This Article Abstract Full Text (PDF) Online Supplement All Versions of this Article: 200302-201OCv1 168/6/645    most recent Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Younes, M. Search for Related Content PubMed PubMed Citation Articles by Younes, M.

Contributions of Upper Airway Mechanics and Control Mechanisms to Severity of Obstructive Apnea

Department of Medicine, University of Manitoba, Winnipeg, Manitoba; and Department of Medicine, University of Toronto, Toronto, Ontario, Canada

Address correspondence and requests for reprints to Magdy Younes, 3611–55 Harbour Square, Toronto, ON, M5J 2L1 Canada. E-mail: mkyounes{at}sympatico.ca

	ABSTRACT

TOP ABSTRACT THEORETICAL CONSIDERATIONS METHODS RESULTS DISCUSSION APPENDIX 1 APPENDIX 2 APPENDIX 3 APPENDIX 4 REFERENCES

 
The contributions of pharyngeal mechanical abnormalities, flow demand, and compensatory effectiveness to obstructive sleep apnea severity were determined in 82 patients. Flow demand was estimated from mean inspiratory flow on continuous positive airway pressure. Mechanical load on upper airway muscles was estimated from minimal effective continuous positive airway pressure, flow demand, and minimum flow observed during brief pressure dial downs. Compensatory effectiveness was estimated by relating polysomnographic severity and mechanical load. Mechanical load was more severe in men, in supine position, and in older and heavier patients. Higher flow demand contributed significantly to mechanical load in men and in those who are obese. At the same mechanical load, severity was independent of age, sex, or body mass index but was greater in the supine position and in REM sleep. Mechanical load accounted for only 34% of variability in severity. Eighty-two percent of patients experienced periods of stable breathing despite mechanical loads that would produce continuous cycling without compensation. I conclude that most patients can adequately compensate for the abnormal mechanics, at least part of the time. Higher flow demand contributes to severity in men and in obesity. Severity is largely due to factors other than mechanical load. Compensatory effectiveness is impaired in the supine position and in REM sleep, but not by age, sex, or body mass index.

Key Words: age • sex • supine position • body mass index • REM sleep

Patients with obstructive sleep apnea/hypopnea (OSA) have, on average, a narrower and more collapsible upper airway relative to others (1–5). They may overcome the abnormal upper airway mechanics during wakefulness by increased activation of upper airway dilators (6) but develop OSA in sleep because of a reduction in basal upper airway muscle activity (7) and attenuation of the mechanisms involved in activating upper airway dilators in response to negative pressure (8–10). Patients with abnormal passive upper airway mechanics are more dependent on upper airway muscle activation to maintain upper airway patency and, hence, are more vulnerable to sleep-related changes in neuromuscular control of pharyngeal muscles (11).

An issue of considerable importance is whether the polysomnographic severity of OSA in a given patient is principally related to the severity of abnormalities in passive upper airway mechanical properties or is a reflection of how much control mechanisms are impaired by sleep. One approach to addressing this issue is to plot the relationship between the mechanical load imposed on upper airway dilators by passive upper airway properties and an index of polysomnographic severity. A greater polysomnographic severity at the same mechanical load denotes less effective compensatory mechanisms and vice versa.

In this study, rapid dial downs of continuous positive airway pressure (CPAP) were used to assess the mechanical load imposed on upper airway control mechanisms by passive upper airway properties. Specifically, and as justified in THEORETICAL CONSIDERATIONS, the lowest inspiratory flow rate observed soon after dial down to near atmospheric pressure (min) is used as an index of passive collapsibility. From min and flow demand (mean inspiratory flow on CPAP), CPAPmin, an estimate of the pressure that upper airway dilators need to generate to permit flow rates that are consistent with stable breathing, is determined. Compensatory effectiveness is determined from the relationship between CPAPmin and the fraction of sleep time during which breathing is stable (T_stable/ST); a higher T_stable/ST at the same CPAPmin reflects greater compensatory effectiveness and vice versa.

The relationships between min and CPAPmin on one hand and T_stable/ST on the other were determined in different body positions and sleep states in 82 patients. These correlations were used to obtain the overall relationship between mechanical load and polysomnographic severity and to identify the mechanisms by which age, sex, body mass index (BMI), body position, and REM sleep affect OSA severity.

	THEORETICAL CONSIDERATIONS

TOP ABSTRACT THEORETICAL CONSIDERATIONS METHODS RESULTS DISCUSSION APPENDIX 1 APPENDIX 2 APPENDIX 3 APPENDIX 4 REFERENCES

 
What Is an Appropriate Index of Polysomnographic Severity of OSA?
Although in some patients repetitive obstructive hypopneas and apneas (cycling) occur throughout sleep, most patients display a mix of periods of stable breathing and periods in which breathing is cycling. In this study, the effectiveness of control mechanisms involved in countering upper airway collapse during sleep (henceforth called "compensatory effectiveness") will be assessed using a new index, namely the time spent in stable sleep (T_stable) as a fraction of sleep time (i.e., T_stable/ST). This approach is justified as follows: Periods of stable breathing indicate effective compensation requiring (1) the ability to increase upper airway dilator activity to a level that permits a sustainable level of ventilation without arousals and (2) a response that occurs in an orderly fashion without ventilatory overshoot to perpetuate cycling (12). That patients alternate between stable breathing and repetitive cycling indicates that some factors that impact compensatory effectiveness vary from time to time during sleep. The more robust compensatory effectiveness is, the more it will withstand these other fluctuating influences and the greater the fraction of time spent in stable breathing. Thus, provided that it can be shown that fluctuations between cycling and stable breathing are not related to spontaneous changes in severity of passive mechanical abnormalities (and this is one of the objectives/findings of this study), T_stable/ST is a specific indicator of compensatory effectiveness.

In contrast, the apnea–hypopnea index (AHI) is the product of frequency of obstructive events (cycling frequency) when breathing is cycling and fraction of time spent cycling:

(1)

Cycling frequency, one of the two key determinants of AHI, is primarily determined by factors that have little to do with upper airway dilator recruitment. Thus, factors that cause blood gas tensions to deteriorate faster (e.g., low FRC, low cardiac output, high metabolic rate, low alveolar PO₂ [13, 14], a low arousal threshold, or high controller gain [i.e., change in chemical drive per unit change in PO₂ or PCO₂]) will speed up termination of the apnea–hypopnea, resulting in a higher cycling frequency, regardless of whether upper airway dilators responded during the obstruction. Because cycling frequency varies over a wide range (40 to more than 150 per hour), differences in AHI among patients need not reflect differences in compensatory effectiveness. Notwithstanding its lack of specificity, AHI data are reported, in addition to T_stable/ST, to provide a familiar reference and for contrast with the new index.

What Is the Load Imposed by the Passive Mechanical Properties?
The pressure required of a muscle to do a task is determined by the magnitude of the task and the passive properties of the system on which the muscle acts. The task of upper airway dilators is to make it possible for flow to traverse the upper airway at a rate that allows a tolerable Pa_CO2 (required flow, _required). If this is not achieved, blood gas tensions deteriorate, leading to arousals and recurrent cycling. The dilators are opposed by the passive properties of the upper airway. These determine how much pressure the dilators must generate to permit _required, that is, the load imposed by that patient's airway for his or her ongoing ventilatory requirements.

Figure 1 shows schematically the relationship between distending pressure across upper airway wall and the maximum flow that can traverse the airway (max). Distending pressure is defined here as the sum of intraluminal pressure and active pressure generated by upper airway dilators. The upper airway is closed when distending pressure is below a finite value (P_close) (4, 15). Above P_close, the upper airway behaves as a Starling resistor that limits flow to a max that is related to distending pressure (16). The diagonal solid line is the hypothetical relationship between distending pressure, above P_close, and max in the passive pharynx. There are no direct measurements of this relationship, and these would be difficult to obtain (Appendix 1). A linear relationship between distending pressure and max is assumed (Appendix 2).

View larger version (12K): [in this window] [in a new window]   Figure 1. Schematic illustration of the theory behind estimation of total mechanical load on upper airway muscles. The diagonal line represents the relationship between pressure distending the pharynx and maximum flow (max) in the passive upper airway. The upper airway is closed below a certain pressure (Pclose). peak is the peak flow obtained on effective continuous positive airway pressure (CPAP). required is the minimum flow consistent with stable breathing. The point on the passive line corresponding to required (i.e., Prequired) is the total pressure that upper airway dilators must generate to permit a flow that is consistent with stable breathing, even if that flow is limited. passive is the maximum flow at atmospheric upper airway pressure with a passive pharynx.

The intersection of the dashed horizontal line and the diagonal line is the total distending pressure required (P_required) to achieve _required (Figure 1). P_required represents the pressure that upper airway dilators must generate to permit stable breathing in the absence of CPAP.

How to Estimate P_required?
Effective CPAP is, or can be defined as, the pressure that just eliminates flow limitation. On effective CPAP, E reflects ventilatory demand, and mean inspiratory flow (VT/TI) is a close approximation of _required (see section A in the online supplement). What remains to be determined, therefore, is the position of the passive P_distending–max relationship, in the vicinity of _required, under conditions comparable to spontaneous breathing. In theory, this can be done by briefly dialing down CPAP to a level associated with a max below _required. This is so because (1) upper airway dilator activity is considerably reduced on effective CPAP (8, 17, 18) and (2) there is no significant change in upper airway dilator activity during the first two breaths after a pressure dial down (17–19). Thus, if CPAP is titrated so that it just eliminates flow limitation, the highest flow observed on CPAP (peak flow, _peak) can be considered as max at the effective CPAP level (Figure 1). Flow observed during the dial down would provide another pressure–flow point below _required at the same upper airway dilator activity. A line joining the two points can then be used to obtain P_required (Figure 1). Although simple, this approach has two limitations: (1) Although upper airway dilator activity is much reduced on effective CPAP, it is not totally absent. Thus, P_required, obtained by this approach (Figure 1), underestimates P_required by the amount of pressure generated by these muscles on effective CPAP. The pressure equivalent of this residual activity has not been directly measured. However, it is almost certainly minimal. Thus, Schwartz and colleagues (19) measured P_close, breath by breath for three breaths, after rapid CPAP dial downs. P_close increased progressively from breath 1 to breath 3, reflecting the progressive decrease in lung volume, and viscoelastic behavior of upper airway (discussed later here). By the third breath, P_close was 1.4 ± 2.9 cm H₂O. P_close measured by Isono and colleagues in paralyzed patients was 2.8 ± 2.8 cm H₂O (4). The difference (1.4 cm H₂O) reflects not only the mechanical impact of residual muscle activity while on CPAP, but also includes a 33% increase in this activity between breath 1 and breath 3 (19). In addition, the effects of lung volume and viscoelastic behavior had not completely dissipated by breath 3. Thus, the mechanical effect of residual activity on CPAP is likely a fraction of 1.4 cm H₂O. Data are also presented from this study to support this assumption (DISCUSSION). (2) The effect of CPAP on max is not mediated exclusively via local distending action. CPAP increases lung volume, and this exerts an independent upper airway dilating force (20–22). This volume-related force can be expressed in pressure units, P_volume, representing the increase in upper airway distending pressure that would duplicate volume's effect on max. Because activation of upper airway dilators is not associated with an increase in lung volume, these muscles must generate extra pressure (corresponding to P_volume) on top of the nominal CPAP value if the same max is to be allowed. Furthermore, upper airway displays viscoelastic behavior (19, 23, 24). With sustained CPAP, upper airway area would be larger than if distending pressure were not sustained. This is equivalent to an additional dilating force (P_viscoelastic), exerted by CPAP, which would not be available without CPAP, and must also be provided by upper airway muscles to obtain the same max. The operation of these additional CPAP effects (P_volume and P_viscoelastic) is illustrated by the observation that when CPAP is rapidly reduced, flow does not decrease to its minimum value during the first breath (19, 25). Rather, the minimum value (min) is reached in three to five breaths (19, 25).

The experimental approach to be used in this study consists of (1) titrating CPAP to the level that just eliminates flow limitation; (2) measuring _peak and VT/TI on effective CPAP, with the latter being used as _required; (3) rapid dial downs of pressure to 1.0 cm H₂O, the closest to atmospheric pressure that can be effected; and (4) measuring maximum flow breath by breath after the dial down to determine the lowest value (min) (because min represents the maximum flow at a common and very low pressure (1 cm H₂O) before any compensatory response [Appendix 3], it is used as a measure of passive collapsibility); and (5) Measuring the change in FRC (FRC) after the dial down. Figure 2 illustrates the theoretical basis for analysis. The dotted diagonal lines represent the relationship between distending pressure and max at constant volume and P_viscoelatic. The leftmost line (CPAP line) is the relationship immediately before dial down, where P_volume and P_viscoelastic are at their highest level. When CPAP is removed, these lines shift progressively to the right (B1 to B4), reflecting the progressive reduction in P_volume and in P_viscoelastic. The magnitude of the shift progressively decreases breath by breath, reflecting the exponential nature of decline in volume and in P_viscoelastic. min is reached at breath 4 in this illustration. The diagonal dotted line at min (spontaneous breathing line) reflects the passive P_distending–max relationship in the absence of CPAP (i.e., after disappearance of P_volume and P_viscoelastic) (Appendix 3). The desired value of P_required is the pressure on the spontaneous breathing line at _required. This value cannot be directly obtained. A line is drawn between the effective CPAP point and the min point (Schwartz and colleagues [19] found a linear relationship between CPAP level and max during such brief dial downs). This line describes the relationship between CPAP level and max, including the associated differences in P_volume and P_viscoelastic. The pressure value on this line at _required (CPAPmin) gives the CPAP level required to obtain _required in the absence of compensation. It underestimates actual P_required by an amount corresponding to the P_volume and P_viscoelastic associated with CPAPmin (distance between CPAPmin and P_required; Figure 2). Thus:

(2)

View larger version (24K): [in this window] [in a new window]   Figure 2. Schematic illustration of sequence of events during a rapid dial down of CPAP to 1.0 cm H2O. Diagonal dotted lines are isolung volume passive Pdistending–max lines. The leftmost of these lines describes what would happen to max if CPAP is reduced but lung volume remained unchanged. When CPAP is reduced, lung volume decreases, removing the distending force associated with lung volume. This causes the passive lines to shift to the right during the dial down. The rightward shift causes max to progressively decrease until a minimum value is reached within a few breaths (min). peak, the highest flow observed on effective CPAP, is considered max at effective CPAP. A line joining the effective CPAP and min points (solid diagonal line) describes the CPAP pressure required to result in different max values in the passive pharynx and includes the effect of changes in lung volume associated with the different levels of CPAP. required is the flow required to maintain stable breathing. CPAPmin is the CPAP required to produce required. The diagonal dashed line is the relationship between CPAP pressure and max of the first breath after dial down. Note that both the solid and dashed lines underestimate the isovolume passive Pdistending–max slope (diagonal dotted lines). B1–B4 = breath number after dial down. Pclose, B0 is closing pressure if CPAP is reduced to the point of closure, but lung volume was held constant.

(3)

In this study, CPAPmin and FRC are used as coindependent variables that collectively reflect the total load imposed by passive properties on upper airway dilators. Regression analysis between these two independent variables and OSA severity, as reflected by T_stable/ST, is done. K₁ and K₂ are considered constant across patients. The implications of this assumption are discussed later (DISCUSSION).

Theoretical Basis for Interpretation
In this model, compensatory effectiveness simply refers to how much pressure upper airway dilators can develop in a sustained fashion in the absence of arousals. Stable breathing develops when this pressure exceeds the pressure required to maintain stable breathing, as given by Equation 3. Because a large amount of stable breathing time (i.e., mild sleep apnea) may reflect either a low P_required or a high degree of compensatory effectiveness, assessment of the latter can only be done by comparing T_stable/ST at similar P_required; a higher T_stable/ST at the same P_required indicates more compensatory effectiveness and vice versa.

Whereas correlations between T_stable/ST and P_required identify the separate contributions of mechanical load and compensatory effectiveness to OSA severity, they do not indicate the separate contributions of passive collapsibility and flow demand to severity. This is because P_required reflects the combined impact of both variables. To ascertain the separate contributions of collapsibility and flow demand, the relationship between T_stable/ST (or AHI) and collapsibility, as reflected in min (Appendices 3 and 4), is plotted. Higher OSA severity at the same min indicates either a higher flow demand or a lower compensatory effectiveness. Should the difference in severity disappear when flow demand is incorporated in the regression or if P_required is used as the independent variable (which accomplishes the same end), differences in severity can be attributed to differences in flow demand between the patients or conditions being compared. Partial reduction of the difference in severity (after inclusion of flow demand) would signify that the difference is in part related to flow demand and in part to differences in compensatory effectiveness.

	METHODS

TOP ABSTRACT THEORETICAL CONSIDERATIONS METHODS RESULTS DISCUSSION APPENDIX 1 APPENDIX 2 APPENDIX 3 APPENDIX 4 REFERENCES

 
Patients
Eighty-two patients referred to the sleep laboratory for possible OSA were studied.

Apparatus
CPAP was applied via nasal mask connected to a special ventilator that allowed reduction of CPAP to 1.0 cm H₂O. A pneumotachograph was inserted upstream from the mask. Flow, airway pressure, and standard polysomnography signals were recorded.

Protocol
CPAP was started after obtaining enough information for clinical evaluation. Pressure was increased in 1 to 2 cm H₂O steps until the first level associated with no flow limitation. During periods of stable sleep, pressure was dialed down to 1 cm H₂O. Dial down was maintained until arousal or 60 seconds. The procedure was repeated when stable sleep resumed.

Analysis
Polysomnography preceding CPAP.
Sleep, arousals, and respiratory events were scored using standard criteria (27–29). Time spent with stable breathing was defined as sleep periods without hypopneas, apneas, or respiratory effort-related arousals, in excess of 3 minutes. AHI, sleep time, and stable time (expressed as fraction of sleep time) were calculated separately for each position and sleep state combination (combo).

Dial downs.
Flow signal was offset by the amount of leak (see section B in the online supplement). Calculated baseline values (from the 60 seconds preceding dial down) included average tidal volume, respiratory rate, inspiratory (TI) and expiratory times, mean inspiratory flow (VT/TI), inspiratory time fraction (TI/Ttot), and electroencephalogram (EEG) delta power (see section B in the online supplement). In addition, the highest peak flow observed during baseline was noted (_peak). Flow and tidal volume were measured breath by breath from the point when airway pressure reached minimum value until the first sign of compensation (increase in flow, with or without arousal). In the majority of dial downs, there was evidence of flow limitation (flat top section in inspiratory flow and failure of flow to increase despite increasing effort, as evident from Respitrace; Figure 3) . The average value during the flow "plateau" in each breath was considered max. The lowest max observed during dial down was identified (min).

View larger version (53K): [in this window] [in a new window]   Figure 3. Tracings from a dial down associated with moderate hypopnea. Top 2 channels are electroencephalogram tracings. The flow signal was corrected for different leaks at the two pressure levels. Note the typical configuration of flow limitation after the dial down and that flow progressively decreases over several breaths. A = arousal; PAW = airway pressure.

CPAP–max relationship and P_close.
Two CPAP–max relationships were calculated in dial downs associated with min of more than 0 (i.e., hypopneas). One, corresponding to the diagonal dashed line in Figure 2, was calculated from [(CPAP_baseline - CPAP_{dial down})/(_peak - _BR1)], where _BR1 is max of breath 1. The other, corresponding to the diagonal solid line in Figure 2, was estimated from [(CPAP_baseline - CPAP_{dial down})/(_peak - min)]. This, together with VT/TI, was used to calculate CPAPmin (Figure 2). Extrapolation of the two lines to zero flow provided boundaries for P_close. The average of the two intercepts was used as P_close. The method of estimating CPAP–max relationship and P_close in dial downs associated with complete obstruction is described section B of the online supplement. When there was no evidence of flow limitation during the dial down, CPAPmin was assigned a value of 1 cm H₂O. Data from all dial downs in a given combo were averaged.

Statistical Analysis
Values are reported as mean ± SD. Multiple regression analysis was used to identify determinants of mechanical load, the relationship between mechanical load and AHI or T_stable/ST, and the effect of different variables on this relationship (see the online supplement).

	RESULTS

TOP ABSTRACT THEORETICAL CONSIDERATIONS METHODS RESULTS DISCUSSION APPENDIX 1 APPENDIX 2 APPENDIX 3 APPENDIX 4 REFERENCES

 
Polysomnography Results
Overall AHI was 46.0 ± 35.0 hour^-1 (Table 1) . In 65 patients, AHI was more than 15 hour^-1. Of the remaining 17 patients, 10 had an AHI of more than 15 hour^-1 in at least one position/sleep state. The remaining seven patients were principally heavy snorers with sporadic hypopneas (AHI = 6.1 ± 3.7 hour^-1).

Periods of stable breathing were observed in 123 of 202 combos. Sixty-four patients (78%) had periods of stable breathing in one or more combos. T_stable/ST was higher on the side than on the back (0.61 ± 0.37 vs. 0.29 ± 0.35, paired n = 68, p < E-9) and in NREM than in REM (0.62 ± 0.37 vs. 0.39 ± 0.44, paired n = 57, p < 0.0001). The difference in T_stable/ST remained significant after excluding "REM-only" OSA patients (0.55 ± 0.38 vs. 0.42 ± 0.46, p < 0.01).

Dial-down Results
There were 749 dial downs. Of these, 148 dial downs were discarded (section C1 in the online supplement). Table 2 shows baseline data. Males had higher E, VT, VT/TI, and lower respiratory rates than females. Effective CPAP was higher on the back, and VT/TI was slightly lower. Otherwise, there were no position-related differences. Breathing in REM sleep was rapid and shallow relative to NREM sleep, and TI/Ttot was larger, and VT/TI was lower. E was not different. Stepwise regression (Table 3) showed that females had lower E and VT/TI after adjusting for differences in other variables. BMI had a highly significant effect on E and VT/TI. Flow rate was lower in REM. The position had no independent effect.

View larger version (52K): [in this window] [in a new window]   Figure 4. Tracings from a dial down associated with complete obstruction. Top 2 channels are EEG tracings. Flow signal was corrected for different leaks at the two pressure levels.

View larger version (51K): [in this window] [in a new window]   Figure 5. Tracings comparing the response to dial downs in REM and non-REM (NREM) sleep in the same position in the same patient. C3/A2 = central electroencephalogram; EOG = electro-occulogram; Note the erratic increases in flow during the dial down in REM (arrows) but not in NREM sleep.

Reproducibility of Dial-down Results
Figure 6 is a plot of the difference between min of individual dial downs and average min of all dial downs in the same combo. In most cases, the results were consistent within the combo. In some cases, however, fairly large differences occurred. The 95% confidence interval was ± 24% baseline. The diagonal solid lines represent the cases where such variability would place flow within 20% (lower line) or 10% of baseline VT/TI. When average min was less than 50%, even the most extreme points did not reach the range that would permit stable breathing (see section A in the online supplement). When average min was 50–80% baseline, some points fell above the diagonal lines, but these were a minority.

View larger version (27K): [in this window] [in a new window]   Figure 6. Reproducibility of dial-down results. Each point is the difference between minimum flow (min) observed in one dial down and the average min in the same combo (patient/position/sleep state combination), plotted against the average min in the combo. Horizontal dashed lines are the 95% confidence interval. The two diagonal lines represent the boundaries for min reaching 80% (lower line) or 90% of baseline flow.

Table 4 shows results of stepwise regression for variables that correlate with collapsibility (min%) and mechanical load (CPAPmin). min decreased, and CPAPmin increased with age, BMI, and body position. Males had a borderline lower min% but a significantly higher CPAPmin, reflecting the fact that CPAPmin incorporates the effect of the higher flow demand in men. Baseline delta power, baseline ventilation, and height had no effect. All of the variables included in the model accounted for only 34% of the variance in min and 41% of the variance in CPAPmin, suggesting that other factors play a more important role in determining collapsibility and mechanical load. Although REM sleep had no effect in stepwise regression, paired comparisons in patients who had both REM and NREM sleep data showed a slightly higher min in REM sleep (p < 0.05). Figure E5 (in the online supplement) shows individual differences in min according to sleep state and body position.

View larger version (21K): [in this window] [in a new window]   Figure 7. Scatter plot of the relationship between minimum flow observed in individual combos (min; average of all dial downs in a combo) and the fraction of sleep time in stable breathing in the same combo during polysomnography. The dashed lines represent the expected fraction of sleep time during which breathing is stable (Tstable/ST), in the absence of any response by upper airway dilators, if maximum flow required for stable breathing is 90% (right line) or 80% of baseline VT/TI. Note that a majority of the points to the left of the 80% vertical line show periods of stable time, indicating that these patients were able to mount effective compensation without arousal for at least part of the time.

(4)

(5)

Tables 6 and 7 show that when age, sex, BMI, sleep state, and body position are included with indices of load, r² still does not exceed 0.40. An attempt was made to determine the extent to which residual variability is inherent to individual patients. There were 34 patients in whom there were dial down and polysomnography data in two positions in the same sleep state. The deviation from the main regression (Table 6, regression employing min and CPAPmin) was calculated for the two combos in each patient. Figure 8 shows the relationship between the two residual values in these 34 pairs. There was a highly significant correlation (r² = 0.49, p < E-5), indicating that overperformers in one combo tend to be overperformers in the other.

View larger version (22K): [in this window] [in a new window]   Figure 8. Thirty-four patients had dial down and polysomnography data on both the side and the back in NREM sleep. The plot compares the results in the two combos, in the same patient, expressed as deviation of Tstable/ST from what is expected. Positive values indicate that the patient had more stable time than expected. The diagonal line is line of identity.

	DISCUSSION

TOP ABSTRACT THEORETICAL CONSIDERATIONS METHODS RESULTS DISCUSSION APPENDIX 1 APPENDIX 2 APPENDIX 3 APPENDIX 4 REFERENCES

Significance of Stable Breathing Time
It has been long recognized that most patients with OSA develop periods of stable breathing. This important observation has received little exploration. In theory, periods of stable breathing can develop if there are time- or state-dependent changes in passive mechanical properties, in baseline upper airway dilator tone (which mitigate or preclude sleep-related hypopneas), or in effectiveness of response to obstructive events. This issue was addressed by determining the reproducibility of min among multiple dial downs in the same position and sleep state (Figure 6). Because all dial downs began from a basal chemical drive, differences in min among observations in the same combo reflect spontaneous changes in passive properties (e.g., because of changes in head and neck position) or in basal dilator tone. The differences observed were not sufficient to explain independently the periods of stable breathing except in a minority of cases. Thus, when average min was less than 50% (most combos), flow in none of the dial downs was close to a level that could be sustained without arousal (Figure 6; see section A in the online supplement). However, patients very frequently displayed periods of stable breathing when average min was less than 50% and even when complete obstruction developed consistently during the dial downs (Figure 7). Although flow exceeded 80% in some dial downs when average min was 50–80%, this was uncommon and is to be contrasted with the extensive stable time observed under similar conditions during polysomnography (Figure 7).

Based on these considerations, it is possible to conclude that development of stable breathing in a patient with significant passive abnormalities (e.g., min of less than 80% baseline) signifies the patient's ability to respond effectively to the hypopneas or apneas in the absence of arousals. In this study, 78% of patients demonstrated periods of stable breathing in at least one combo. Thus, the majority of OSA patients can mount effective responses without arousals. This is an important conclusion given the prevailing view that arousal is the main mechanism by which flow is restored in these patients. Twenty-two percent of patients showed no stable breathing at any time. It is possible that a minority of patients are dependent on arousals to open the airway. Alternatively, strong responses may be present in these patients but are overwhelmed by the passive abnormality or the response may be too brisk, resulting in a ventilatory overshoot, which perpetuates cycling (12).

If a patient can respond effectively at some point, why does he or she revert to recurrent cycling at other times? Reference was made to the occurrence of time-dependent changes in min. It is also likely that compensatory effectiveness varies from time to time. Compensatory effectiveness may be adequate to restore flow to a tolerable level given a certain mechanical load. A small change in mechanical load or in compensatory effectiveness may result in flow falling short of the tolerable level, precipitating cycling.

The association between delta sleep and stable breathing in OSA patients is well known. The mechanism is not known. EEG baseline delta power varied considerably among dial downs in the same combo. This offered the opportunity to explore the mechanism of this association. The effect of delta activity during baseline on min was negligible, amounting to a difference of approximately 1% of baseline flow in the face of large differences in baseline delta activity (stage 2 to stage 4). Thus, the increased upper airway dilator tone during delta sleep (30) is not intrinsic to this state. Rather, it likely reflects a permissive role whereby, by delaying arousal, delta sleep allows upper airway response to evolve in an orderly fashion to a steady state.

Indices of the Challenge to Upper Airway Dilators
This study demonstrated that baseline flow demand (VT/TI) affects OSA severity. Variability in VT/TI was mostly related to differences in baseline E, which ranged from 3.3 to 9.3 L/minute (Table 2). However, differences in TI/Ttot (range, 0.25–0.55) accounted for 18% of the variability. Thus, differences in TI/Ttot indirectly contribute to OSA severity.

The lack of predictive value for FRC, when combined with CPAPmin, was surprising because changes in lung volume are believed to exert an independent dilating action (20–22), and there was a wide range of FRC (0.03 to 1.19 l). This was not because of an excellent correlation between FRC and CPAPmin; r² of this relationship was only 0.25, likely reflecting the wide range of respiratory compliance, near FRC, in these patients with widely different BMI (Table 1). It is possible that changes in FRC with CPAP provide no extra dilating force in these patients; mechanical coupling between chest and pharynx may be different from that in normal subjects or animals. In support of this possibility, Series and colleagues (31) found no reduction in AHI in patients with OSA when lung volume was increased by negative body surface pressure (i.e., without an increase in pharyngeal pressure). The other possibility is that the caudal traction effect is not related to volume, per se, but to the associated change in respiratory system elastic recoil (i.e., K1 of Equation 3 is proportional to respiratory elastance). This would result in P_volume being proportional to CPAPmin (as FRC = CPAPmin/respiratory elastance), whereby the effect of volume becomes incorporated in the CPAPmin coefficient. Regardless of which possibility is true, the demonstration that FRC has no independent predictive value simplifies the selection of a comprehensive index of the mechanical load because the caudal traction effect, if any, is a constant fraction of CPAPmin. Thus, CPAPmin can be such an index.

CPAPmin should be distinguished from the lowest CPAP required in practice to establish stable breathing. CPAPmin reflects CPAP required for stable breathing, albeit with flow limitation, in the absence of compensatory responses. Because flow limitation may evoke compensatory responses, CPAP required to stabilize breathing in practice may be lower than CPAPmin.

min continued to have an independent predictive value after it was combined with CPAPmin (Table 5). Figure 9 illustrates possible explanations. CPAPmin reflects the pressure needed to increase max to baseline VT/TI. In reality, a lower flow can be tolerated (Figure 9, dashed horizontal line). For the same CPAPmin, if upper airway dilator response is less than CPAPmin (Figure 9, vertical dashed line), flow is more likely to reach the tolerable range in a patient with higher min. A higher min at the same CPAPmin also signifies that the change in flow per unit change in distending pressure is less. This may be stabilizing because a ventilatory overshoot, with consequent hypocapnia, would be less likely, reducing the chance of self-perpetuating cycling.

View larger version (14K): [in this window] [in a new window]   Figure 9. Schematic illustration of possible reason why minimum flow during the dial down (min) offers predictive (of severity) value in addition to the value provided by CPAPmin. CPAPmin reflects the pressure required of upper airway dilators to achieve a normal mean inspiratory flow (baseline VT/TI). In reality, a flow that is somewhat lower can be tolerated (tolerable VT/TI). In the event upper dilator response falls short of CPAPmin, a patient with higher min has a better chance of reaching the tolerable flow.

Relative Contributions of Mechanical Load and Control Mechanisms to OSA Severity
Table 5 shows that approximately 34% of the variability in OSA severity is related to differences in load, thereby suggesting that most of the variability is due to differences in compensatory effectiveness. Residual variance was only in small part related to body position and sleep state (r² improved from 0.34 to 0.38, Tables 5–7, respectively). Approximately half the residual variability appeared inherent to individual patients (Figure 8). Thus, some patients are good responders, whereas others are not. This is in keeping with previous observations that the magnitude of upper airway dilators' response to negative pressure varies among subjects but is reproducible in the same subject (32). The remaining variability probably reflects differences, during polysomnography, in variables that may alter compensatory effectiveness (e.g., arousal threshold, chemical control gain, discussed later here).

Several studies reported the relation between an index of collapsibility and an index of OSA severity (4, 33, 34). In all but one (4), the collapsibility index used (closing pressure, P_crit) was estimated under conditions in which compensatory responses were well under way (33, 34). Such correlations do not reflect the relative contributions of passive abnormalities and control mechanisms because the index used incorporates both elements. Thus, a low P_crit could result from mild passive abnormalities or from severe passive abnormalities with excellent compensation. A low AHI in such a case does not help sort out the underlying mechanism.

The only previous study in which an index of passive collapsibility was related to OSA severity was by Isono and colleagues (4). These authors found a correlation between P_close of the velopharynx, measured under anesthesia, and a desaturation index. Although the correlation was significant (p < 0.005), r² was only 0.14, indicating that this index accounted for only 14% of the variability in severity. In this study, the correlation was stronger (r² = 0.34; Table 5). The difference is likely related to our use of a more comprehensive index that accounted not only for P_close but also for differences in the P_distending–max relationship and in flow demand.

Validity of Assumptions
The conclusion that mechanical load accounts for 34% of OSA variability is subject to the validity of assumptions made in estimating the mechanical load. The following are the main assumptions and their limitations:

1. Residual upper airway dilator activity on CPAP is, mechanically, negligible. Reference was made earlier to the data of Schwartz and colleagues (19) in support of this assumption (see THEORETICAL CONSIDERATIONS). In addition, in this study, P_close was estimated in combos associated with AHI of more than 20 hour^-1. On average, it was 1.3 ± 3.6 cm H₂O, very similar to the values obtained by Schwartz and colleagues (1.4 ± 2.9 cm H₂O) (19). In patients with the same severity (AHI of more than 20 hour^-1), studied during paralysis, P_close was 2.8 ± 2.8 cm H₂O (4). The difference was small and not significant.
   
2. VT/TI on CPAP reflects the flow (max) that can be tolerated in the steady state (i.e., _required). Arguments are presented in section A of the online supplement in support of this assumption. However, the relationship between unloaded VT/TI and _required is complex and may be influenced by (1) how much reduction in ventilation can be tolerated without arousal, (2) how much TI/Ttot increases in the presence of flow limitation, and (3) the shape of the inspiratory flow waveform (see section A in the online supplement). To the extent that these may differ among patients, variability in VT/TI may not entirely reflect variability of _required. The implication of this uncertainty is that some of the residual variance in correlations between CPAPmin and OSA severity may be related to differences in mechanical load that are not reflected by CPAPmin.
   
3. The factor that governs conversion of lung volume to distending upper airway force (K1 in Equation 3) is the same in different patients. This assumption has been rendered irrelevant by the finding that FRC has no independent predictive value.
   
4. K2 (Equation 3) is constant. To the extent that differences may exist among patients in the extent of upper airway viscoelastic behavior, some of the residual variance in correlations between CPAPmin and OSA severity may be related to differences in K₂, which is a characteristic of the passive system. There is no information on this topic. However, because such variability operates on a relatively small component of the total load, its overall impact on the correlation coefficient is likely small.
   

It must be pointed out that the main implication of uncertainties associated with assumptions 2 and 4 is that the relationship between mechanical load and OSA severity (r²) may be somewhat higher than the 34% value found using CPAPmin. These uncertainties do not undermine any of the other conclusions as similar findings were obtained using the combination of min and VT/TI (Tables 6 and 7). min and VT/TI are straight measurements that reflect collapsibility and ventilatory demand and are not subject to these assumptions.

Relationship between Pharyngeal Distending Pressure and Maximum Flow
The P_distending–max relationship is important for two reasons. First, it determines how much pressure pharyngeal dilators need to generate, above P_close, to permit a given flow. This is an important component of the load. Second, because in the presence of pharyngeal flow limitation total E is determined by pharyngeal max, the P_distending–max relationship determines the overall ventilatory response to changes in upper airway dilator activity. This is an important determinant of ventilatory stability (12, 14). Thus, if the slope of the P_distending–max relationship (in cm H₂O · l^-1 · second) is high, the load on upper airway dilators is high, but ventilatory control is more stable.

The P_distending–max relationship estimated from the first breath after a rapid dial down was 27.2 ± 11.5 cm H₂O · l^-1 · second. Because this value underestimates the true slope (Figure 2 and related discussion), the actual slope must be even higher. Thus, the load imposed on the upper airway dilators is high; the dilators need to generate, on average, 14 cm H₂O to increase max by 0.5 L/second. The range of the slope was quite wide (9.7–62.1 cm H₂O · l^-1 · second). Thus, the slope likely plays an important role in determining whether effective compensation develops.

Using a roughly similar approach, Boudewyns and colleagues (25) reported lower slope values (15.9 and 21.7 cm H₂O · l^-1 · second in the supine and lateral positions, respectively). This difference is likely due to the fact that they constructed the relationship from data obtained in the third to fifth breath after the dial down, as opposed to our use of the first breath. This would result in greater underestimation of the true slope (compare the solid and dashed diagonal lines in Figure 2).

Age and OSA
The prevalence of OSA increases with age (35, 36). In awake normal subjects, pharyngeal cross-sectional area decreases (37, 38) and upper airway resistance increases (39) with age. Whether this translates to increased collapsibility during sleep is not known. Upper airway resistance during sleep increases with age (40). It is not clear whether this is due to changes in passive mechanics or to less effective compensatory responses. In this study, after controlling for sex, height, BMI, body position, and sleep state, min was significantly lower in older patients (Table 4). At the same mechanical load, however, older patients did not have more severe OSA (Tables 6 and 7). Thus, the compensatory effectiveness is not affected by age. These findings suggest that the increased prevalence with age may be related to deterioration in upper airway mechanics. However, because the patients studied were not randomly selected from the general population, these conclusions need not reflect a general age effect.

Sex and OSA
The prevalence of OSA is higher in men (36). Studies exploring possible mechanisms in awake normal subjects produced inconsistent results (reviewed in 41). In sleeping normal subjects, Pillar and colleagues (42) found that men more readily increased upper airway resistance and developed flow limitation when an inspiratory resistance was added. In contrast, Rowley and colleagues (40) found no difference between sexes in upper airway closing pressure or upper airway resistance during sleep. These authors, however, reported later that upper airway compliance is higher in sleeping men, suggesting that it may be more collapsible (43). There are no studies regarding differences between sexes in passive mechanics or upper airway control, during sleep in OSA patients. The current population was typical of sleep laboratory populations. It was predominantly male (4:1 ratio). Women were older (51.4 ± 8.8 vs. 48.5 ± 12.0 years) and heavier (BMI of 37.3 ± 4.8 vs. 33.0 ± 5.9). After allowing for differences in age and BMI, men had more severe OSA; AHI was greater by 28.0 ± 9.1 hour^-1 (p = 0.003). Thus, the patients of this study offered an opportunity to examine mechanisms of sex difference in severity, at least in the sleep laboratory population. min was lower in males, but the difference was small and just shy of significance (Table 4). At the same min, however, males had more severe OSA (Tables 6 and 7, first column). The difference was large (AHI greater by 17 hour^-1). When differences in flow demand were incorporated in the regression, by using CPAPmin or by adding baseline flow as an independent variable, sex difference disappeared (Tables 6 and 7). Thus, the greater severity in males was mostly because of their higher flow demand (Table 2). There was no sex difference in compensatory effectiveness.

BMI and OSA
There is a strong association between BMI and severity and prevalence of OSA (reviewed in 36). Studies in awake subjects failed to demonstrate a BMI effect on pharyngeal size (44, 45). However, several studies reported that obesity increases upper airway collapsibility in awake (46–48), sleeping (49) and anesthetized, and paralyzed subjects (50). It is not known, however, whether increased collapsibility is the only relevant mechanism. These patients covered a wide range of BMI (Table 1), making it possible to address this issue. As expected, BMI had a highly significant effect on min (Table 4). The effect of obesity was not entirely because of the increased collapsibility. At the same min, obese patients had less stable time (Table 6, first column) and more AHI (Table 7, first column). As with the sex effect, the effect on T_stable/ST disappeared when the higher flow demand (Table 3) was taken into account (subsequent columns in Table 6), indicating that compensatory effectiveness is not impaired. With respect to AHI, allowance for the higher flow decreased the BMI coefficient, but a significant BMI effect remained (Table 7). This indicates a higher cycling frequency when breathing is unstable. This is likely due to the lower FRC, higher metabolic rate, and lower PO₂, characteristic of obesity. These factors would speed up the increase in chemical drive when a hypopnea or apnea develops, causing earlier termination and a shorter cycle duration. In summary, the effect of obesity is related to three factors: increased passive collapsibility, increased flow demand, and increased cycling frequency. There is no evidence of impairment of compensatory mechanisms.

Body Position and OSA
OSA is more severe in the supine than in the lateral position (51–53). Investigations in awake subjects produced inconsistent results (45, 54, 55). There is, however, compelling evidence that collapsibility is increased in the supine position during sleep (25, 56, 57) and under anesthesia (58). Whether increased collapsibility accounts for the entire positional effect on OSA severity is not known. In the current patients, min was reduced by 27% baseline in the supine position, consistent with greater collapsibility (Table 4 and Figure E5). However, at the same min, OSA was considerably more severe in the supine position (Tables 6 and 7), indicating that compensatory effectiveness is considerably impaired in this position.

The mechanism by which the supine position impairs compensatory mechanisms is not clear. One possibility is that the mechanical advantage of upper airway dilators is reduced. This could happen, for example, if the site of closure/flow limitation moves to a locale where upper airway dilators are less effective or if upper airway muscle length is reduced in the supine position. This would reduce the muscle's pressure-generating capacity (59). Alternatively, the reduced effectiveness may be related to changes in chemical control in the supine position through, for example, a lower FRC or lower PaO₂. Such changes would increase chemical control loop gain, resulting in greater ventilatory instability (12).

REM Sleep and OSA
Because upper airway muscle activation and response to negative pressure are depressed more in REM sleep than in NREM sleep (60, 61), OSA should be more severe during REM. However, reported differences in AHI have been inconsistent, with some reporting no difference (25, 62) and others observing higher AHI in REM (63). There is a subgroup of patients who have OSA predominantly, or exclusively, in REM (63, 64). In this study, patients who slept in both states had significantly higher AHI in REM. However,