Receiver operating characteristics of impulse oscillometry parameters for predicting obstructive sleep apnea in preobese and obese snorers

Obstructive sleep apnea (OSA) is a common disorder characterized by a repetitive collapse of the pharyngeal airway during sleep. It manifests as reduced (hypopnea) or absent (apnea) airflow at the nose/mouth, despite ongoing inspiratory efforts, which is terminated by a transient arousal from sleep and restoration of upper-airway patency [1, 2]. Consequently, this pathophysiological process causes disturbances in blood gases and sleep structure and has also been associated with cardio-cerebrovascular complications [3‐5].

Nocturnal polysomnography (PSG) in a sleep laboratory is the standard method for diagnosing OSA. However, PSG is expensive, time-consuming, and labor-intensive. There is a high prevalence of OSA and the associated syndrome (OSAS) in obese populations, yet few centers can afford to perform the diagnostic procedure on all patients presenting with OSA. In this context, simpler and less expensive tests are needed.

Even through physical upper-airway stenosis and increased pharyngeal airway collapsibility during expiration plays an important role in these patients, [6, 7] it is expected that the cycle of obstruction and restoration of upper-airway patency from arousal will be accompanied by large swings in intrathoracic pressure and changes in the mechanical properties of the respiratory system [2]. Van Noord et al. reported that respiratory resistance (Rrs) at low oscillatory-frequencies measured by the forced-oscillation technique (FOT) increased with a decrease in oscillatory frequency in patients with upper-airway obstruction; Also, this patients was showed a decrease in respiratory reactance (Xrs) with a increase of oscillatory frequency [8]. Zerah et al. demonstrated that Rrs increases linearly with a decrease of oscillatory frequency over the 4–16 Hz frequency range in obese subjects. This characteristic can be used to extrapolate total Rrs, that is, the zero-frequency resistance (Rrs0), which was found to increase with the level of obesity as a result of a reduction in lung volume [9]. Consequently, Zerah et al. analyzed Rrs data obtained in obese OSA patients with a sitting posture and back-extrapolated the regression line to 0 Hz to obtain the Rrs0 parameter, and its inverse–respiratory conductance (Grs) as well as specific Grs (sGrs) was calculated: the ratio of Grs over functional residual capacity (FRC), and those FOT parameters was found to be independent from their BMI associated with OSA severity [10]. Additionally, the parameter of sGrs was showed a high predictive value for OSAS diagnos, defined as the Apnea–Hypopnea Index (AHI) ≥15 [11]. These results suggest that the caliber of both pharyngeal airway and intrathoracic airways in obese OSA patients are commonly prone to collapse on the exhale, due primarily to decrease in lung volumes. The cross-sectional area of pharyngeal airway as well as peripheral airways are well known to varies considerably with alterations in lung volume. The lumen size in those structures are proved to a decrease when the end-expiratory lung volume (EELV) or FRC are artificially lowered, that either caused by negative expiratory pressure (NEP) or by positive extrathoracic pressure on exhalation, and manifests as expiratory flow limitation (EFL) in the both structures as well as airflow resistance are markedly increased [12‐14]. These phenomena appear to be more pronounced in obese OSAS patients while adopt on the supine position. It has been demonstrated in obese subjects with and without OSA that lung volume, in term of EELV or FRC, would be a further decrease when the supine position is take on from a seated. Such a lung volume decrease may facilitate pharyngeal airway and intrathoracic airways to collapse or even closure, due to loss effects of caudal traction tension on both structures, and contribute to resistance increase in the airways.

Recently, we identified some parameters of impulse oscillometry (IOS) measured in obese OSA patients with sitting position, such as respiratory impedance (Zrs), Rrs, and Xrs at 5 Hz were significantly correlated with OSA severity as defined by AHI [15]. IOS, a type of FOT is used to determine the mechanical properties of the respiratory system during tidal breathing in recent years [16]. It is easy to perform, requires minimal subject cooperation, and the measurements can be recorded in any body position. More than twice the number of respiratory events in majority of the obese OSA population are occur in the supine position, [17, 18] the underlying mechanisms of it has been attributed to decrease in lung volume with a reduce favorable effects of caudal tracheal traction on pharyngeal airway [19]. It can be speculated from such evidence that the mechanical properties of the intrathoracic airways would also be implicated by the same mechanisms mentioned above when obese OSA patients sleeping in the supine. Hence, we hypothesize that parameters of IOS obtained in the supine could higher the capacity of IOS test in the seated position to differentiate between obese snoring persons and those more severely affected – namely, patients with OSAS.

The aim of this study was to develop a diagnostic tool, not requiring polysomnography, for identifying OSAS in snoring patients whose risk of OSAS is not too high. We have therefore undertaken a large prospective study to determine the optimal operating characteristics for the use of IOS as a screening tool for detecting OSAS in preobese and obese snoring patients.

In total, 230 patients who were preobese had a BMI 25.1 − 29.9 kg/m² and obese had a BMI ⩾30 kg/m² with normal lung function defined by FEV₁/FVC >80 % of the predicted values were included in the study, while 16 patients were not included in analyses due to an absence of IOS data in the supine position. Anthropometric characteristics, lung function and volume data, and PSG data of all patients are summarized in Table 1. Of those 84 subjects were identified as healthy snorers whose RDI less than 5 events/h according to PSG results, 37 subjects with a RDI of 5–14 events/h were identified as mild–OSA, and 93 subjects with a RDI ⩾15 events/h were identified as OSAS patients. No significant differences were found between the three groups with respect to age, height, sex, FEV₁/FVC, and lung volume data. There were significant differences in weight and BMI were existed only between healthy snorers and OSAS patients. The mechanical properties of the respiratory system measured by IOS for snorers, mild–OSA and OSAS patients in the sitting and the supine positions, are shown in Fig. 1a–f. Rrs0 was back-extrapolated separately, for each subject in each different condition, by applying linear regression equations of resistance vs frequency over the range of 5–15 Hz (Fig. 2a, b). The Rrs0 in the supine position were significantly higher with increasing severity of OSA as defined by RDI

In this study, we systematically appraised the ability of IOS parameters, obtained in supine and seated positions, compared with full polysomnography, for the detection of OSAS in preobese and obese snorers.

The parameter Rrs0, derived from linear-regression analysis of resistance versus frequency (over 5–15 Hz; Fig. 2a, b), its inverse Grs, and the IOS parameters at 5 Hz in the supine position, showed potential usefulness. We demonstrated that peripheral-airway parameter measurements by IOS in the sitting position were significantly and moderately correlated with the severity of OSAS as defined by RDI (Table 2). These associations were stronger when the IOS parameters were obtained in the supine position, and a highly significant association was found between X5 and RDI (r = −0.517, P = 0.000). The correlation coefficient between RDI and the peripheral-airway resistance parameters (R5), total respiratory resistance (Rrs0 and Zrs5), and respiratory conductance (Grs and Gz) were 0.378, 0.425, 0.395, −0.425, and −0.395, respectively, which are similar to results reported by Cao et al. in a previous study [22].

Shifting the posture from sitting to supine has been demonstrated to increase Rrs on FOT [23, 24]. This postural influence on respiratory mechanics has been attributed mainly to the effects of a reduction in lung volume [23‐25]. Lung volume, in terms of FRC and ERV, significantly decreases when an obese subject adopts the supine posture due to a decreased outward recoil from the chest wall mass loading, as well as the gravitational effects of the abdominal contents. This results in a relaxed diaphragm that takes up a more expiratory position [26‐28]. Because obese and obese OSAS patients already have a reduced FRC and ERV in erect and sitting postures, [15, 26] any further fall in FRC might lead to considerable airway closure or collapse. This could markedly increase airflow resistance in the intrathoracic airway, and may be more pronounced in the peripheral airways due to the lack of supporting structures such as the larger central airways. In the present study, there was a small decline in Rrs with increasing oscillation frequency between 5–15 Hz in the sitting position, which was exaggerated when our OSAS patients adopted the supine position. A striking rise in Rrs at 5 Hz was observed, exhibiting a more negative frequency dependence. In the non-OSAS groups (snorers and mild–OSA) there was little variation of Rrs with frequency and the Rrs at all frequencies was significantly lower than in the OSAS group in either the sitting or the supine position. Indeed, Rrs is roughly constant at lower-frequencies (from 4 to 16 Hz) on FOT in subjects with normal lungs, but shows a marked negative frequency dependence in patients with upper-airway obstruction and when the EELV was artificially lowered [14, 24]. Rrs at lower-frequencies in our OSAS patients presented a more negative frequency dependence than the non-OSAS groups. It was evident that most OSAS patients were experiencing low lung volumes, which may further decrease when those patients lie down, thereby contributing to a rise in Rrs at the serial site of the intrathoracic airways.

Other investigators have demonstrated that pharyngeal-airway collapsibility varies inversely with EELV. Reductions in EELV, by means of NEP or positive extrathoracic pressure on exhalation, causes decreases in caudal traction on the upper-airway and concomitant increases in upper-airway collapsibility, [29] presenting either with EFL or with airflow resistance increase in the upper-airway [6, 7, 30]. This is more pronounced in obese OSAS patients. The evidence implies that highly positive intrathoracic pressure, or driving pressure, would be generated when these patients begin expiration. Any additional pressure caused by NEP or positive extrathoracic pressure that favored the intrathoracic pressure or driving pressure to further evaluation, this may facilitate the collapse of the upper-airway as well as the early closure of small airways. Therefore, EFL induced by NEP is common in obese OSAS patients and can predict the severity of OSAS, as a high association between the degree of EFL measured in the supine position and AHI has been found [6, 7]. We found a remarkable significant increase in total respiratory resistance (Rrs0 and Zrs5) and a decrease in its inverse-respiratory conductance with increasing RDI in the OSAS group when they adopted a supine position. They also had a greater BMI than the non-OSAS groups. This may add evidence for the potential impact of obesity and postural changes on upper-airway patency and OSA susceptibility. Our results suggest that obesity, which substantially lowers EELV, can increase OSA susceptibility markedly by increasing pharyngeal collapsibility. This effect might be most marked when obese subjects sleep in the supine position, when EELV is the lowest [31].

Indeed, obesity has been proven to be the most common risk factor for OSAS. Obese supine subjects usually respire with a lower compliance of the lung and chest wall [32, 33]. Breathing with such a respiratory system determines larger changes in intrathoracic pressure, and even generates positive intrathoracic pressure surrounding the lung while relaxed at FRC level. Concomitantly, the high pleural pressure would cause tidal breathing to be initiated from low EELV where the lungs are less compliant and airways are prone to close on exhalation [34]. Breathing at lower end-expiration volumes has been demonstrated to significantly influence upper-airway patency, through loss of tension of caudal traction in the pharyngeal-airway, making it more folded and leading to increased pharyngeal collapsibility. In this study, the reactance (Xrs) at all oscillatory-frequencies measured in the supine position was found to be significantly decreased in the obese OSAS group. This decrease was remarkable for Xrs measured at 5 Hz (X5), and was found to strongly correlate with the severity of OSAS as defined by RDI and shows good predictive value for the diagnosis of OSAS in obese snorers. The X5 is a component of the out-of-phase air flow and pressure signal, and is numerically a negative value that reflects the sum elastance or compliance of the respiratory system. X5 values that are more negative indicate reduced respiratory system compliance or increased lung elasticity recoil pressure [14‐16]. Our findings thus lead us to speculate that a narrowing of the upper-airway due to obesity causes inspiratory resistive breathing, and leads to respiratory compliance reduction in OSAS patients. This, in turn, may aggravate the upper-airway as well as peripheral-airway collapsibility, due to a loss of caudal traction tension on those structures at the end of expiration when the lung volume is lowest.

The ROC curves we constructed for Rrs0, Zrs5, R5, Grs, Gz, and X5 in both positions showed how these parameters can discriminate OSAS in preobese and obese snorers who are suspected of having this condition, with the diagnostic accuracy estimated by the AUC (Fig. 3). Only the parameters obtained in the supine position have an acceptable discriminating power making them suitable for clinical use. OSAS patients were optimally detected using the cut-off points of −0.23 kPa.s.L⁻¹ for X5 with a good sensitivity 77.7 % and specificity of 79. 4 % and with a highly diagnostically accurate of 0.811 (95 % CI: 0.682–0.813). Among the predictors, X5 showed good PPV (76.9 %) and NPV (80 %) for the polysomnography diagnosis of OSAS.

A number of studies have used different and appealing approaches employing clinical and FOT parameters, or the degree of EFL, to predict the presence or absence of OSAS in suspected patients. Lorino et al. [30] reported that the FOT parameters of Rrs0 and Xrs at low-frequencies in the seated position were independently associated with the severity of OSAS. The ROC curves for those parameters were demonstrated to correctly discriminate patients with OSAS with sensitivities and specificities for Rrs0 and Xrs of 67 % and 93 %, and 90 % and 67 %, respectively. Zerah et al. [11] introduced that the parameter of sGrs in the seated position, combined with daytime oxygen-saturation, shows a 100 % NPV and a 86 % PPV with 100 % sensitivity and 84 % specificity for the diagnosis of OSAS in a large cohort of obese snorers. This suggests that obstruction in both structural upper and peripheral airways is common in obese OSAS patients. Previously, many studies have applied a small additional NEP to obese OSA patients in the sitting and supine, to induce EFL; these studies showed that the degree of EFL is a good predictor of this disorder [6, 7, 35‐38]. The NEP technique is initially used for early detection of intrathoracic EFL in patients with obstructive lung disease who usually have lower lung compliance. The increased intrathoracic pressure gradients on exhalation caused by NEP facilitate the early closure of peripheral airways and result in expiratory flow decreased. EFL has been demonstrated to be common in obese OSAS patients when they experience NEP in the supine position, although the source of an EFL (intrathoracic or extrathoracic) from NEP has not been determined [35]. Our results are in line with Liistro et al. [35] and Verin et al. [6] who suggested that obese OSAS patients may undergo pharyngeal-airway and peripheral-airway collapse when sleeping supine, due to breathing at lower end-expiration volumes with less respiratory compliance and loss of caudal traction tension on those structures. In recent years, interest in the interpretation of Xrs parameters in obstructive disease has increased. Of particular clinical interest is Xrs measured at 5 Hz as a surrogate marker of EFL with NEP [39‐42]. In the present study, we also used a statistical model based on lung function parameters in the supine position and combined this with anthropometry data to predict OSAS in preobese and obese snorers. This demonstrated that X5 combined with sex and ERV yields a good specificity (83.3 %) and sensitivity (76.8 %). Moreover, other predictors (Rrs0, Zrs5, and R5) combined with sex and BMI yielded acceptable sensitivities, but had lower specificity than X5 for predicting OSAS.

In conclusion, the results of this study demonstrate that respiratory resistance parameters (Rrs0, Zrs5) and their inverses describing respiratory conductance (Grs and Gz), and the reactance parameter X5, measured in the supine position measured by IOS significantly correlate with the severity of OSAS as defined by RDI. Both the ROC curves we constructed and a statistical model based on daytime respiratory function abnormalities were effective for predicting OSAS in preobese and obese snorers who were suspected of having this condition. This study appears to indicate that the upper-airway, as well as the peripheral-airway, may experience obstruction or collapse simultaneously when OSAS patients are in the supine position as obesity increases the elastic load on these structures. Thus, reactance (X5) measurements are related to the degree of airflow obstruction, which had a moderate sensitivity and specificity for detecting OSAS. Therefore, IOS measurements might be useful as a screening test for OSAS during wakefulness.