Discriminating the severity of pharyngeal collapsibility in men using anthropometric and polysomnographic indices

Abstract

Study Objectives:

Although obstructive sleep apnea results from the combination of different pathophysiologic mechanisms, the degree of anatomical compromise remains the main responsible factor. The passive pharyngeal critical closing pressure (Pcrit) is a technique used to assess the collapsibility of the upper airway and is often used as a surrogate measure of this anatomical compromise. Patients with a low Pcrit (ie, less collapsible airway) are potential candidates for non–continuous positive airway pressure therapies. However, Pcrit determination is a technically complex method not available in clinical practice. We hypothesized that the discrimination between low and high Pcrit can be estimated from simple anthropometric and polysomnographic indices.

Methods:

Men with and without obstructive sleep apnea underwent Pcrit determination and full polysomnography. Receiver operating characteristics analysis was performed to select the best cutoff of each variable to predict a high Pcrit (Pcrit ≥ 2.5 cmH₂O). Multiple logistic regression analysis was performed to create a clinical score to predict a high Pcrit.

Results:

We studied 81 men, 48 ± 13 years of age, with an apnea-hypopnea index of 32 [14–60], range 1–96 events/h), and Pcrit of −0.7 ± 3.1 (range, −9.1 to +7.2 cmH₂O). A high and low Pcrit could be accurately identified by polysomnographic and anthropometric indices. A score to discriminate Pcrit showed good performance (area under the curve = 0.96; 95% confidence interval, 0.91–1.00) and included waist circumference, non–rapid eye movement obstructive apnea index/apnea-hypopnea index, mean obstructive apnea duration, and rapid eye movement apnea-hypopnea index.

Conclusions:

A low Pcrit (less collapsible) can be estimated from a simple clinical score. This approach may identify candidates more likely to respond to non–continuous positive airway pressure therapies for obstructive sleep apnea.

Citation:

Genta PR, Schorr F, Edwards BA, Wellman A, Lorenzi-Filho G. Discriminating the severity of pharyngeal collapsibility in men using anthropometric and polysomnographic indices. J Clin Sleep Med. 2020;16(9):1531–1537.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Upper airway anatomical compromise is an important factor leading to obstructive sleep apnea in many patients and can be quantified by measuring the passive pharyngeal critical closing pressure. Non–continuous positive airway pressure therapeutic options for obstructive sleep apnea may perform better for patients without a highly collapsible airway. However, passive pharyngeal critical closing pressure is only performed in a few research laboratories because of technical challenges.

Study Impact: We developed a simple clinical score that uses anthropometric and polysomnographic indices that can accurately discriminate a low from a high passive pharyngeal critical closing pressure. Using a clinical score to discriminate passive pharyngeal critical closing pressure may improve the selection of non–continuous positive airway pressure alternatives for patients with obstructive sleep apnea.

INTRODUCTION

Obstructive sleep apnea (OSA) is a common disorder among adults, with adverse health consequences. Although many therapeutic options are available, OSA treatment remains a challenge because of variable adherence and efficacy of available therapies.^1,2 Although continuous positive airway pressure (CPAP) therapy is highly effective, it is often not well tolerated.¹ Despite being preferred by patients, non-CPAP alternatives such as mandibular advancement devices, are less often prescribed because of lower efficacy and lack of good predictors of outcome.²

The pathogenesis of OSA can be thought of as resulting from at least 4 major pathophysiologic traits: abnormalities of upper airway size/collapsibility, arousal threshold, upper airway muscle responsiveness, and ventilatory control sensitivity or loop gain.³ Although the amount that each trait contributes to an individual’s OSA is highly variable, the upper airway size/collapsibility is by far the most important feature among all traits linked to OSA.³ The most common way that the upper airway anatomy/collapsibility is assessed is by measuring the individual passive critical closing pressure (Pcrit).⁴ The interest of determining Pcrit is to grade the severity of pharyngeal collapsibility. Patients with a high Pcrit (ie, more collapsible airway) may only respond to CPAP. In contrast, patients with a less compromised Pcrit are more likely to respond to non-CPAP therapies, including mandibular advancement devices,^5,6 weight loss, or even medications.⁷ Despite the usefulness of Pcrit, its determination is laborious, requires specialized equipment, and is only performed in a few physiologic laboratories in the world.⁸ Therefore, a simple tool to discriminate low and high Pcrit from existing clinical and polysomnographic (PSG) data is of significant interest.

PSG is currently the gold standard study to diagnose OSA. Although several variables are described in a polysomnography report, the apnea-hypopnea index (AHI) is one of the few variables that is considered in the management of patients with OSA. Although obstructive apneas imply a more significant respiratory compromise than hypopneas, the absolute number of apneas and hypopneas are summed together to provide a single index of severity that quantifies the frequency of events but not the severity of airway obstruction. This could help to explain why the AHI does not correlate well with OSA symptoms and cannot accurately predict response to a non-CPAP therapeutic option such as mandibular advancement, surgery, and pharmacologic interventions.^9–12 Pcrit has been shown to be higher among patients with a predominance of obstructive apneas compared with those with a predominance of hypopneas, highlighting the different anatomical compromise between hypopneas and apneas.¹³ Therefore, the frequency of hypopneas and obstructive apneas may predict Pcrit. In addition, Pcrit has been shown to be associated with obesity.¹⁴ We hypothesized that obesity-related anthropometric indices and polysomnographic indices will predict Pcrit. To address this hypothesis, we performed full PSG and Pcrit measurements in a group of participants with a wide range of AHI (from normal to severe OSA). The importance of this research is that it will allow us to distinguish patients with OSA with a low and high upper airway collapsibility from currently available clinical information (ie, anthropometric and PSG indices) and, perhaps more importantly, assist with a simple way to estimate response to non-CPAP therapies.

METHODS

Participants

All participants provided informed written consent before study entry, which was approved by the Hospital das Clínicas Ethics Committee.

PSG

All participants underwent a baseline standard overnight polysomnogram (Alice 5, Philips Respironics, Murrysville, PA). Monitoring included electroencephalography, electrooculography, electromyography, oximetry, airflow (oronasal thermistor and pressure cannula), and rib cage and abdominal movements, as previously described.¹⁵ Apneas were defined as complete cessation of airflow (thermistor) for at least 10 seconds. Hypopneas were defined as a >30% reduction in airflow (nasal pressure) for at least 10 seconds associated with oxygen desaturation of 3% or cortical arousal. AHI was calculated as the total number of respiratory events (apneas plus hypopneas) per hour of sleep. OSA was defined as an AHI > 15 events/h.

Upper airway collapsibility determination (passive Pcrit)

Sleep induction

Sleep induction for Pcrit determination was performed in the sleep laboratory, starting at 8:00 am following the diagnostic PSG, as previously described.¹⁵ A peripheral vein was cannulated and maintained with a continuous infusion of saline solution. Midazolam (diluted in a saline solution with a concentration of 1 mg/10 mL) was infused slowly as follows: sleep induction started with the infusion of 5 mL solution over 5 minutes, which was repeated as needed until a sleep-like state was detected through online PSG monitoring. Midazolam infusion was restarted for sleep maintenance only if the patient awoke and was not able to fall asleep again for 10 minutes.

Pcrit

During Pcrit determinations, the same polysomnographic channels used in the diagnostic PSG were recorded except for nasal pressure and thermistor. Pcrit measurements were performed with patients in the supine position. Subjects were fitted with a nasal mask attached to a heated pneumotachograph (3700A, Hans Rudolf, Kansas City, MO) and a differential pressure transducer (MP45-14-871, Validyne, Northbridge, CA) for measurement of airflow. Mask pressure was measured by another pressure transducer (Validyne). Respiratory signals (airflow and mask pressure) were recorded on a personal computer using an analog-to-digital converter (PCI-6014, National Instruments, Austin, TX) and data acquisition software (LabVIEW, National Instruments). A modified CPAP device (Philips Respironics) that could deliver both positive and negative airway pressure was connected to the mask. After sleep onset, airway pressure was increased to abolish airflow limitation. This level was used as the holding pressure for each patient. Once stable stage N2 or N3 sleep was observed at the holding pressure, CPAP was abruptly reduced by 1–2 cmH₂O during expiration and was held at this level for 5 breaths. The pressure was then returned to the holding pressure for 1 minute before being dropped a further 1–2 cmH₂O for another 5 breaths. This process of progressively dropping CPAP continued until obstructive apnea occurred. If an arousal occurred during any pressure drop, the CPAP was returned to the holding level until stable sleep resumed. The entire procedure of progressively dropping CPAP until obstruction occurred was repeated 3–5 times in each participant. Data were analyzed using custom-designed software (Matlab, The MathWorks, Inc., Natick, MA) to determine the peak inspiratory flow for breaths 3–5 during the pressure drop. Each of these breaths was assessed for the presence or absence of inspiratory flow limitation defined as the flattening of inspiratory flow after initial peak.¹⁶ Breaths associated with arousal were excluded from the analysis. Nasal mask pressure for flow-limited breaths was plotted against V_imax. Pcrit was determined as the zero-flow intercept from the linear regression of V_imax vs nasal mask pressure as previously described.^4,15

Statistical analysis

Continuous variables are reported as mean ± standard deviation or median [interquartile range] as appropriate. Normal distribution was tested using the Shapiro-Wilk test. PSG-derived variables were used regardless of body position. Correlation analysis was performed using Spearman rank correlation to test for associations between variables. To test if Pcrit would differ according to the proportion of non–rapid eye movement sleep (NREM) obstructive apnea index (OAI) to AHI (NREM-OAI/AHI), Pcrit was compared among NREM-OAI/AHI quartiles using the Kruskal-Wallis H test. Post hoc comparisons between groups was performed using Dunn’s test with Bonferroni adjustment. Receiver operating characteristic (ROC) curves were plotted to test the performance of AHI, OAI, NREM-OAI/AHI ratio, minimum O₂ saturation, and mean apnea duration to predict Pcrit. Because there is no well-defined cutoff to define a low and high Pcrit threshold, a range of Pcrit cutoffs was tested (1.0, 1.5, 2.0, and 2.5 cmH₂O). Among the tested cutoffs, an optimum Pcrit cutoff was chosen based on the highest area under the curve (AUC). We then selected cutoffs for each variable tested to predict Pcrit (as defined by the optimum cutoff) based on the best combination of sensitivity and specificity obtained from ROC analysis. A univariate logistic regression analysis was performed for each variable considering Pcrit at the defined optimum cutoff as the dependent variable. Multiple logistic regression analysis was then performed to determine independent predictors of Pcrit using variables that had P < .1 at univariate analysis. A clinical score was then constructed observing the odds ratios obtained at the multiple logistic regression model. ROC analysis was done to test the performance of the clinical score to predict a Pcrit > 2.5 cmH₂O. Cross-validated estimates of sensitivity and specificity were obtained by the “leave-one-out” method.

RESULTS

Eighty-one male participants were studied. Participants were overweight (BMI [body mass index] = 28.5 [26.6–32.2] kg/m²), with a mean age of 48 ± 13 years. The median AHI was 32.0 [13.5–60.1] and ranged from 1.3 to 96.1 events/h. Mean Pcrit was −0.7 ± 3.1 and ranged from −9.1 to +7.2 cmH₂O.

We initially tested the associations of Pcrit and PSG indices. Pcrit was associated with NREM-AHI, REM-AHI, NREM-OAI/AHI, mean obstructive apnea duration, BMI, and waist circumference (P < .01; Figure 1). Pcrit was also associated with NREM-OAI (r = .662, P < .001), minO₂ saturation (r = −.557, P < .001), and neck circumference (r = .451 P < .001). When overall AHI instead of NREM-AHI was used, similar associations were observed between Pcrit and overall AHI (r = .574, P < .001) and between Pcrit and overall OAI/AHI (r = .635, P < .001). Pcrit was not associated with NREM-AHI (r = .215, P = .055). Pcrit differed significantly according to the proportion of NREM obstructive apneas (NREM-OAI/AHI) quartiles (P < .05; Figure 2).

Figure 1. Associations between Pcrit and polysomnographic and anthropometric indices.

AHI = apnea-hypopnea index, NREM = non–rapid eye movement sleep, OA = obstructive apnea, Pcrit = passive pharyngeal critical closing pressure, REM = rapid eye movement sleep.

Figure 2. Pcrit according to NREM-OAI/AHI quartiles. Individual data are shown as hollow circles and median values as solid markers.

AHI = apnea-hypopnea index, NREM = non–rapid eye movement sleep, OAI = obstructive apnea index, Pcrit = passive pharyngeal critical closing pressure.

Using ROC analysis, the best threshold (from 1 to 2.5 cmH₂O) for Pcrit (reference variable) according to different classification variables (AHI, OAI, and NREM-OAI/AHI) was then tested. The Pcrit threshold with the highest AUC for all tested classification variables was 2.5 cmH₂O (Table 1). Anthropometric, PSG, and Pcrit data for the whole group, as well as subdivided according to a Pcrit ≤2.5 or >2.5 cmH₂O are shown in Table 2. A cutoff for each classification variable was then selected from the ROC curves using the best combination of sensitivity and specificity. Univariate logistic regression analysis was then performed using the cutoff of each variable to predict a Pcrit > 2.5 cmH₂O (Table 3).

Table 1.

Area under the curve values obtained from ROC analysis using different Pcrit cutoffs.

	Pcrit > 1	Pcrit > 1.5	Pcrit > 2.0	Pcrit > 2.5
Age	0.67	0.66	0.65	0.64
BMI	0.68	0.69	0.74	0.87
Neck circumference	0.72	0.71	0.73	0.82
Waist circumference	0.74	0.75	0.78	0.85
NREM-AHI	0.78	0.82	0.88	0.92
REM-AHI	0.62	0.60	0.64	0.68
NREM-OAI	0.83	0.86	0.91	0.92
mOA duration	0.74	0.76	0.82	0.86
NREM-OAI/AHI	0.83	0.85	0.90	0.91
MinO2sat	0.80	0.79	0.85	0.89

Pcrit = passive pharyngeal critical closing pressure, BMI = body mass index, NREM = non–rapid eye movement sleep, AHI = apnea-hypopnea index, REM = rapid eye movement sleep, OAI = obstructive apnea index, mOA duration = mean obstructive apnea duration, MinO₂sat = minimum oxygen saturation.

Table 2.

Anthropometric, polysomnographic, and Pcrit data for the whole group and according to a Pcrit ≤ or >2.5 cmH₂O.

Variables	Whole group (n = 81)	Pcrit ≤ 2.5 cmH2O (n = 70)	Pcrit > 2.5 cmH2O (n = 11)	P
Age, yr	48 ± 13	47 ± 14	54 ± 8	.097
BMI, kg/m2	28.5 [26.6–32.2],	28.0 [26.1–30.7]	33.0 [31.3–33.8]	<.001
Neck circumference, cm	41 [39–44]	41 [39–43]	44 [44–46]	<.001
Waist circumference, cm	101 ± 11	100 ± 10	114 ± 9	<.001
AHI, events/h	32.0 [13.5–60.1]	24.5 [12.3–48.6]	78.5 [66.6–82.8]	<.001
NREM-AHI, events/h	33.3 [12.3–63.8]	24.1 [10.6–50.5]	81.4 [65.6–83]	<.001
REM-AHI, events/h	25.2 [10.1–47.7]	23.4 [10.1–44.1]	47.7 [39.9–62.5]	.063
NREM-OAI, events/h	3.6 [0.2–19.8]	1.1 [0–13.4]	49.6 [31.5–68.5]	<.001
mOA duration, s	18.2 [10.8–22.7]	17.3 [0–20.5]	26.5 [22.1–31.8]	<.001
NREM-OAI/AHI	0.11 [0.01–0.44]	0.08 [0–0.27]	0.66 [0.44–0.88]	<.001
MinO2 sat, %	83 [73–88]	84 [76–88]	65 [60–73]	<.001
Pcrit, cmH2O	−0.7 ± 3.1	−1.5 ± 2.6	4.0 ± 1.5	<.001

Data is shown as mean ± SD or median [25th–75th percentile]. Pcrit = passive pharyngeal critical closing pressure, BMI = body mass index, AHI = apnea-hypopnea index, NREM = non–rapid eye movement sleep, REM = rapid eye movement sleep, OAI = obstructive apnea index, mOA duration = mean obstructive apnea duration, MinO₂sat = minimum oxygen saturation.

Table 3.

Area under the curve values obtained from ROC analysis and odds ratio obtained from univariate logistic regression analysis using a Pcrit cutoff of 2.5 cmH₂O.

	AUC (95% CI)	Cutoff	Sensitivity (%)	Specificity (%)	OR (95% CI)	P
Age	0.64 (0.50–0.78)	54	63	63	3.0 (0.8–11.1)	.107
BMI	0.87 (0.78–0.95)	31.3	82	79	16.5 (3.2–84.6)	<.001
Neck circumference	0.82 (0.72–0.92)	43.5	82	76	14.0 (2.8–71.4)	.001
Waist circumference	0.85 (0.76–0.94)	106	82	74	13.0 (2.6–65.9)	<.01
NREM-AHI	0.92(0.85–0.98)	63.8	91	84	24.1 (4.6–127.2)	<.001
REM-AHI	0.68(0.47–0.89)	39.9	80	73	12.1 (2.4–61.1)	.003
NREM-OAI	0.92(0.86–0.99)	31.5	82	87	30.5 (5.7–164.4)	<.001
mOA duration	0.86 (0.76–0.96)	22.1	82	79	16.5 (3.2–84.6)	.001
NREM-OAI/AHI	0.91(0.82–0.99)	0.44	82	84	24.1 (4.6–127.2)	<.001
MinO2 sat	0.89 (0.81–0.97)	73	83	82	21.8 (4.2–113.6)	<.001

P values obtained from logistic regression analysis. Cutoffs and sensitivity/specificity for each specific variable are shown. ROC = receiving operator characteristic, Pcrit = passive pharyngeal critical closing pressure, AUC = area under the curve, CI = confidence interval, OR = odds ratio, BMI = body mass index, NREM = non–rapid eye movement sleep, AHI = apnea-hypopnea index, REM = rapid eye movement sleep, OAI = obstructive apnea index, mOA duration = mean obstructive apnea duration, MinO₂sat = minimum oxygen saturation.

Multiple logistic regression analysis revealed that NREM-OAI/AHI, waist circumference, obstructive apnea duration, and REM-AHI were independent predictors of Pcrit > 2.5 cmH₂O (Table 4). A clinical score was generated according to the observed odds ratios for each variable. The presence of NREM-OAI/AHI > 0.44 yielded 2 points, and waist circumference > 106 cm, mean obstructive apnea duration > 22.1 seconds, and REM-AHI > 39.9 yielded 1 point each. ROC analysis to test the performance of the clinical score to predict a Pcrit > 2.5 cmH₂O is shown in Table 5 (AUC = 0.96). A score ≥3 showed a sensitivity of 90.9% and a specificity of 84.3%. Cross-validation using the leave-one-out method yielded AUC between 0.95 and 0.98, sensitivity between 80.0% and 90.0%, and specificity between 94.2% and 95.7%. To address the contribution of different variables to predict Pcrit, we performed a multiple linear regression analysis with the same variables used in the multiple logistic regression model (Pcrit > 2.5 cmH₂O as the dependent variable and NREM-OAI/AHI, waist circumference, obstructive apnea duration, and REM-AHI as independent predictors). Obstructive apnea duration did not remain a significant predictor in the linear model and therefore was excluded from the model. The remaining model showed that NREM-OAI/AHI had the highest standardized β-coefficient (0.431). Waist circumference and REM-AHI had similar influence to predict Pcrit (standardized B coefficient of 0.247 and 0.237, respectively; Table 6).

Table 4.

Multiple logistic regression analysis to predict a Pcrit > 2.5 cmH₂O.

Predictors	OR (95% CI)	P
NREM-OAI/AHI > 0.44	26.9 (2.2–327.6)	.010
Waist circumference > 106 cm	15.4 (1.2–200.0)	.037
mOA duration > 22.1 s	14.1 (1.0–193.6)	.048
REM-AHI > 39.9 events/h	15.3 (1.2–201.1)	.038

Pcrit = passive pharyngeal critical closing pressure, OR = odds ratio, CI = confidence interval, NREM = non–rapid eye movement sleep, OAI = obstructive apnea index, AHI = apnea-hypopnea index, mOA duration = mean obstructive apnea duration, REM = rapid eye movement sleep.

Table 5.

Performance of the clinical score to predict a Pcrit > 2.5 cmH₂O.

Cutpoint	Sensitivity (%)	Specificity (%)
≥0	100	0
≥1	100	45.7
≥2	100	70.0
≥3	90.9	84.3
≥4	81.8	94.3
≥5	36.4	100

Area under the curve = 0.96 (95% CI, 0.91−1.00). Point system: NREM-OAI/AHI > 0.44, 2 points; waist circumference > 106 cm, 1 point; mean obstructive apnea duration > 22.1 seconds, 1 point; REM-AHI > 39.9 events/h, 1 point. Pcrit = passive pharyngeal critical closing pressure, CI = confidence interval, NREM = non–rapid eye movement sleep, OAI = obstructive apnea index, AHI = apnea-hypopnea index, REM = rapid eye movement sleep.

Table 6.

Multiple linear regression analysis to predict a Pcrit > 2.5 cmH₂O.

Predictors	B	Standard Error	Standardized β	P
Constant	-0.067	0.041		.106
NREM-OAI/AHI > 0.44	0.343	0.071	0.431	<.001
Waist circumference > 106 cm	0.179	0.067	0.247	.009
REM-AHI > 39.9 events/h	0.171	0.066	0.237	.012
Model			R = 0.648	<.001

Final model characteristics appear in bold. Pcrit = passive pharyngeal critical closing pressure, NREM = non–rapid eye movement sleep, OAI = obstructive apnea index, AHI = apnea-hypopnea index, REM = rapid eye movement sleep.

DISCUSSION

The main finding of this study is that PSG indices and anthropometric variables were able to accurately discriminate patients with a high or low Pcrit. A simple clinical score with 4 easily obtainable variables was able to discriminate patients with a highly collapsible airway (Pcrit > 2.5 cmH₂O). This approach could potentially be used by clinicians to improve the selection of treatment for patients with OSA by allowing the identification of patients with a low collapsible airway that are potential candidates for non-CPAP therapies.

A major interest to determine Pcrit among patients with OSA is to assess the degree of upper airway anatomical compromise, which may influence the choice of therapy offered. The associations between Pcrit and both BMI and waist circumference reported in the present study confirms the influence of obesity on Pcrit and have been reported before.^17,18 We showed that Pcrit was higher among patients with a predominance of obstructive apneas compared with those with a predominance of hypopneas, which is in line with the study by Gleadhill et al.¹³ In addition, the preponderance of apneas had the largest impact to predict Pcrit in our study. Some studies that assessed a non-CPAP–specific intervention for OSA reported a greater effect on the hypopnea index compared with the apnea index,^10,19,20 suggesting that patients with a higher proportion of hypopneas are better candidates for such therapeutic alternatives. In our study, Pcrit was also associated with NREM AHI, REM-AHI, and mean apnea duration, which confirms previous reports.^17,18 Readily available PSG indices beyond the AHI offer useful additional information that may be clinically useful.

Measuring the underlying causes of OSA has been proposed to allow better treatment selection.³ The degree of upper airway compromise, we believe, will dictate the relative importance of the other traits. Individuals with a highly collapsible airway will be less likely to respond to treatments that aim to reduce loop gain or increase the arousal threshold, as well as therapies that offer limited improvement of upper airway collapsibility (mandibular advancement, weight loss, oropharyngeal surgery).^3,5,6,8 We showed that a clinical score that includes the proportion of obstructive apneas, waist circumference, mean apnea duration, and REM-AHI had good performance to distinguish low and high Pcrit. A Pcrit cutoff (+2.5 cmH₂O) was chosen to define a highly collapsible airway among those that had the best AUC using ROC analysis. This value is very similar to the +2.0 cmH₂O Pcrit cutoff considered as a highly collapsible airway in a previous study.³ Therefore, the prediction of a high Pcrit is possible through a simple clinical score that may be useful in the selection of non-CPAP alternatives for patients with OSA.

CPAP is usually recommended as the first therapeutic choice for patients with severe OSA. In contrast, CPAP is usually not the patient´s preference when an alternative is offered.²¹ In addition, adherence is frequently lower for CPAP compared with other choices.^22,23 However, a wide range of Pcrit values can be found among patients with severe OSA (Figure 1). Non-CPAP therapies may still be considered as alternatives to CPAP for selected patients with severe OSA. Although generally considered for mild-to-moderate OSA severity, non-CPAP alternatives may be effective in patients with severe OSA.^5,6,10,19,24 Recent studies using non-CPAP alternatives for OSA assessed the key OSA phenotypical traits including Pcrit. Both a low Pcrit and low loop gain were predictors of positive response to oral appliances in a study that involved patients with severe OSA.⁵ Responders to the combination of a sedative (eszopiclone) and oxygen had a lower pharyngeal collapsibility and greater upper airway responsiveness.⁷ Taken together, patients with OSA with a low Pcrit, even with severe disease, are more likely to respond to non-CPAP alternatives.

Our study has several limitations that must be considered. First, we studied only men. Because men and women have several differences in the presentation and pathophysiology of OSA, the results of our study may only be applicable among men. Second, our Pcrit measurements were performed after sleep induction with midazolam rather than during natural sleep. However, we have previously shown that Pcrit measured using small doses of midazolam as used in the present study are comparable to upper airway collapsibility values obtained during natural sleep.¹⁵ Third, the impact of Pcrit discrimination on non-CPAP (weight loss, mandibular advancement devices, surgery) therapeutic outcomes was not tested in the present study. Fourth, we did not include patients with OSA who were morbidly obese in our study, limiting the application of our findings among these patients. Last, the cutoffs defined by our study should be confirmed in a different sample that includes both men and women to confirm their performance and expand the generalizability of our findings.

In conclusion, we were able to develop a simple clinical score using anthropometric and PSG indices to discriminate patients with a high Pcrit. Because several treatment options for OSA depend on a non–severely compromised airway to be effective, this approach may improve the selection of OSA therapy and provide more treatment options for a significant number of patients with OSA.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at Laboratorio do Sono, LIM 63, Pulmonary Division, Heart Institute (InCor), Hospital das Clínicas HCFMUSP, Universidade de Sao Paulo, Sao Paulo, SP, Brazil. Dr. Edwards receives grant support from the National Health and Medical Research Council of Australia and ApniMed. Dr. Wellman works as a consultant for Somnifix, Cambridge Sound Management, Nox, Bayer, Philips, Galvani and Inspire and has received grants from Philips and Somnifix. He also has a financial interest in Apnimed Corp., a company developing pharmacologic therapies for sleep apnea. Dr. Wellman’s interests were reviewed and are managed by Brigham and Women’s Hospital and Partners HealthCare in accordance with their conflict of interest policies. All other authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

AUC

area under the curve

CPAP

continuous positive airway pressure

NREM

non–rapid eye movement sleep

OAI

obstructive apnea index

OSA

obstructive sleep apnea

Pcrit

passive pharyngeal critical closing pressure

PSG

polysomnography

REM

rapid eye movement sleep

ROC

receiver operating characteristic