Airway cell involvement in intermittent hypoxia-induced airway inflammation.

PURPOSE: Respiratory inflammation has been described in patients with obstructive sleep apnea syndrome, but it is unknown whether the increased neutrophil and interleukin (IL)-8 levels observed in induced sputum reflect systemic or local airway inflammation. We assessed the potential role of resident cells in intermittent hypoxia-induced airway inflammation. METHODS: Airway epithelial cells (AEC) and bronchial smooth muscle cells (BSMC) were exposed to intermittent hypoxia (IH) in vitro. Cell supernatants were assessed for matrix metalloproteinase, growth factor, and cytokine expression. The role of IH on neutrophil and BSMC migration capacities was evaluated, and the effect of supernatants from IH-exposed or control AEC was tested. RESULTS: Compared to normoxic conditions, 24 h of exposure to IH induced a significant increase of MMP-9 and MMP-2 expression and pro-MMP-9 activation (p < 0.05), and IL-8 (p < 0.05), platelet-derived growth factor (PDGF)-AA (p < 0.05), and vascular endothelial growth factor (VEGF) (p < 0.05) expression by AEC and VEGF expression (p = 0.04) by BSMC. Neutrophil chemotaxis and BSMC migration were enhanced by IH and supernatants of IH-exposed AEC (112.00 ± 4.80 versus 0.69 ± 0.43 %, p = 0.0053 and 247 ± 76 versus 21 ± 23, p = 0.009 respectively). This enhanced BSMC migration was totally abolished in the presence of an antibody blocking PDGF-AA. CONCLUSIONS: These data suggest a specific inflammatory response of airway cells to IH, independently of systemic events.

Effect of manufacturer-inserted mask leaks on ventilator performance.

Most pressure-support devices use a single circuit with an exhalation port integrated in the mask. The aim of the current study was to compare the effects of masks having different manufacturer-inserted leaks on ventilator performance. We simulated chronic obstructive pulmonary disease and restrictive disease. Four ventilators (VENTImotion (Weinmann, Hamburg, Germany), VPAP III STA (ResMed, Saint Priest, France), Synchrony 2 (Respironics, Nantes, France) and Vivo 40 (Breas, Saint Priest)) were tested with the recommended masks and with the masks having the largest and smallest leaks. Tests were performed with pressure support levels of 10, 15 and 20 cmH(2)O. The in vivo evaluation compared two ventilators using recommended masks opposed in terms of exhaled port resistance. The ventilators were tested with their recommended mask, and after mask exchange. The mask with the largest leak induced auto-triggering and/or increased inspiratory-trigger sensitivity was the VENTImotion under both simulated conditions and VPAP III STA under the simulated obstructive-disease condition. The mask with the smallest leak-increased inspiratory-trigger delay was Synchrony 2 in the simulated obstructive-disease condition and increased rebreathing. The in vivo study confirmed the bench results. When switching to a mask that has a different leak, evaluation is needed to adjust trigger sensitivity and pressurisation level and to check the absence of rebreathing.

Prestress and adhesion site dynamics control cell sensitivity to extracellular stiffness.

This study aims at improving the understanding of mechanisms responsible for cell sensitivity to extracellular environment. We explain how substrate mechanical properties can modulate the force regulation of cell sensitive elements primarily adhesion sites. We present a theoretical and experimental comparison between two radically different approaches of the force regulation of adhesion sites that depends on their either stationary or dynamic behavior. The most classical stationary model fails to predict cell sensitivity to substrate stiffness whereas the dynamic model predicts extracellular stiffness dependence. This is due to a time dependent reaction force in response to actomyosin traction force exerted on cell sensitive elements. We purposely used two cellular models, i.e., alveolar epithelial cells and alveolar macrophages exhibiting respectively stationary and dynamic adhesion sites, and compared their sensitivity to theoretical predictions. Mechanical and structural results show that alveolar epithelial cells exhibit significant prestress supported by evident stress fibers and lacks sensitivity to substrate stiffness. On the other hand, alveolar macrophages exhibit low prestress and exhibit sensitivity to substrate stiffness. Altogether, theory and experiments consistently show that adhesion site dynamics and cytoskeleton prestress control cell sensitivity to extracellular environment with an optimal sensitivity expected in the intermediate range.

Performance of ventilators for noninvasive positive-pressure ventilation in children.

The aim of the present study was to evaluate the performance characteristics of all the ventilators proposed for home noninvasive positive-pressure ventilation in children in France. The ventilators (one volume-targeted, 12 pressure-targeted and four dual) were evaluated on a bench which simulated six different paediatric ventilatory patterns. For each ventilator, the quality of the inspiratory and expiratory trigger and the ability to reach and maintain the preset pressures and volumes were evaluated with the six patient profiles. The performance of the ventilators showed great variability, and depended upon the type of trigger (flow or pressure), type of circuit and patient profile. Differences were observed between the preset and measured airway pressure and between the tidal volume measured by the ventilator and on the bench. Leaks were associated with an inability to detect the patient's inspiratory effort or autotriggering. No single ventilator was able to adequately ventilate the six paediatric profiles. Only a few ventilators were able to ventilate the profiles simulating the youngest patients. A systematic paediatric bench evaluation is recommended for every ventilator proposed for home ventilation, in order to detect any dysfunction and guide the choice of the appropriate ventilator for a specific patient.

Mechanical and structural assessment of cortical and deep cytoskeleton reveals substrate-dependent alveolar macrophage remodeling.

The sensitivity of alveolar macrophages to substrate properties has been described in a recent paper (Féréol et al., Cell Motil. Cytoskel. 63 (2006), 321-340). It is presently re-analyzed in terms of F-actin structure (assessed from 3D-reconstructions in fixed cells) and mechanical properties (assessed by Magnetic Twisting Cytometry experiments in living cells) of cortical and deep cytoskeleton structures for rigid plastic (Young Modulus: 3 MPa) or glass (70 MPa) substrates and a soft (approximately 0.1 kPa) confluent monolayer of alveolar epithelial cells. The cortical cytoskeleton component (lowest F-actin density) is represented by the rapid and softer viscoelastic compartment while the deep cytoskeleton component (intermediate F-actin density) is represented by the slow and stiffer compartment. Stiffness of both cortical and deep cytoskeleton is significantly decreased when soft confluent monolayer of alveolar epithelial cells replace the rigid plastic substrate while F-actin reconstructions reveal a consistent actin cytoskeleton remodeling observable on both cytoskeleton components.

Evaluation of carbon dioxide rebreathing during pressure support ventilation with airway management system (BiPAP) devices.

The purpose of this study was to evaluate whether carbon dioxide (CO2) rebreathing occurs in acute respiratory failure patients ventilated using the standard airway management system (BiPAP pressure support ventilator; Respironics; Murrysville, Pa) with positive inspiratory airway pressure and a minimal level of positive end-expiratory pressure (PEEP) and whether any CO2 rebreathing may be efficiently prevented by the addition of a nonrebreathing valve to the BiPAP system circuit. In the first part of the study, the standard device was tested on a lung model with a nonrebreathing valve (BiPAP-NRV) and with the usual Whisper Swivel connector (BiPAP-uc). With the BiPAP-uc device, the resident volume of expired air in the inspiratory circuit at the end of expiration (RVEA) was 55% of the tidal volume (VT) when the inspiratory pressure was 10 cm H2O and the frequency was at 15 cycles per minute. The BiPAP-NRV device efficiently prevented CO2 rebreathing but resulted in a slight decrease in VT, which was due to a significant increase in external PEEP (2.4 vs 1.3 cm H2O) caused by the additional expiratory valve resistance. For similar reasons, both the pressure swing necessary to trigger pressure support and the imposed expiratory work were increased in the lung model when the nonrebreathing valve was used. In the second part of the study, seven patients weaned from mechanical ventilation were investigated using a randomized crossover design to compare three situations: pressure support ventilation with a conventional intensive care ventilator (CIPS), BiPAP system use, and BiPAP-NRV. When we compared the BiPAP system use with the other two systems, we observed no significant effect on blood gases but found significant increases in VT, minute ventilation, and work of breathing. These findings are experimental and are clinical evidence that significant CO2 rebreathing occurs with the standard BiPAP system. This drawback can be overcome by using a non-rebreathing valve, but only at the expense of greater expiratory resistance.

Assessment of high-frequency neonatal ventilator performances.

OBJECTIVE: To assess the efficacy and reliability of neonatal high-frequency ventilators. DESIGN: Bench evaluation of neonatal high-frequency ventilators. SETTING: Physiology department and university hospital neonatal intensive care unit. INTERVENTIONS: HFV-Babylog 8000 (Dräger Medical), OHF 1 (Dufour), and SensorMedics 3100A (Sensor-Medics) ventilators were connected to a neonatal test-lung. Tidal volume, peak-to-peak pressure amplitude, and mean airway pressure were measured for several ventilator settings, endotracheal tube sizes, and lung compliances. MEASUREMENTS AND RESULTS: Increasing peak-to-peak pressure resulted in a linear increase in tidal volume delivery in the 0-30% range of maximum amplitude. No significant increase in tidal volume was observed with the HFV-Babylog 8000 when pressure amplitude was above 50%. The maximum tidal volume delivered was substantially smaller with the HFV-Babylog 8000 than with the OHF 1 or SensorMedics 3100A. Tidal volume increased with endotracheal tube size with all three ventilators. Increasing test-lung compliance resulted in lower tidal volumes only with OHF 1. Decreasing mean airway pressure was responsible for a decrease in tidal volume delivery with HFV-Baby-log 8000. CONCLUSION: We found that under our test conditions two of the three ventilators delivered adequate tidal volumes at the usual frequency of 15 Hz, regardless of the size of the endotracheal tube and of the mechanical properties of the respiratory system. When lung compliance increased or mean airway pressure decreased, both of which are common events during the recovery phase of hyaline membrane disease, we found that the intrinsic properties of two of the ventilators tested were responsible for a decrease in tidal volume. This decrease may account for some cases of heretofore unexplained hypercapnia.

Interaction of oscillatory and steady turbulent flows in airway tubes during impedance measurement.

Measurement of input respiratory impedance is carried out by superimposing forced oscillations on spontaneous breathing. The latter thus acts as a quasi-steady unidirectional flow component, with effects on the measured impedance that are habitually neglected (linearity assumption). We examined the validity of that assumption in the case of a turbulent steady flow. We tested the validity of a fluid dynamics criterion previously proposed in water channel experiments for gas flow in a tube. This criterion states that oscillatory and continuous turbulent flow may or may not interact if the Stokes boundary layer (ls) is embedded within the viscous sublayer (lv), i.e., if lS+ = lS/lv < or = 10, implying Re7/8 < or = (100 alpha/square root of 2), for a fully developed hydraulically smooth turbulent flow in a tube (where alpha is Womersley parameter and Re is Reynolds number of the steady-flow component). Experiments were performed in long rigid circular and semicircular tubes by superimposing two independent well-defined flows: 1) laminar oscillatory flow obeying the linear transmission line model (frequency = 1.5-250 Hz, i.e., alpha = 6-80) and 2) fully developed turbulent flow characterized by Blasius resistance formula (Re = 3,000-16,000). Confirming the validity of the criterion above, we found that the real and the imaginary parts of the long-tube impedance did not differ from those measured in the absence of a steady-flow component, provided lS+ < or = 10. On the contrary, the real parts measured with and without the continuous component differed greatly as soon as lS+ > 10, both for circular and semicircular tubes and for outward as well as inward steady flows. We concluded that the proposed criterion is pertinent for predicting appropriate oscillation frequency for a given rate of spontaneous flow, such that oscillatory and turbulent flows do not interact. Application of the forced oscillation measurement technique during spontaneous breathing requires use of a range of oscillatory frequencies higher than the frequency range classically used during apnea.

Effect of air entrainment on airway pressure during endotracheal gas injection.

Turbulent jets in endotracheal tubes induce air entrainment and airway pressure changes. We attempted to understand the physical explanation for these effects, which open up to a wide range of applications in intubated patients. An in vitro study was performed on standard size endotracheal tubes with diameters of 8, 7, and 3 mm and several capillaries molded into the wall (less than 1 mm diam) allowing gas injection at approximately 1-2 cm from the tracheal end of the endotracheal tube. This produced a jet velocity-dependent gain in tracheal pressure (Ptr) during inspiration. Data have been interpreted with a theory, based on the classic momentum theorem, which indicates that the mechanisms involved resemble those of axisymmetrical confined jets: air entrainment by turbulent friction with a longitudinal increase in lateral pressure. The difference with axisymmetrical systems lies in the nonconservation of the total thrust in our system because, secondary to wall friction and to the nonaxial incidence of the jets, only a fraction of the jet momentum flux is transformed into pressure. This suggests faster mixing in the present lateral jet system, as shown by 1) the independence of Ptr on tracheal geometry and 2) the very rapid increase in lateral pressure. The present study supports the idea that pressure changes in the airways, which are potentially beneficial in intubated patients, can be satisfactorily generated by turbulent jets.

Respiratory response to inhaled CO2 during positive inspiratory pressure in humans.

To investigate ventilatory CO2 sensitivity during inspiratory pressure support (IPS), we administered inspiratory CO2 [fractional concn (FICO2) 0.01, 0.03, or 0.05] in eight normal subjects without (CTRL) or with (Pinsp) positive inspiratory airway pressure (5 or 10 cmH2O). At CTRL and low IPS, CO2 inhalation led to a significant increase in tidal volume (VT) with nearly identical slopes in the plot of VT vs. end-tidal PCO2. At the high IPS level, VT at FICO2 of 0 was significantly above the value at lower Pinsp and did not increase with CO2 unless FICO2 was elevated to > 0.03. There was very little effect of either Pinsp or FICO2 on respiratory frequency and respiratory timing. The data suggest that the CO2 sensitivity of ventilation is similar at low levels of IPS as during CTRL. However, at high levels of IPS, VT is determined largely by the passive inflation and, thus, independent of CO2. CO2 has to be elevated to increase the respiratory drive before VT becomes CO2 sensitive.

Shunt effect of gas compression inside pneumotachographs during forced oscillations.

Determination of the frequency response of pneumotachographs is needed whenever they are used to measure high-frequency flows, such as in the forced oscillation method. When screen and capillary pneumotachographs are calibrated using an adiabatic compression in a closed box as a reference impedance, they can be adequately described by a series of inertial-resistive elements. However, this type of reference impedance strongly differs from the actual respiratory impedance (ZL). We studied the frequency response of pneumotachographs up to 250 Hz in reference to the impedance of a compressible gas oscillating in a long tube, taken as a more generalizable model of actual ZL. We found that, with this device, the series resistance-inertance models fail to describe the frequency response of the pneumotachograph. However, when compressible effects in the pneumotachograph are taken into account by adding to the resistive models a compliance (Cpn) corresponding to the compression in half of the inner volume of the pneumotachograph, the agreement with experiments becomes satisfactory. Gas compression-related phenomena were demonstrated to be negligible only when the parameter omega Cpn magnitude of ZL is much smaller than 1 (omega pulsation). Results obtained in normal humans have shown that such a correction is required above 100 Hz. Similar correction at lower frequency might also be necessary in cases of large respiratory impedance (e.g., babies, subjects with pathological lungs, and intubated subjects).

Dependence of central airway resistance on frequency and tidal volume: a model study.

The resistance of a hollow cast of human central airways was measured during true sinusoidal airflow oscillations over a wide range of frequencies (0.5-40 Hz) and for various flow amplitudes up to 8 l/s. Pressure and flow were measured in the trachea with high-performance transducers, digitized and averaged over 100 cycles. Data were studied at two points in the flow cycle: at peak inspiratory and expiratory flows and in the two neighborhoods around zero flow where airway resistance (Rv approximately equal to o) was taken as the average slope of the pressure-flow (P-V) curve in each zone. When data obtained near peak flow were plotted in terms of dimensionless pressure drop vs. peak Reynolds number (Rem) and compared with steady-state data, we found no difference up to 2 Hz as previously reported (Isabey and Chang, J. Appl. Physiol. 51: 1338-1348, 1981), a slight decay in pressure drop between 4 and 8 Hz, a frequency-dependent increase in peak flow resistance at high frequencies (10-40 Hz) governed by the Strouhal number alpha 2/Rem beyond alpha 2/Rem = 0.5. On the other hand RV approximately equal to o was found to increase relative to steady state as local acceleration increases, e.g., as peak flow increases at a fixed frequency; this differs from the classical linear theory of oscillatory flow in a long straight tube. To explain these results, we had to use, as in our previous study, an alternative expression for the Strouhal number, i.e., epsilon = L X A X (dV/dt)/V2 (where L and A are the length and cross-sectional area of the trachea and V is a constant flow range over which resistance around flow reversal was computed), which accurately reflects the ratio of local acceleration [d(V/A)/dt)] to convective acceleration [(V/A)2/L] in developing branching flow. Finally, to delineate the regions of dominance of each of the dimensionless parameters, we compiled frequency-tidal volume diagrams for peak flows as well as for reversal. Epsilon, which is negligible near peak flows, appeared to govern the oscillatory P-V relationship near flow reversal in a transitional region of the diagram located between regions of steadiness, or moderate unsteadiness, and a region of dominant unsteadiness governed by alpha.

Physiological effects of alveolar, tracheal, and "standard" pressure supports.

Pressure support (PS) is characterized by a pressure plateau, which is usually generated at the ventilator level (PS(vent)). We have built a PS device in which the pressure plateau can be obtained at the upper airway level (PS(aw)) or at the alveolar level (PS(A)). The effect of these different PS modes was evaluated in seven healthy men during air breathing and 5% CO(2) breathing. Minute ventilation during air breathing was higher with PS(A) than with PS(aw) and lower with PS(vent) (16 +/- 3, 14 +/- 3, and 11 +/- 2 l/min, respectively). By contrast, there were no significant differences in minute ventilation during 5% CO(2) breathing (25 +/- 5, 27 +/- 7, and 23 +/- 5 l/min, respectively). The esophageal pressure-time product per minute was lower with PS(A) than with PS(aw) and PS(vent) during air breathing (29 +/- 26, 44 +/- 44, and 48 +/- 30 cmH(2)O. s, respectively) and 5% CO(2) breathing (97 +/- 40, 145 +/- 62, and 220 +/- 41 cmH(2)O. s, respectively). In conclusion, during PS, moving the inspiratory pressure plateau from the ventilator to the alveolar level reduces pressure output, particularly at high ventilation levels.

Nonchemical influence of inspiratory pressure support on inspiratory activity in humans.

To determine whether nonchemical inhibition of respiratory activity occurs during inspiratory pressure support (IPS) ventilation (IPSV), respiratory motor output (in 9 subjects), obtained by calculating transdiaphragmatic pressure-time products, and central respiratory output (in 5 subjects), obtained by integrating the electromyographic activity of the diaphragm (EMGdi) during mechanical inspiratory time, EMGdi per minute, and electrical inspiratory time, as determined from onset to peak EMGdi, were compared during spontaneous ventilation (control) and IPSV with (IPS+CO2) and without (IPS) correction of hypocapnia. Both IPS and IPS+CO2 induced significant decreases in transdiaphragmatic pressure-time products (46 +/- 31 and 53 +/- 23%, respectively), EMGdi during mechanical inspiratory time (49 +/- 12 and 57 +/- 14%, respectively), EMGdi per minute (65 +/- 22 and 69 +/- 15%, respectively), and electrical inspiratory time (73 +/- 8 and 65 +/- 6%, respectively). Because correction of hypocapnia failed to eliminate the marked inhibition of both respiratory and central motor output seen with IPS, we conclude that nonchemical inhibition of respiratory activity occurs during IPSV.

Gravity effects on upper airway area and lung volumes during parabolic flight.

We measured upper airway caliber and lung volumes in six normal subjects in the sitting and supine positions during 20-s periods in normogravity, hypergravity [1.8 + head-to-foot acceleration (Gz)], and microgravity ( approximately 0 Gz) induced by parabolic flights. Airway caliber and lung volumes were inferred by the acoustic reflection method and inductance plethysmography, respectively. In subjects in the sitting position, an increase in gravity from 0 to 1. 8 +Gz was associated with increases in the calibers of the retrobasitongue and palatopharyngeal regions (+20 and +30%, respectively) and with a concomitant 0.5-liter increase in end-expiratory lung volume (functional residual capacity, FRC). In subjects in the supine position, no changes in the areas of these regions were observed, despite significant decreases in FRC from microgravity to normogravity (-0.6 liter) and from microgravity to hypergravity (-0.5 liter). Laryngeal narrowing also occurred in both positions (about -15%) when gravity increased from 0 to 1.8 +Gz. We concluded that variation in lung volume is insufficient to explain all upper airway caliber variation but that direct gravity effects on tissues surrounding the upper airway should be taken into account.

Dual assessment of airway area profile and respiratory input impedance from a single transient wave.

This report concerns the inference of geometric and mechanical airway characteristics based on information derived from a single transient planar wave recorded at the airway opening. We describe a new method to simultaneously measure upper airway area and respiratory input impedance by performing dual analysis of a single pressure wave. The algorithms required to reconstruct airway dimensions and mechanical characteristics were developed, implemented, and tested with reference to known physical models. Our method appears suitable to estimate, even under severe intensive care unit conditions, the respiratory system frequency response (above 10 Hz) in intubated patients and the patency of the endotracheal tube used to connect the patients to the ventilator.

Segmental analysis of nasal cavity compliance by acoustic rhinometry.

To explore the determinants of possible collapse of the nasal valve region, a common cause of nasal obstruction, we evaluated the mechanical properties of the nasal wall. In this study, we determined the nasal cross-sectional area-to-negative pressure ratio (nasal wall compliance) in the anterior part of the nose in six healthy subjects by measuring nasal area by acoustic rhinometry at pressures ranging from atmospheric pressure to a negative pressure of -10 cmH(2)O. Measurements were performed at baseline and after nasal mucosal decongestion (oxymetazoline). At baseline, nasal wall compliance increased progressively from the nasal valve (0.031 +/- 0.016 cm2/cmH(2)O, mean +/- SD) to the anterior and medial part of the inferior turbinate (0.045 +/- 0.024 cm2/cmH(2)O) and to the middle meatus region (0.056 +/- 0.029 cm2/cmH(2)O). After decongestant, compliances decreased and became similar in the three regions. On the basis of these results, we hypothesize that compliance of the nasal wall is partly related to mucosal blood volume and quantity of vascular tissue, which differ in the three regions, increasing from the nasal valve to the middle meatus.

Estimation of inspiratory pressure drop in neonatal and pediatric endotracheal tubes.

Endotracheal tubes (ETTs) constitute a resistive extra load for intubated patients. The ETT pressure drop (DeltaP(ETT)) is usually described by empirical equations that are specific to one ETT only. Our laboratory previously showed that, in adult ETTs, DeltaP(ETT) is given by the Blasius formula (F. Lofaso, B. Louis, L. Brochard, A. Harf, and D. Isabey. Am. Rev. Respir. Dis. 146: 974-979, 1992). Here, we also propose a general formulation for neonatal and pediatric ETTs on the basis of adimensional analysis of the pressure-flow relationship. Pressure and flow were directly measured in seven ETTs (internal diameter: 2.5-7.0 mm). The measured pressure drop was compared with the predicted drop given by general laws for a curved tube. In neonatal ETTs (2.5-3.5 mm) the flow regime is laminar. The DeltaP(ETT) can be estimated by the Ito formula, which replaces Poiseuille's law for curved tubes. For pediatric ETTs (4.0-7.0 mm), DeltaP(ETT) depends on the following flow regime: for laminar flow, it must be calculated by the Ito formula, and for turbulent flow, by the Blasius formula. Both formulas allow for ETT geometry and gas properties.

Heavy snoring with upper airway resistance syndrome may induce intrinsic positive end-expiratory pressure.

We studied eight heavy snorers with upper airway resistance syndrome to investigate potential effects of sleep on expiratory airway and lung resistance, intrinsic positive end-expiratory pressure, hyperinflation, and elastic inspiratory work of breathing (WOB). Wakefulness and non-rapid-eye-movement sleep with high- and with low-resistance inspiratory effort (H-RIE and L-RIE, respectively) were compared. No differences in breathing pattern were seen across the three conditions. In contrast, we found increases in expiratory airway and lung resistance during H-RIE compared with L-RIE and wakefulness (56 +/- 24, 16 +/- 4, and 11 +/- 4 cmH2O . 1(-1) . s, respectively), with attendant increases in intrinsic positive end-expiratory pressure (5.4 +/- 1.8, 1.4 +/- 0.5, and 1.3 +/- 1.3 cmH2O, respectively) and elastic WOB (6.1 +/- 2.2, 3.7 +/- 1.2, and 3.4 +/- 0.7 J/min, respectively). The increase in WOB during H-RIE is partly caused by the effects of dynamic pulmonary hyperinflation produced by the increased expiratory resistance. Contrary to the Starling model, a multiple-element compliance model that takes into account the heterogeneity of the pharynx may explain flow limitation during expiration.