-aged and older humans (60). Children and young adults may have less OSA because of generally deeper sleep (61). Likewise, although differences in anatomic severity may account, in part, for why some older humans have OSA, whereas others do not, it is possible that many older subjects suffer from severe mechanical abnormalities but are protected by a relatively high arousal threshold. If so, OSA may legitimately be considered an arousal disorder.

Drug therapy for OSA. Use of sedatives/hypnotics in OSA is severely discouraged primarily because of the conviction that arousals are needed to open the airway and partly because limited trials gave largely negative results (6265). The current findings should in no way encourage the clinical use of hypnotics in patients with OSA. However, they should be a stimulus to reevaluate the current position and to encourage research into potential benefits of different drugs. Not all patients, if any, may benefit. Theoretically, the ideal patient is one who has some stable breathing time during sleep in the supine position and in whom obstructive events are brief and terminated by arousal. This patient will have proven that he/she can mount effective stable compensation, without arousals, for the most severe anatomic abnormalities he/she encounters (supine position). Likewise, not all drugs that raise arousal threshold should eliminate cycling. A drug that concurrently increases, by the same amount or more, the UA opening threshold (through depression of the UA motoneurons or their input), will simply increase the duration of obstructive events without eliminating cycling. The ideal drug is one that raises arousal threshold while not increasing UA opening threshold as much. In this fashion, sufficient separation between the two thresholds is created to permit orderly compensation by the automatic mechanisms. It is possible that in earlier clinical trials the results were negative because the wrong type of patient or drug was used.

The focus of the current research into pharmacotherapy of OSA is to develop drugs that stimulate UA dilators (66). The present study suggests that not all types of enhancement of dilator responses should be effective and some may be counterproductive. Thus, even without arousal, UA opening occurred in a timely manner and the flow response was already excessive in most cases. Agents that increase sensitivity and gain of UA dilators' response when obstructive apneas–hypopneas occur are not likely to be helpful and may exacerbate the problem by promoting an even greater overshoot. The ideal agent would be one that prevents obstructive events outright by increasing tonic activity of these muscles (i.e., analogous to CPAP) while, at the same time, leaving unchanged or even reducing the response gain to avoid self-perpetuating cycling, should the tonic effect not be entirely effective.

In summary, in the vast majority of patients, if not in all patients, arousal is required neither to initiate UA opening nor to obtain adequate flow. UA opening would occur at approximately the same time regardless of when or whether arousal occurs and the flow response in most patients would still be timely and adequate. Arousals are incidental events that occur when the thresholds for arousal and arousal-independent opening are close to each other, as they appear to be in patients with OSA. By promoting an unnecessarily high flow response at UA opening, arousals help perpetuate cycling and likely exacerbate OSA.

The author is grateful to the sleep lab technologists at the Health Sciences Centre in Winnipeg (Wayne Thompson, Warren Shewchuck, Perry Sahni, Colleen Leslie, Tamara Allen, and Tanya Dykes) for conducting the studies and to John Kun for writing computer programs to facilitate the analysis.

1. Remmers JE, deGroot WJ, Sauerland EK, Anch AM. Pathogenesis of upper airway occlusion during sleep. J Appl Physiol 1978;44:931–938.
2. Rees K, Spence DP, Earis JE, Calverley PM. Arousal responses from apneic events during non-rapid-eye-movement sleep. Am J Respir Crit Care Med 1995;152:1016–1021.
3. Carlson DM, Onal E, Carley DW, Lopata M, Basner RC. Palatal muscle electromyogram activity in obstructive sleep apnea. Am J Respir Crit Care Med 1995;152:1022–1027.
4. Berry RB, Gleeson K. Respiratory arousal from sleep: mechanisms and significance. Sleep 1997;20:654–675.
5. Martin SE, Engleman HM, Kingshott RN, Douglas NJ. Microarousals in patients with sleep apnoea/hypopnoea syndrome. J Sleep Res 1997;6:276–280.
6. Gould GA, Whyte KF, Rhind GB, Airlie MA, Catterall JR, Shapiro CM, Douglas NJ. The sleep hypopnea syndrome. Am Rev Respir Dis 1988;137:895–898.
7. Dingli K, Fietze I, Assimakopoulos T, Quispe-Bravo S, Witt C, Douglas NJ. Arousability in sleep apnoea/hypopnoea syndrome patients. Eur Respir J 2002;20:733–740.
8. Dingli K, Assimakopoulos T, Fietze I, Witt C, Wraith PK, Douglas NJ. Electroencephalographic spectral analysis: detection of cortical activity changes in sleep apnoea patients. Eur Respir J 2002;20:1246–1253.
9. Younes M. Contributions of upper airway mechanics and control mechanisms to severity of obstructive apnea. Am J Respir Crit Care Med 2003;168:645–658.
10. Gugger M, Molloy J, Gould GA, Whyte KF, Raab GM, Shapiro CM, Douglas NJ. Ventilatory and arousal responses to added inspiratory resistance during sleep. Am Rev Respir Dis 1989;140:1301–1307.
11. Berry RB, Bonnet MH, Light RW. Effect of ethanol on the arousal response to airway occlusion during sleep in normal subjects. Am Rev Respir Dis 1992;145:445–452.
12. Berry RB, Asyali MA, McNellis MI, Khoo MC. Within-night variation in respiratory effort preceding apnea termination and EEG delta power in sleep apnea. J Appl Physiol 1998;85:1434–1441.
13. Onal E, Lopata M, O'Connor T. Pathogenesis of apneas in hypersomnia-sleep apnea syndrome. Am Rev Respir Dis 1982;125:167–174.
14. Okabe S, Chonan T, Hida W, Satoh M, Kikuchi Y, Takishima T. Role of chemical drive in recruiting upper airway and inspiratory intercostal muscles in patients with obstructive sleep apnea. Am Rev Respir Dis 1993;147:190–195.
15. Okabe S, Hida W, Kikuchi Y, Taguchi O, Takishima T, Shirato K. Upper airway muscle activity during REM and non-REM sleep of patients with obstructive apnea. Chest 1994;106:767–773.
16. Berry RB, McNellis MI, Kouchi K, Light RW. Upper airway anesthesia reduces phasic genioglossus activity during sleep apnea. Am J Respir Crit Care Med 1997;156:127–132.
17. Vincken W, Guilleminault C, Silvestri L, Cosio M, Grassino A. Inspiratory muscle activity as a trigger causing the airways to open in obstructive sleep apnea. Am Rev Respir Dis 1987;135:372–377.
18. Gleeson K, Zwillich CW, White DP. The influence of increasing ventilatory effort on arousal from sleep. Am Rev Respir Dis 1990;142:295–300.
19. Wilcox PG, Pare PD, Road JD, Fleetham JA. Respiratory muscle function during obstructive sleep apnea. Am Rev Respir Dis 1990;142:533–539.
20. Berry RB, Light RW. Effect of hyperoxia on the arousal response to airway occlusion during sleep in normal subjects. Am Rev Respir Dis 1992;146:330–334.
21. Berry RB, Mahutte CK, Light RW. Effect of hypercapnia on the arousal response to airway occlusion during sleep in normal subjects. J Appl Physiol 1993;74:2269–2275.
22. Kimoff RJ, Cheong TH, Olha AE, Charbonneau M, Levy RD, Cosio MG, Gottfried SB. Mechanisms of apnea termination in obstructive sleep apnea: role of chemoreceptor and mechanoreceptor stimuli. Am J Respir Crit Care Med 1994;149:707–714.
23. Wheatley JR, Tangel DJ, Mezzanotte WS, White DP. Influence of sleep on response to negative airway pressure of tensor palatini muscle and retropalatal airway. J Appl Physiol 1993;75:2117–2124.
24. Wheatley JR, Mezzanotte WS, Tangel DJ, White DP. Influence of sleep on genioglossus muscle activation by negative pressure in normal men. Am Rev Respir Dis 1993;148:597–605.
25. Horner RL, Innes JA, Morrell MJ, Shea SA, Guz A. The effect of sleep on reflex genioglossus muscle activation by stimuli of negative airway pressure in humans. J Physiol 1994;476:141–151.
26. Innes JA, Morrell MJ, Kobayashi I, Hamilton RD, Guz A. Central and reflex neural control of genioglossus in subjects who underwent laryngectomy. J Appl Physiol 1995;78:2180–2186.
27. Philip-Joet F, Marc I, Series F. Effects of genioglossal response to negative airway pressure on upper airway collapsibility during sleep. J Appl Physiol 1996;80:1466–1474.
28. Malhotra A, Pillar G, Fogel RB, Beauregard J, Edwards JK, Slamowitz DI, Shea SA, White DP. Genioglossal but not palatal muscle activity relates closely to pharyngeal pressure. Am J Respir Crit Care Med 2000;162:1058–1062.
29. Stanchina ML, Malhotra A, Fogel RB, Ayas N, Edwards JK, Schory K, White DP. Genioglossus muscle responsiveness to chemical and mechanical stimuli during non-rapid eye movement sleep. Am J Respir Crit Care Med 2002;165:945–949.
30. Younes M. The physiologic basis of central apnea and periodic breathing. Curr Pulmonol 1989;10:265–326.
31. Khoo MC, Gottschalk A, Pack AI. Sleep-induced periodic breathing and apnea: a theoretical study. J Appl Physiol 1991;70:2014–2024.
32. Khoo MC, Koh SS, Shin JJ, Westbrook PR, Berry RB. Ventilatory dynamics during transient arousal from NREM sleep: implications for respiratory control stability. J Appl Physiol 1996;80:1475–1484.
33. Younes M, Ostrowski M, Thompson W, Leslie C, Shewchuk W. Chemical control stability in patients with obstructive sleep apnea. Am J Respir Crit Care Med 2001;163:1181–1190.
34. Younes M. Response to de novo upper airway collapse in patients with obstructive sleep apnea [abstract]. Am J Respir Crit Care Med 2003;167:A599.
35. Rechtschaffen A, Kales A. A manual of standardized terminology, techniques, and scoring system for sleep stages of human subjects. Washington, DC: U.S. Government Printing Office; 1968. NIH Publication No. 204.
36. Arousals EEG. Scoring rules and examples: a preliminary report from the sleep disorders atlas task force of the American sleep disorders association. Sleep 1992;15:174–184.
37. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. The report of an American Academy of Sleep Medicine Task Force. Sleep 1999;22:667–689.
38. Svanborg E, Guilleminault C. EEG frequency changes during sleep apneas. Sleep 1996;19:248–254.
39. Sforza E, Jouny C, Ibanez V. Cardiac activation during arousal in humans: further evidence for hierarchy in the arousal response. Clin Neurophysiol 2000;111:1611–1619.
40. Davies RJ, Belt PJ, Roberts SJ, Ali NJ, Stradling JR. Arterial blood pressure responses to graded transient arousal from sleep in normal humans. J Appl Physiol 1993;74:1123–1130.
41. Morgan BJ, Crabtree DC, Puleo DS, Badr MS, Toiber F, Skatrud JB. Neurocirculatory consequences of abrupt change in sleep state in humans. J Appl Physiol 1996;80:1627–1636.
42. Pitson D, Chhina N, Knijn S, van Herwaaden M, Stradling J. Changes in pulse transit time and pulse rate as markers of arousal from sleep in normal subjects. Clin Sci 1994;87:269–273.
43. Catcheside PG, Chiong SC, Mercer J, Saunders NA, McEvoy RD. Noninvasive cardiovascular markers of acoustically induced arousal from non-rapid-eye-movement sleep. Sleep 2002;25:797–804.
44. O'Driscoll DM, Meadows GE, Corfield DR, Simonds AK, Morrell MJ. The cardiovascular response to arousal from sleep under controlled conditions of central and peripheral chemoreceptor stimulation in humans. J Appl Physiol 2003 Oct 24 (Epub ahead of print).
45. Badr MS, Morgan BJ, Finn L, Toiber FS, Crabtree DC, Puleo DS, Skatrud JB. Ventilatory response to induced auditory arousals during NREM sleep. Sleep 1997;20:707–714.
46. Basner RC, Onal E, Carley DW, Stepanski EJ, Lopata M. Effect of induced transient arousal on obstructive apnea duration. J Appl Physiol 1995;78:1469–1476.
47. Stradling JR, Pitson DJ, Bennett L, Barbour C, Davies RJ. Variation in the arousal pattern after obstructive events in obstructive sleep apnea. Am J Respir Crit Care Med 1999;159:130–136.
48. Montserrat JM, Kosmas EN, Cosio MG, Kimoff RJ. Mechanism of apnea lengthening across the night in obstructive sleep apnea. Am J Respir Crit Care Med 1996;154:988–993.
49. Sforza E, Krieger J, Petiau C. Arousal threshold to respiratory stimuli in OSA patients: evidence for a sleep-dependent temporal rhythm. Sleep 1999;22:69–75.
50. Mezzanotte WS, Tangel DJ, White DP. Influence of sleep onset on upper-airway muscle activity in apnea patients versus normal controls. Am J Respir Crit Care Med 1996;153:1880–1887.
51. Jelev A, Sood S, Liu H, Nolan P, Horner RL. Microdialysis perfusion of 5-HT into hypoglossal motor nucleus differentially modulates genioglossus activity across natural sleep-wake states in rats. J Physiol 2001;532:467–481.
52. Khoo MC, Shin JJ, Asyali MH, Kim TS, Berry RB. Ventilatory dynamics of transient arousal in patients with obstructive sleep apnea. Respir Physiol 1998;112:291–303.
53. Horner RL, Rivera MP, Kozar LF, Phillipson EA. The ventilatory response to arousal from sleep is not fully explained by differences in CO2 levels between sleep and wakefulness. J Physiol 2001;534:881–890.
54. Issa FG, Sullivan CE. Arousal and breathing responses to airway occlusion in healthy sleeping adults. J Appl Physiol 1983;55:1113–1119.
55. Gugger M, Bogershausen S, Schaffler L. Arousal responses to added inspiratory resistance during REM and non-REM sleep in normal subjects. Thorax 1993;48:125–129.
56. Issa FG, Edwards P, Szeto E, Lauff D, Sullivan C. Genioglossus and breathing responses to airway occlusion: effect of sleep and route of occlusion. J Appl Physiol 1988;64:543–549.
57. Wiegand L, Zwillich CW, Wiegand D, White DP. Changes in upper airway muscle activation and ventilation during phasic REM sleep in normal men. J Appl Physiol 1991;71:488–497.
58. Kass JE, Akers SM, Bartter TC, Pratter MR. Rapid-eye-movement-specific sleep-disordered breathing: a possible cause of excessive daytime sleepiness. Am J Respir Crit Care Med 1996;154:167–169.
59. O'Connor C, Thornley KS, Hanly PJ. Gender differences in the polysomnographic features of obstructive sleep apnea. Am J Respir Crit Care Med 2000;161:1465–1472.
60. Young T, Peppard PE, Gottlieb DJ. Epidemiology of obstructive sleep apnea: a population health perspective. Am J Respir Crit Care Med 2002;165:1217–1239.
61. Bliwise DL. Normal aging. In: Kryger MH, Roth T, Dement WC, editors. Principles and practice of sleep medicine, 2nd ed., Chap 63, Philadelphia, PA: WB Saunders; 1994. p. 642–656.
62. Cirignotta F, Mondini S, Zucconi M, Gerardi R, Farolfi A, Lugaresi E. Zolpidem-polysomnographic study of the effect of a new hypnotic drug in sleep apnea syndrome. Pharmacol Biochem Behav 1988;29:807–809.
63. Series F, Series I, Cormier Y. Effects of enhancing slow-wave sleep by gamma-hydroxybutyrate on obstructive sleep apnea. Am Rev Respir Dis 1992;145:1378–1383.
64. Berry RB, Kouchi K, Bower J, Prosise G, Light RW. Triazolam in patients with obstructive sleep apnea. Am J Respir Crit Care Med 1995;151:450–454.
65. Hedner J, Grunstein R, Eriksson B, Ejnell H. A double-blind, randomized trial of sabeluzole–a putative glutamate antagonist–in obstructive sleep apnea. Sleep 1996;19:287–289.
66. Veasey SC. Pharmacotherapeutic trials for sleep-disordered breathing. In: Pack AI, editor. Sleep apnea: pathogenesis, diagnosis, and treatment. New York: Dekker; 2002, Lung Biology in Health and Disease. Vol. 166, Chap. 21. p. 607–622.
Correspondence and requests for reprints should be addressed to Magdy Younes, M.D., Ph.D., 3611-55 Harbour Square, Toronto, ON, Canada M5J 2L1. E-mail:


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record