INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The obstructive sleep apnea (OSA) syndrome is a disorder in which there is repetitive collapse and closing of the pharynx during sleep. Although the upper airway (UA) occlusion may be due to several factors such as anatomic abnormalities and obesity, there is growing evidence that even when these factors are taken into consideration, individual collapsibility is a key factor for the UA occlusion (1). The importance of an abnormal pharyngeal susceptibility to collapse in the pathogenesis of obstructive apneas was demonstrated by studying the pharyngeal critical pressure (Pcrit) in patients with OSA and in control subjects (4). Applying a theoretical model of upper airway occlusion in which the Starling resistor system is implied, the investigators suggested that the upper airway can be represented as a simple tube with a collapsible segment. Flow cannot occur until the pressure upstream of the collapsible segment exceeds the surrounding pressure, i.e. Pcrit. These investigators demonstrated that in normal subjects, Pcrit is negative (5), that is, the airways tend to stay open. In snorers the critical pressure is less negative, which implies that their airway is more susceptible to collapse. In patients with OSA, the Pcrit is positive, inducing collapse and occlusion of the UA during sleep (6). On the basis of these results, the investigators suggested that Pcrit is an index of abnormal upper airway collapsibility allowing differentiation between control subjects, snorers, and patients with OSA.

Continuous positive airway pressure (nCPAP) is currently used as the treatment of choice for OSA, but the level of pressure to prevent collapse of the UA varies among patients. Recent studies have indicated that neck circumference, body mass index (BMI), and apnea plus hypopnea index (7) play an important role in modulating the effective nCPAP, whereas observations by our group (8) have shown that length of the uvula and inspiratory effort developed against occlusion are particularly important in modulating the level of efficacious pressure. Applying the hypothesis of the UA functioning as a Starling resistor, Gold and Schwartz (9) suggested that the measurement of Pcrit may be a useful tool in the diagnostic and therapeutic approach to patients with OSA. According to these investigators UA obstruction will be removed when a differential pressure of 8 cm between nasal pressure and Pcrit is applied during the night on nCPAP titration (9).

Despite the evident role of UA collapsibility in OSA, the use of pharyngeal critical pressure has not yet been extensively investigated in clinical practice. In the previous studies greater Pcrit was consistently associated with greater severity of OSA, and this factor may have influenced the results. Moreover, no data about the effect of anthropometric variables and sex in this measurement of UA collapsibility are available.

The present study was undertaken to examine the influence of age and anthropometric variables on UA collapsibility and to evaluate whether Pcrit by itself might provide diagnostic information enabling us to differentiate patients with different severity of the disease. An additional goal was to determine whether measurement of UA collapsibility during sleep might provide more than anthropometric variables, apnea density, and respiratory effort diagnostic information on the level of efficacious nCPAP necessary to restore upper airway patency.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We reviewed the polygraphic recordings of 200 patients referred to our center between January 1996 and December 1997 because of snoring or daytime sleepiness associated or not with reported apneas. The patients were eligible for the study if they had: (1) a nocturnal polysomnography using a pneumotachograph and esophageal pressure recordings without technical problems during the monitoring; (2) a titration night consecutive to diagnostic study showing no mask and mouth leaks; (3) the presence of an apnea plus hypopnea index (AHI) > 10 with at least 85% of apneas as obstructive. The exclusion criteria were previous treatment for sleep apnea with nCPAP or surgery, a history or clinical evidence of neuromuscular disease, and signs of acute lung or cardiopulmonary disease. Each subject signed a consent form outlining that some of the collected data would be used for research purposes.

Of the original sample, 106 patients (89 men and 17 women) fulfilling the above criteria were examined. All patients underwent a clinical history and physical examination, including measurement of neck size, and a full evaluation, including polysomnography, blood gas analysis, and respiratory function tests.

Polysomnographic Study

Nocturnal recording was carried out for two nights, the first one without nCPAP and the second one with nCPAP, as part of the diagnostic work-up. Polysomnography included an electroencephalogram, electrooculogram, and electromyogram of chin muscles for conventional sleep staging. Breathing was analyzed with a Fleisch no. 2 pneumotachograph and an electronic integrator (Godart Statham, Bilthoven, The Netherlands) attached to a face mask. Oxygen saturation was measured continuously with an ear oximeter (Biox III; Ohmeda, Boulder, CO). Esophageal pressure was measured with a 10-cm latex balloon placed in the lower third of the esophagus, inflated with 1 ml of air, and connected to a pressure transducer (Validyne MP 45; Validyne, Northridge, CA). The reference pressure for esophageal pressure measurements was the atmospheric pressure, and consequently all inspiratory pressures were negative. To facilitate analysis, all pressure values will be given in absolute values so that increase in respiratory effort corresponds to increase in esophageal pressure swings.

During the titration night the nasal pressure (Pn) was measured through a side port in the nasal mask. A thermocouple was placed over the lips to ascertain that mouth breathing did not occur during the measurements. nCPAP was increased from the initial value of 2 cm H₂O with 1 cm H₂O increments up to the efficacious level (nCPAP_eff) at which the following conditions were obtained: apnea and snoring were abolished, no inspiratory flow limitation occurred, and the amplitude of esophageal pressure swings did not exceed twice the awake value during quiet breathing.

Sleep was scored using the criteria of Rechtschaffen and Kales (10) for 10-s epochs and the following variables were calculated: total sleep time, wake time after sleep onset (WASO), sleep efficiency (SE: total sleep time/total recording time × 100), and the percentage of Stages 1, 2, and 3-4 and REM sleep.

Respiratory events were scored using standard criteria. Apneas were defined as the total cessation of airflow lasting for  10 s. Hypopneas were defined as a 50% or greater reduction in tidal volume from the baseline value lasting at least 10 s. The apnea index (AI) and the AHI were established as the ratio of the number of apneas and apneas plus hypopneas per hour of sleep. Apnea time was expressed as the sleep time spent in apnea. As indices of nocturnal hypoxemia, we consider the mean Sa_O2 (Sa_O2 mean) during sleep and the minimal value recorded during sleep (Sa_O2 min).

During the nCPAP titration night, we measured pharyngeal critical pressure according to described methods (4) by relating changes in maximal inspiratory airflow (VI_max) to varying levels of nCPAP by least squares regression analysis; Pcrit representing the extrapolated pressure at zero flow (Figure 1). At the level of nCPAP at which apneas and hypopneas were recorded, we averaged the maximal flow for the last four breaths for at least five hypopneas at each level of nCPAP (Figure 2a). During periods free of apneas and hypopneas we measured maximal inspiratory flow for several randomly chosen breaths (at least 30 breaths for each nCPAP increase), with flow limitation defined as a reduced and plateaued inspiratory airflow while esophageal pressure still increased (Figure 2b). Analysis was done during Stage 2 of non-rapid-eye-movement (NREM) sleep in the supine position. In all patients the r of the regression analysis was greater than 0.89. Upper airway resistance (Rus) was determined as the reciprocal of the slope (Pn/VI_max cm H₂O/L/s) in the regression equation. Because error estimation of the measurement may be affected by mask or mouth leaks, special care was taken to prevent air leaks around the mask, and the subjects were observed during sleep to prevent mask shifts. The use of a thermistor indicating mouth leaks allowed us to exclude patients with increased mouth breathing.

View larger version (11K): [in this window] [in a new window]   Figure 1.   Representative graph of maximal inspiratory flow (VImax) versus nasal pressure (Pn) for a patient with obstructive sleep apnea syndrome. Pcrit is represented by the level on Pn below which VImax becomes zero (x-intercept). In this case, Pcrit was 2 cm H2O and VImax varied in proportion to the level of Pn with a correlation of 0.98.

View larger version (34K): [in this window] [in a new window]   Figure 2.   Polygraphic recording showing the periods of hypopneas (a) and flow limitation (b) used for Pcrit measurement.

Once Pcrit was measured, the nCPAP estimated by the analysis of Pcrit (nCPAP_est) was defined as the value obtained adding 8 cm H₂O to Pcrit value. Tolerance for estimated nCPAP was relaxed to ± 1 cm H₂O.

From among the patients we selected a subgroup of 68 subjects in which at least 30 obstructive apneas during NREM sleep were available during the diagnostic night to respiratory effort analysis. In these patients the respiratory effort during obstructive apneas was measured according to the method previously described (8) considering apneas occurring during the Stages 1 and 2 of the NREM sleep in which three or more obstructed inspiratory efforts occurred. Apneas with artifacts on esophageal pressure or oxygen saturation tracings were rejected from the analysis. During apnea, the swings in esophageal pressure during all occluded efforts were measured and the following indices were calculated: the minimum of the esophageal pressure swings (Pes_min) recorded at the start of apnea, the maximum of the final one (Pes_max) recorded at the end of apnea, the difference (Pes) between the minimum and the maximum of the esophageal pressure swings, and the rate of increase in intrathoracic pressure (RPes) defined as the ratio of Pes to the duration of apnea. We also measured the lowest Sa_O2 value recorded after each apnea and the value before occlusion to determine for each apnea the decrease in oxygen saturation (Sa_O2) and its rate of decrease (RSa_O2) defined as the ratio of Sa_O2 to the apnea duration.

Statistical Analysis

Data in the text are expressed as means ± SEM. Data from nontreatment and treatment nights were compared using Student's paired t test. Comparison within male and female subjects was made using one-way analysis of variance with post-hoc Student-Newman-Keuls multiple range test. Bivariate correlation analysis using Pearson's correlation coefficient was performed to find the correlations between variables, and multiple regression analysis was done to define the contribution of Pcrit and other variables in explaining apnea density and efficacious nCPAP. The statistical significance level selected for all analysis was p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographics and Polysomnographic Variables

Clinical, anthropometric variables and polysomnographic data of the study patients are shown in Table 1.

The patients, 53.2 ± 1.0 yr of age with a mean BMI of 33.3 ± 0.7 kg/m², had a mean AHI of 71.5 ± 3.1 ranging from 12 to 140 and a mean AI of 43.2 ± 3.2. The mean time of sleep spent in apnea was 66.9 ± 5.8 ranging from 0 to 269 min. A wide range of nocturnal hypoxemia severity was present with a mean Sa_O2 of 93.3 ± 0.3% and a minimal Sa_O2 of 75.1 ± 1.2%. As a group, the patients had normal blood gas determinations, FEV₁, and FEV₁/FVC%. Eleven patients were hypoxemic, defined as having a Pa_O2  65 mm Hg (mean Pa_O2, 62.0 ± 0.8 mm Hg), 20 were hypercapnic, defined as having a Pa_CO2 45 mm Hg (mean Pa_CO2, 47.0 ± 0.3 mm Hg), and 19 patients had restrictive lung disease defined as a FEV₁/FVC% < 65 (mean FEV₁/FVC%, 56 ± 0.2%).

In the untreated night, nocturnal sleep structure was disrupted, with a mean total sleep time of 278 min, a mean sleep efficiency of 69.0 ± 0.1%, and a mean WASO of 176.2 ± 6.3 min. Sleep architecture showed a low amount of slow-wave sleep and rapid-eye-movement (REM) sleep and a high percentage of light sleep. REM sleep and sleep Stage 3-4 increased significantly with nCPAP treatment (p = 0.001), whereas NREM Stages 1 and 2 decreased significantly (p = 0.001). The application of effective nCPAP significantly reduced the AHI from 71.5 to 7.5 (p = 0.001).

When we compared the 89 male and the 17 female groups, they did not differ in terms of age, AHI, apnea time, and indices of nocturnal hypoxemia. Women were heavier (BMI, 38.7 ± 2.0 versus 32.3 ± 0.7; p = 0.0008), and they had lower neck circumference (39.8 ± 0.8 versus 43.8 ± 0.5 cm; p = 0.0005), lower FEV₁ (2.4 ± 1.4 versus 3.2 ± 0.9 L; p = 0.0002), and lower FVC (3.1 ± 1.6 versus 4.6 ± 1.0 L; p = 0.0001).

Pharyngeal Critical Pressure, Demographic and Anthropometric Variables, and AHI

In the study group the average measured Pcrit was 2.09 ± 0.1 cm H₂O, ranging from 0 to 4.5, and the mean Rus was 11.1 ± 0.5 cm H₂O/L/s, ranging from 2.9 to 35. When Pcrit and Rus were compared between women and men, Pcrit was more positive in men (2.2 ± 0.9 versus 1.6 ± 1.2 cm H₂O; p = 0.03), whereas upstream resistance was higher in the female subgroup (14.7 ± 5.8 versus 10.5 ± 5.2 cm H₂O/L/s, p = 0.003).

Correlation of Pcrit with the nocturnal and diurnal variables in the whole group of patients indicated a higher linear relationship with AHI and apnea time (Figure 3). There were no correlations between Pcrit and age (r = 0.08), BMI (r = 0.04), and neck circumference (r = 0.15).

View larger version (17K): [in this window] [in a new window]   Figure 3.   Scatterplots showing the correlation between AHI, apnea time, nCPAPeff and Pcrit.

To examine which variables affected most the severity of OSA, we performed Pearson's correlation analysis (Table 2) and a stepwise regression analysis considering AHI as a continuous variable. In the total group the AHI was correlated positively with BMI, neck circumference, and Pcrit and negatively with FVC, FEV₁, and daytime hypoxemia. The regression model for the entire sample predicted that 23% of the variation in AHI could be explained by three variables. Stepwise results indicated that the most important variable was neck circumference (R² = 0.17) followed by a small contribution of Pa_O2 (R² = 0.3) and Pcrit (R² = 0.3). No effect of sex and airway obstructive disease was revealed by regression analysis.

Pharyngeal Critical Pressure and Efficacious Nasal CPAP

In the group of patients as a whole, the mean level of effective nCPAP applied was 9.4 ± 0.2 cm H₂O, ranging from 5 to 16. Univariate correlation analysis (Table 2) was used to examine the effect of nocturnal and diurnal variables on nCPAP_eff. No effect was found for age and Pa_CO2. Regarding diurnal variables, the level of nCPAP_eff was correlated positively with BMI and neck circumference and negatively with Pa_O2, FEV₁, and FVC. Regarding nocturnal variables, the nCPAP_eff was strongly and positively correlated with the AHI and the apnea time and negatively with the indices of nocturnal hypoxemia. When we considered the measurement of upper airway collapsibility, we noted, as illustrated in Figure 3, that the nCPAP_eff was correlated significantly with the Pcrit. When all the above variables were considered in the multiple regression analysis, the regression model predicted that 36% of the variability in nCPAP_eff could be explained by four variables. Stepwise results indicated that the AHI was the most important variable predicting the 21% of the variation in the therapeutic nCPAP level, followed by BMI (R² = 0.8), Rus (R² = 0.4), and Pcrit (R² = 0.3).

Applying the nCPAP estimation by using the analysis of Pcrit, the mean level of nCPAP obtained after titration was not significantly different from the one obtained by the analysis of Pcrit value (nCPAP_est, 10.1 ± 0.1 versus nCPAP_eff, 9.4 ± 0.7 cm H₂O) (Figure 4a). However, only 36% of the patients were correctly estimated by the criteria (mean nCPAP_est, 10.2 ± 0.2; mean nCPAP_eff, 10.3 ± 0.2 cm H₂O). For 19% of the patients the criteria underestimated the level of efficacious pressure (mean nCPAP_est, 9.7 ± 0.3 versus 12.9 ± 0.4 cm H₂O) and in 45% of the patients it overestimated the pressure level (mean nCPAP_est, 10.1 ± 0.1 versus 7.3 ± 0.2 cm H₂O). Underestimation and overestimation occurred, respectively, for patients having a nCPAP_eff greater than 13 and lower than 7 cm H₂O (Figure 4b).

View larger version (26K): [in this window] [in a new window]   Figure 4.   The upper panel (a) illustrates the relationship between nCPAPeff and nCPAPest. The solid line is the identity line and the dashed line is the regression line. The bottom panel (b) shows the difference between nCPAPeff and nCPAPest at each titrated pressure level. These panels illustrate that despite a good correlation between nCPAPeff and nCPAPest, patients having a nCPAPeff above 13 cm H2O and below 7 cm H2O are either underestimated or overestimated.

To assess the direct relationship between nCPAP_eff and diurnal and nocturnal variables, including respiratory effort, we performed a correlation and stepwise regression analysis of 68 patients with OSA with complete data. These patients, 53.8 ± 1.2 yr of age, had a mean AHI of 81.6 ± 3.6, a minimal Sa_O2 of 72.2 ± 1.8, a mean nCPAP_eff of 9.8 ± 0.3 cm H₂O, and a mean Pcrit of 2.25 ± 0.12 cm H₂O. During the diagnostic night, 9,032 obstructive apneas in NREM sleep were analyzed, with a mean duration of 20.5 ± 0.5 s and a mean Sa_O2 of 8.7 ± 0.5%. On average, negative esophageal pressure increased gradually during apnea up to a mean maximal value recorded at the end of apnea of 37.9 ± 1.9 cm H₂O. The mean Pes was 18.4 ± 1.1 cm H₂O and the average rate of increase in intrathoracic pressure (RPes) was 0.91 ± 0.06 cm H₂O/s.

Correlation analysis in this group (Table 3) indicated a higher linear positive relationship between nCPAP_eff and AHI, BMI, and neck circumference and a negative one with Pa_O2, FEV₁, and FVC. Moreover, a strong correlation was noted between nCPAP_eff and all indices of respiratory effort. In this subgroup Pcrit was correlated positively with apnea time (r = 0.32, p = 0.008), Pes (r = 0.44, p = 0.001), Pes_max (r = 0.28, p = 0.02), and RPes (r = 0.37, p = 0.002) (Figure 5) and negatively with Sa_O2 (r = 0.34, p = 0.004). The regression model for this group predicted that 42% of the variance in nCPAP_eff was explained by RPes and BMI, with RPes making the largest contribution (R² = 0.33).

View larger version (11K): [in this window] [in a new window]   Figure 5.   The figure depicts the relationship between pharyngeal critical pressure and the maximal effort developed at the end of apnea (Pesmax) and the rate of increase in respiratory effort (RPes) in patients having analysis of respiratory effort.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have reported here the first large series of patients with obstructive sleep apnea in whom pharyngeal critical pressure has been determined. The large sample was expected to allow comparison of patients with different severities of the disease in adequate number to lead to a better characterization of the shape of the relationship linking Pcrit, anthropometric variables, AHI, and efficacious nCPAP. Our major findings were as follows. First, although male patients showed a greater level of pharyngeal collapsibility, neither age nor anthropometric variables, including BMI and neck circumference, contributed to variability in Pcrit. Second, despite Pcrit permitting the establishment of values to distinguish patients with mild disease from those with severe disease, the upper airway collapsibility has relatively little effect on the frequency of sleep-related breathing disorders. Finally, though Pcrit may influence the level of efficacious pressure, obesity and respiratory effort contributed mostly to variations in nCPAP.

Pharyngeal Critical Pressure and Anthropometric Variables

Correlation coefficients calculated to evaluate the relationship between individual pharyngeal collapsibility, age, and anthropometric variables revealed no significant relationship between these variables, whereas sex seems to affect UA compliance, with greater Pcrit in men.

The absence of a relationship between Pcrit, age, and obesity was in some way surprising since it is known that the frequency of sleep-disordered breathing increases with age and that Pcrit decreases after weight loss. Studies performed during wakefulness using tomography (11) or acoustic reflection techniques (12) have shown that the cross-sectional area and the upper airway caliber decrease with age, suggesting an age-related impairment of upper airway dilating muscle activity. However, our results do not show any impact of age on Pcrit. The discrepancy between the present results and the study quoted above is probably related to methodologic problems since the above techniques were performed while the subjects were awake, and thus they cannot be safely extrapolated to sleep.

Obesity, through fat deposits around the upper airway in the neck, may narrow the pharynx and alter its shape and mechanical properties, predisposing the patients to sleep-related obstruction. Ryan and Love (13) showed that obese patients with large necks tend to have more collapsible velopharynx during wakefulness. Of interest was our finding that the variables that may affect the AHI, i.e., neck circumference and obesity, were not correlated with Pcrit. This finding partially contrasts with a previous report showing a reduction in apnea severity when Pcrit falls to negative values after weight loss (14). The decrease in pharyngeal collapsibility after weight loss has been explained as a consequence of reduced fat deposition on the pharynx and of an improvement in the activity of UA muscles. However, the absence of a link between pharyngeal collapsibility, as measured by Pcrit, and obesity, suggests that the effect of weight loss on pharyngeal collapsibility may act more through changes in lung function (15) than in upper airway patency or muscle activity.

The strong male predominance in OSA may be related to sex differences in upper airway size and collapsibility. Previous studies have shown differences in pharyngeal size (16) and in pharyngeal resistance (17) in men, accounting for the male predisposition to upper airway occlusion. Studies analyzing electromyogram genioglossal activity during wakefulness have demonstrated that even though no differences in upper airway resistance are present between men and women, men present greater collapsibility (18). In keeping with this report we noted that our measurement of upper airway collapsibility performed during sleep, i.e., Pcrit, was significantly greater in men, confirming that women have more stable airways and less collapsible pharynxes.

Influence of Pharyngeal Critical Pressure on Upper Airway Occlusion

In the present study we observed an increase in Pcrit related to the density of apneas and hypopneas similar to those described in previous reports (4), suggesting that individual upper airway collapsibility may affect the severity of the syndrome. These results have clinical implication since the measurement of Pcrit may be a powerful tool for discriminating patients with mild to moderate OSA from those with severe OSA. However, the strength of the association was not large and most of the changes in the AHI were related to neck circumference rather than to Pcrit. We are not certain why we observed different results from those reported previously. A potential explanation for the difference includes the fact that subjects suspected of having OSA were predominantly studied. However, despite our recruitment method, the patients belonged to a heterogeneous group in terms of BMI and AHI and subjects were not excluded from the study on the basis of pulmonary function test or blood gas analysis. Moreover, the use of a positive pressure for Pcrit measurement may be considered as the major limit of our study. From previous studies (4) we know that Pcrit is positive or near to zero in patients with hypopnea and apnea syndrome, whereas it exhibits negative values in patients who snore. Consequently, the determination of Pcrit during routine nCPAP titration with a start pressure level of 2 cm H₂O might be unsuitable for an accurate estimation of pharyngeal collapsibility in patients who snore and have upper airway resistance syndrome. Even though the methodology used in this study may artificially increase Pcrit, our values in patients with OSA are slightly higher than those previously reported by using subatmospheric pressure.

Although Pcrit increased with the severity of the disease and influenced the AHI and the apnea time, this measurement does not wholly explain why one individual mainly has obstructive apneas. In our group, neck circumference was the stronger predictor of AHI, with Pcrit explaining the 3% of the variance in apnea frequency. This would suggest that though individual susceptibility to pharyngeal collapse may predispose to UA occlusion, tendency of upper airways to occlude may be a multifactorial phenomenon requiring the operation of several factors (19). In the model of the upper airway as a Starling resistor system, VI_max is modulated by the Pcrit and by the resistance upstream to the flow-limiting site (6). One possibility that would resolve the discrepancy between the minor effect of Pcrit described herein and the relationship with AHI is that several factors may affect this measurement such as neck extension (20), body position, changes in lung volume (21), and mouth opening (22). Moreover, mechanisms acting downstream from the flow-limiting site may influence Pcrit and consequently pharyngeal collapsibility. We know that VI_max may decrease when intrathoracic pressure is more negative (23, 24) as a consequence of the changes in transmural pressure (25) or in the pharyngeal cross-sectional area (26). The close relationship found between Pcrit, RPes, and Pes that reflect the mechanical effects of inspiratory drive (27), suggest that changes in central inspiratory drive (28) or in the mechanical response to obstruction (29) may modify upper airway muscle activity and therefore upper airway resistance and compliance.

Pharyngeal Critical Pressure and Efficacious nCPAP

Another major finding of our study was that, despite a significant relationship between Pcrit and nCPAP pressure, pharyngeal collapsibility alone does not significantly influence the level of optimal nCPAP. In the total study sample only 3% of the variation in nCPAP_eff was explained by Pcrit, whereas AHI and BMI were the stronger predictors. Moreover, when we consider the patients undergoing analysis of respiratory effort, Pcrit was not sensitive enough to predict the level of nCPAP_eff, whereas, as previously described (8), the indices of respiratory effort strongly affected the efficacious pressure, the greatest level of nCPAP occurring in the subjects with higher RPes. The physiologic significance of this observation is not clear, but the close relationship demonstrated that additional factors might oppose to pharyngeal patency restoration in an occluded upper airway. In the pure Starling resistor model only two states are possible: the upper airway collapses when intraluminal pressure is lower than external pressure, i.e., Pn, and distends when intraluminal pressure exceeds external pressure. However, studies in patients with OSA (2) have demonstrated that a subatmospheric pressure is necessary to close the pharynx, but the return to atmospheric pressure does not fully open the airways. Furthermore, using forced oscillation technique, Farré and coworkers (30) showed that when nCPAP pressure was raised just above Pcrit, the effective critical pressure, i.e., the pressure at which airflow limitation or collapse occurs, is higher than the value obtained when nCPAP is below Pcrit. These observations may suggest that during nCPAP titration the increased inspiratory effort may compress the distended segment, reducing intraluminal pressure and increasing collapsibility. Only when nCPAP is raised sufficiently above Pcrit to compensate for the major inspiratory effort is the upper airway fully distended (31). As a consequence, greater levels of nCPAP would be required to abolish flow-limitation (32) and snoring (33) in comparison to periods of sleep with complete occlusion.

Given the little contribution of pharyngeal critical pressure to nCPAP_eff, it is not surprising that Pcrit does not adequately predict the optimal level of nCPAP. According to the criteria of Gold and Schwartz (9), only 36% of the patients have clinically accurate estimations. The major disagreement was noted for patients having nCPAP_eff values above 13 and below 7 cm H₂O. One explanation for the difference in the estimation may be the effect of the treatment on pharyngeal collapsibility. Analyzing the relationship between VI_max and the level of nCPAP, Schwartz and coworkers (34) showed that at higher levels of nCPAP, an increase in VI_max is present, in spite of the increase in Pcrit. They explained these contradictory results as being a consequence of a drop in upstream resistance determined by reduced activation of upper airway dilator muscles. However, if this hypothesis could explain the inaccuracy of the criterion for patients with higher levels of nCPAP, an error of estimation is still present for subjects with lower levels. Moreover, since accurate estimation of the nCPAP level is much more frequent for a pressure of 10 cm H₂O, the mean level often used during nCPAP titration, we cannot exclude that an nCPAP_eff within 10 cm H₂O is more likely to occur simply by chance in an OSA population.

In conclusion, we have found in patients with OSA a rise in pharyngeal critical pressure related to apnea time and apnea frequency, suggesting that the measurement of Pcrit may be a useful tool in clinical practice. However, even though the Pcrit measure permitted a certain level of discrimination between patients, pharyngeal collapsibility alone does not optimally assess the severity of sleep apnea and the efficacious nasal continuous positive airway pressure. The clinical implications of the measurement of pharyngeal collapsibility in clinical practice would require a larger sample of patients with a wider AHI distribution and the use of other parameters of upper airway function and structure in order to determine the primary contribution of pharyngeal collapsibility in the pathogenesis of obstructive sleep apnea.