INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients suffering from the sleep apnea-hypopnea syndrome (SAHS) experience recurrent elevation of airway obstruction associated with partial or total collapse of the upper airway (1- 3). These obstructive events are a consequence of increased airway collapsibility (4). Indeed, in patients with SAHS the critical opening pressure (P_crit), which is the minimal intraluminal pressure required to maintain upper airway patency, may be increased to values higher than atmospheric pressure during sleep. Consequently, prevention of airway obstruction in these patients requires application of a nasal pressure (P_n) high enough to result in an intraluminal pressure greater that P_crit at all times. To this end, P_n should be at least equal to P_crit plus the pressure drop across the resistance of the segment of upper airway from the nostrils to the collapsible airway segment. As this dynamic pressure drop depends on the breathing flow, the pressure at the nasal mask that is just necessary to keep the airflow open varies along the breathing cycle.

The most widespread palliative treatment applied in SAHS, continuous positive airway pressure (CPAP [5, 6]) is not adapted to the fact that the P_n exactly required to maintain airway patency depends at any time on the breathing flow. To prevent airway obstruction, CPAP should be high enough to compensate for the maximal inspiratory dynamic pressure possible (at inspiratory peak flow) and, consequently, the applied P_n is much higher than necessary during most of the breathing cycle. Excessive P_n may unnecessarily worsen the patient's compliance and increase the consequences of high intrathoracic pressure (increased work of breathing due to hyperinflation and possible long-term hemodynamic effects). Bilevel positive airway pressure (7), which consists of applying a lower constant pressure during expiration than inspiration, is also employed to maintain airway patency in patients with SAHS (8). Although this mode of nasal pressure support takes into account the phase of breathing, bilevel pressure may be considered as a variant of CPAP since the applied P_n is not adapted to the continuously changing breathing airflow and is, therefore, not the pressure which is just required to maintain airway patency at any time. Moreover, the sudden changes in P_n between inspiration and expiration, which are characteristic of bilevel pressure, are transmitted to the collapsible airway with the risk of possible instability in the upper airway (9).

The aim of this work was to design and assess a procedure for applying a nasal pressure specifically adapted to maintain airway patency in SAHS. Accordingly, the applied P_n consisted of a constant term to account for the critical pressure of the collapsible upper airway and a time varying term, which was proportional to the instantaneous flow, to account for the dynamic pressure drop due to breathing. This flow-dependent positive airway pressure (FDPAP) was tested on a model study with realistic analogs of the collapsible upper airway, and its practical applicability was assessed on nine patients with SAHS during sleep.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Rationale

According to the Starling model of collapsible airway (Figure 1), which is commonly used (4) to interpret upper airway collapsibility in SAHS, a nasal static pressure P_n  P_crit would be enough to maintain the airway patency in the absence of flow. However, this is not the case when breathing because a positive pressure gradient is required to inspire through the resistance (R_n) of the upper airway segment extended from the nostrils to the collapsible section. Assuming linearity, the difference from nasal pressure to the intraluminal pressure (P) in the collapsible airway (Figure 1) for a given flow (;  > 0 during inspiration) is

(1)

View larger version (13K): [in this window] [in a new window]   Figure 1.   Respiratory model based on a Starling resistor for the collapsible upper airway. Rn = upstream (nasal) resistance; Ptr = tracheal pressure; Pcrit = critical pressure of the collapsible airway; P = intraluminal pressure in the collapsible airway; PS = constant pressure source to modify Pcrit.

and, hence, intraluminal pressure P is

(2)

As airway patency requires an intraluminal pressure always equal to or greater than the critical opening pressure (P  P_crit), it follows from Equation 2 that

(3)

and consequently, P_n should be at least equal to P_crit plus the dynamic pressure drop across the resistance of the upper airway segment R_n:

(4)

As illustrated in Figure 2 (left) no external P_n is required to maintain airway patency in healthy subjects. The figure simulates a breathing flow with normal frequency and amplitude (15 breaths/min, ± 0.5 L/s), and the corresponding pressure drop (P_n  P; Equation 1) across a typical nasal resistance R_n = 8 cm H₂O · s/L (10). In the normal situation with P_n = 0, intraluminal pressure (P; Equation 2) reaches a minimum value of 4 cm H₂O which is much higher than the typical critical pressure (about 13 cm H₂O [1, 11]) in healthy subjects. By contrast, artificial support by means of a positive P_n is required to prevent airway collapse and to allow normal breathing during sleep in patients with SAHS who exhibit a critical pressure higher than normals (12, 13). For instance, in the simulation of Figure 2 (left) a patient with P_crit = 6 cm H₂O would be unable to breathe owing to airway collapse (obstructive apnea) since at any time P < P_crit.

View larger version (24K): [in this window] [in a new window]   Figure 2.   Simulation of , P n, P, and dynamic pressure drop (Pn  P) from the nostrils to the collapsible airway. The critical pressure (6 cm H2O) is indicated by dashed lines.

When CPAP is applied, a nasal pressure:

(5)

where I_max is maximal inspiratory flow, ensures that at any time intraluminal pressure P  P_crit (Equation 4). In the simulation example of Figure 2 (center), CPAP = 10 cm H₂O would increase intraluminal P to a value always greater than the P_crit = 6 cm H₂O. Nevertheless, applying such a CPAP (Equation 5) is not an optimal pressure support mode since it does not provide the P_n that is exactly required for airway patency, but a value of P_n that is higher than necessary during most of the breathing cycle. Indeed, intraluminal pressure, which according to Equations 2 and 5, is in this case

(6)

is greater than P_crit always except at peak inspiratory flow (Figure 2, center).

The designed flow-dependent support mode (FDPAP) was aimed to continuously adapt P_n to the specific value required to avoid airway obstruction at any time of the breathing cycle. According to Equation 4, the ideal pressure support to maintain airway patency should consist of a static component to account for the upper airway critical pressure plus a flow-dependent dynamic component to account for the pressure drop from the nostrils and the collapsible segment (P_n  P_crit + R_n · ). Consequently, the FDPAP ventilator was designed to apply a nasal pressure consisting of a constant term and a flow proportional term

(7)

where P₀ and k are the parameters of the FDPAP ventilator that can be modified. Ideally, the optimal pressure would be applied for ventilator settings P₀ = P_crit and k = R_n since in this case the nasal pressure applied (Equation 7) would coincide with the one required (Equation 4) to allow P = P_crit, and therefore airway patency would be ensured at any time with minimal nasal pressure. However, the actual values of P_crit and R_n for a given patient are not known a priori and, consequently, application of the FDPAP mode of support requires an estimate of the values of these parameters. To this end, we designed the following procedure for setting P₀ and k. Initially, k was set to k = 0 (i.e., P_n = P₀ = CPAP) to determine the optimal CPAP (CPAP_opt), defined as the minimal P_n to avoid flow limitation, specifically at inspiratory peak flow. This would correspond to CPAP = 10 cm H₂O in the simulation of Figure 2 (center) because this is the minimum value of nasal pressure to ensure P  P_crit at any time of the breathing cycle. Once CPAP_opt was determined, we followed the algorithm indicated in Figure 3. From the starting point k = 0 and P₀ = CPAP_opt, P₀ was decreased by steps of 0.5 cm H₂O and k was increased at each step so as to obtain P_n = CPAP_opt at peak inspiration:

(8)

View larger version (22K): [in this window] [in a new window]   Figure 3.   Algorithm for setting the parameters k and P0 in the FDPAP support mode. (1) model study, (2) patient study.

This process of modifying P₀ and k was progressively repeated. The values of P₀ and k that were finally selected were the last settings for which no flow limitation of sleep events were observed, suggesting that P₀ was close to P_crit and k was close to R_n. Accordingly, the P_n applied is the one exactly required for keeping intraluminal pressure P at its minimal possible level (P = P_crit), as illustrated in the simulation of Figure 2 (right). This figure also shows that the mean nasal pressure with FDPAP (6 cm H₂O) is much lower than the one required to maintain airway patency with CPAP (CPAP_opt = 10 cm H₂O, Figure 2, center).

Setup

The ventilation system developed to generate FDPAP is schematically shown in Figure 4. It was based on a modified bilevel positive airway pressure (BiPAP) ventilator (Respironics, Inc., Murrysville, PA) with its conventional tubing and whisper swivel. A Fleisch-II pneumotachograph (Metabo, Epalinges, Switzerland) plus pressure transducer (MP-45, ± 2 cm H₂O; Validyne, Northridge, CA) and another pressure transducer (MP-45, ± 50 cm H₂O; Validyne) allowed the recording of and P_n, respectively. Signals and P_n were analogically low-pass filtered (Butterworth 8-poles, 32 Hz) and sampled at 256 Hz by a computer (486-type PC, 2831 Data Translation board, Marlboro, MA). The frequency responses of the pressure measuring systems were flat (± 2%) up to 20 Hz. The BiPAP device was modified to drive its valve coil (14) by means of an external feedback controller. The controller, which was fed with and P_n as input signals, consisted of the computer, including its A/D-D/A board and specific software, and a power amplifier. The system allowed us to continuously modify the settings of P₀ and k according to the described procedure (Figure 3) and to drive the valve coil with the voltage required to generate a nasal pressure P_n = P₀ + k ·  (Equation 7). A minimum P_n = 3 cm H₂O was allowed to avoid rebreathing.

View larger version (18K): [in this window] [in a new window]   Figure 4.   Setup for applying FDPAP. WS = whisper swivel; PNT = pneumotachograph; PT = pressure transducer.

Model Study

The model study was carried out on an analog of the respiratory system shown in Figure 1. A resistor (R_n) accounted for the resistance from the nose to the collapsible upper airway. R_n was followed by a collapsible segment with an external pressure (P_crit) representing the critical opening pressure of the collapsible upper airway (1, 4, 12). P_crit was imposed by a constant pressure source consisting of an independent CPAP device connected to a partially open shunt valve in parallel. The pump to generate the breathing flow () was a computer-controlled mechanical syringe (PWG; MH Custom Design & Mfg L.C., Midvale, UT). In the model P_n corresponded to nasal pressure, and pressure at the outlet of the collapsible segment (P_tr) corresponded to the pressure at the airway downstream of the collapsible upper airway, i.e., the trachea.

Six models (M₁-M₆) of the collapsible airway analog (Figure 1) with various values of R_n, P_crit, and collapsible tubes were built to test FDPAP under different mechanical conditions. In model M₁ a mesh-wire screen linear resistance R_n = 10.5 cm H₂O · s/L was connected to a collapsible segment made of 21 cm of latex Penrose tube of 19 mm interior diameter (ID) (Sherwood Medical, Tullamore, Ireland) attached to rigid cylindrical tubes (21 mm ID), subjected to a strain of 5% and enclosed in a small chamber (50 mm ID, 25 cm in length) where P_crit was set to 5 cm H₂O. The elastance of the tube (E_w) was virtually nil. The relative weights of the static (P_crit) and dynamic (R_n · ) pressure components were modified by building models M₂-M₄ with the linear resistors (R_n = K₁ + K₂ · ; with K₂ = 0) specified in Table 1. To investigate the role of a nonlinear upstream resistance, an orifice-type nonlinear element was included in model M₅ (Table 1). In order to implement a model with a low compliance collapsible segment, in model M₆ (Table 1) the collapsible resistor of models M₁-M₅ was replaced by a flat and thick rubber tube described in detail in (15) which was characterized by a high elastance and an almost nil section in the absence of transmural pressure.

For each of the models (M₁-M₆), we applied the described procedure (Figure 3) to determine the ventilator settings (P₀ and k) for a sinusoidal breathing with a frequency of 15 breaths/min and a minute volume (Vmin) of 11.25 L/min. The effect of an air leak at the mask level was assessed in model M₁ by placing a nonlinear leak resistance [K₁ = 7.3 cm H₂O · s/L, K₂ = 117.5 cm H₂O · (s/L)²] between the pneumotachograph and R_n while maintaining the settings previously fixed in the absence of leak. To analyze the response of the CPAP and FDPAP modes to an increased level of breathing, the ventilator settings determined previously were maintained and pump ventilation was increased by 33% from nominal Vmin = 11.25 to 15 L/min. In addition to CPAP and FDPAP, model M₁ was also subjected to two settings of bilevel pressure: inspiratory pressure equal to CPAP_opt and expiratory pressure of 8 and 4 cm H₂O below inspiratory pressure.

Assessment of the performance of the different ventilation modes was carried out by: (1) computing the mean P_n over an entire breathing cycle (P_n,mean); (2) analyzing the flow curve pattern () to detect the presence of flow limitation; and (3) computing the inspiratory pressure swings (P_tr) and the work of breathing (W =  P_tr · dV) done by the simulated patient (pump) on the upper airway and the ventilator. P_tr was measured and recorded with equipment identical to that described for P_n. W was computed as the integral W =  P_tr ·  · dt over an entire breathing cycle ( · dt = dV) divided by the tidal volume to obtain the work per liter of ventilation.

Patient Study

Nine male patients with severe SAHS, previously documented by full polysomnography, were included in the study (Table 2). The patients, not previously treated for SAHS, had no other active medical problems and had no obstructive pattern in routine pulmonary function tests.

In patients the FDPAP ventilator (Figure 4) was connected to a conventional nasal mask and and P_n (sampled at the mask) were recorded as described for the models. Before beginning the study, with the patient awake in the recumbent position and with all the equipment connected, we carefully fitted the nasal mask to minimize leaks. To this end, we applied a CPAP of 5 cm H₂O, required the patient to stop breathing for a brief period and, if necessary, adjustments of the mask were made until leak with CPAP = 5 cm H₂O was less than 50 ml/s. In two of the patients (Patients 8 and 9) esophageal pressure (P_es) was estimated by means of an esophageal balloon connected to a pressure transducer similar to the one used to measure P_n. The balloon was positioned and validated as described in the literature (16). P_es was sampled and recorded in the same way as and P_n. Application of FDPAP was carried out during a nap period. After the patient achieved a well-established sleep phase according to full polysomnography, at least 1 h of sleep was recorded. Most of the study was carried out in non-REM sleep. Electroencephalogram (EEG) (C4/A1, C3/A2), chin electromyogram (EMG), and electro-oculogram (EOG) for sleep staging according to standard criteria (17) were recorded. Sa_O2 was measured continuously with a finger probe (504; Critical Care Systems, Inc., Waukesha, WI). Rib cage and abdominal motion were monitored by bands placed over the thorax and abdomen. The parameters were recorded continuously on a polygraph (SleepLab 1000P; Aequitron, Plymouth, MN). Respiratory events were scored according to commonly used criteria, with apnea defined as cessation of airflow lasting for 10 s or more and hypopnea as a clear reduction in airflow lasting 10 s or more, in association with an arousal or with a cyclical dip in Sa_O2 (> 4%). Arousals were scored according to the American Sleep Disorders Association recommendations (18).

The patient was initially subjected to a CPAP = 4 cm H₂O by setting P₀ = 4 cm H₂O and k = 0. Once a well-established sleep stage II was achieved, P₀ (k = 0) was progressively increased to determine the optimal CPAP (CPAP_opt), which in patients was defined as the one required to avoid arousals, apneas, hypopneas, and flow limitation as detected by the flow curve (19, 20). The settings of P₀ and k were determined as previously described (Figure 3). The process of modifying P₀ and k was repeated until further reduction of P₀ resulted in flow limitation provided that sleep remained stable according to the polysomnographic criteria (absence of arousals, apneas, hypopneas, and flow limitation). The final settings of P₀ and k were maintained for the rest of the nap. In the two patients with esophageal balloon, the work of breathing (W_br) was computed as the integral W_br =  P_es · · dt over entire breathing cycles divided by the tidal volume. Results were expressed as mean ± SD and statistical comparisons of P_n,mean were made with paired t test. Differences were considered significant when p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Model Study

Figure 5 shows P_n, , and P_tr recorded when model M₁ was subjected to the optimal CPAP (12 cm H₂O), to bilevel pressure (12-4 cm H₂O), and to FDPAP (P₀ = 5.5 cm H₂O, k = 11.0 cm H₂O · s/L). With CPAP_opt, the pressure swing P_tr required for breathing was 13.2 cm H₂O and W = 1.03 J/L. With bilevel pressure we observed that and P_tr showed sudden transients at end-inspiration and end-expiration as a result of the bilevel P_n applied. With this mode P_n,mean, P_tr, and W were considerably reduced to 7.9 cm H₂O, 5.8 cm H₂O, and 0.302 J/L, respectively. With FDPAP, both P_n,mean and W showed a further decrease to 6.6 cm H₂O and 0.270 J/L, respectively. Moreover, with FDPAP and P_tr did not show the sudden transients observed with bilevel pressure. Placing a leak between the pneumotachograph and R_n on FDPAP mode resulted in an increase in P_n,mean from 6.6 to 9.0 cm H₂O and in a decrease in W to virtually zero.

View larger version (23K): [in this window] [in a new window]   Figure 5.   Pn, , and P tr when model M1 was subjected to the optimal CPAP, BiPAP, and FDPAP. Dashed lines correspond to the critical pressure of the collapsible airway. Note that with FDPAP, Pn was maintained above 3 cm H2O to avoid rebreathing.

For the analog configurations including a linear R_n and an almost ideal Starling resistor (M₁-M₄), the values of P₀ and k determined according to the setting procedure were close (within 0.5 cm H₂O in P₀ and 0.5 cm H₂O · s/L in k) to the actual P_crit and R_n (R_n = K₁ + K₂ · ), respectively (Table 1). In model M₅, which included a nonlinear R_n, the resulting settings were P₀ = 6.5 cm H₂O and k = 16.4 cm H₂O · s/L. In model M₆ including a stiff collapsible segment the final settings were P₀ = 8.0 cm H₂O and k = 15.0 cm H₂O · s/L. The results shown in Table 1 and Figure 6A, which is a scatter plot of work versus mean nasal pressure for the different models (M₁-M₆), illustrate the performances of CPAP and FDPAP. When compared with optimal CPAP, FDPAP allowed the same level of ventilation by applying a lower mean nasal pressure and by requiring lower inspiratory pressure swings and work of breathing in all the models.

View larger version (24K): [in this window] [in a new window]   Figure 6.   Scatter plot of the work of breathing performed by the simulated patient and Pn,mean for the different airway models when subjected to optimal CPAP (open circles) and to FDPAP (closed circles). Numbers 1 through 6 correspond to models M1 through M6, respectively. (A) Reference minute ventilation (11.2 L/min). (B) Increased minute ventilation (15 L/min).

The main advantage of FDPAP when compared with CPAP or bilevel pressure was more apparent as ventilation was increased. Figure 7 shows the signals recorded when model M₁ was subjected to a 33% increase in ventilation while maintaining the ventilator settings. With bilevel pressure we observed clear flow limitation during the greater part of the inspiration. Such an inspiratory flow limitation was also observed in CPAP_opt since inspiratory P_n was the same. P_tr increased considerably to 26.1 cm H₂O and, consequently, W rose to 1.76 J/L. By contrast, FDPAP was able to automatically adapt the nasal pressure so as to account for the increase in dynamic pressure corresponding to a greater flow. The associated increase in inspiratory nasal pressure with FDPAP (Figure 7) determined that P_n,mean was similar for FDPAP and bilevel pressure (7.6 and 8.0 cm H₂O, respectively). Nevertheless, the model could be ventilated sinusoidally with almost the same breathing effort as in the situation of lower ventilation: P_tr was 7.7 cm H₂O and W = 0.404 J/L.

View larger version (26K): [in this window] [in a new window]   Figure 7.   Pn, , and P tr when model M1 with increased minute ventilation was subjected to the same settings of BiPAP and of FDPAP as in Figure 5. Dashed lines correspond to the critical pressure of the collapsible airway.

The response of FDPAP when applied to the other airway models tested under conditions of increased breathing flow was similar to that found in model M₁ (Figure 7). As shown in Figure 6B, when ventilation was increased by 33% the response of FDPAP was always better than that of CPAP_opt for all the investigated models.

Patient Study

Table 2 shows that, as expected in patients with severe SAHS, CPAP_opt was high (9.1 ± 1.2 cm H₂O). During the procedure for setting P₀ and k (Figure 3), P₀ could be decreased (normal breathing pattern and polysomnography) down to a value for which the patient exhibited an inspiratory flow limitation profile, hypopneas or apneas. We observed that decreasing P₀ by another step of 0.5 cm H₂O usually resulted in apneas and clear polysomnographic events. As described in METHODS the final settings of FDPAP (Table 2) were the last values of P₀ and k before the appearance of flow limitation or sleep events (Figure 3). According to the polysomnography criteria, all the patients showed normalized sleep when subjected to FDPAP. Figure 8 shows an example of the flow and nasal pressure recorded in one of the patients during application of FDPAP. The figure also shows signals P_n and recorded when a CPAP with a P_n,mean similar to that of FDPAP was applied to the same patient. With FDPAP, the patient was able to breathe with a normal flow pattern. By contrast, when subjected to the CPAP mode with a similar P_n,mean the flow signal displayed the typical pattern of inspiratory flow limitation.

View larger version (19K): [in this window] [in a new window]   Figure 8.   Example of Pn and recorded in a patient when subjected to FDPAP and to a CPAP level with almost the same mean nasal pressure as with FDPAP.

As shown in Figure 9, mean nasal pressure systematically and considerably decreased (p < 0.01) by application of FDPAP when compared with CPAP_opt. On average P_n,mean was reduced from 9.1 ± 1.2 cm H₂O for CPAP to 6.6 ± 1.2 cm H₂O for FDPAP. In the two patients with esophageal balloon we found that, in addition to reducing P_n,mean, application of FDPAP decreased the esophageal pressure swing (P_es) and the work of breathing W_br. In Patient 8, P_es were 8.4 and 6.6 cm H₂O and W_br were 0.54 and 0.45 J/L for CPAP_opt and FDPAP, respectively. In Patient 9, P_es were 8.3 and 6.5 cm H₂O and W_br were 0.63 and 0.55 J/L for CPAP_opt and FDPAP, respectively.

View larger version (18K): [in this window] [in a new window]   Figure 9.   Mean nasal pressure (Pn,mean) in the patients when applying optimal CPAP and when applying FDPAP. Open circles: individual values; closed circles: mean ± SD.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To maintain airway patency the intraluminal pressure in the collapsible airway must be at least equal to the critical opening pressure of the airway. However, the nasal pressure that should be applied to avoid airway obstruction in patients with SAHS does not coincide with the value of the critical airway pressure (Equation 4). This is due to the dynamic pressure drop associated with airflow through the resistance from the nostrils to the collapsible airway (Equation 1). Consequently, the rationale of FDPAP is based on adapting the nasal pressure to the instantaneous flow so that the amplitude of the applied pressure is minimized and its time pattern is matched with the actual patient's flow.

An advantage of FDPAP, when compared with CPAP and bilevel pressure, is that the FDPAP mode of support is aimed at automatically compensating for any change in the patient's ventilation. Indeed, as illustrated in the model study the optimal CPAP and the inspiratory bilevel pressure are titrated to eliminate flow limitation at peak inspiration for a certain level of flow. Consequently, the CPAP and the bilevel models of support decreased P_tr just to P_crit (Figure 5). Therefore, increasing ventilation, as in Figure 7, while maintaining the ventilator settings resulted in flow limitation. In this condition of increased ventilation, maintaining airway patency with CPAP or bilevel pressure would require a higher P_n to prevent flow limitation. This, however, would further increase nasal and intrathoracic pressures. By contrast, FDPAP was able to autoregulate P_n (Equation 7) when ventilation was increased (Figure 7). If, instead of increasing, flow decreased, the CPAP and bilevel modes would apply a P_n greater than necessary any time, even at peak inspiration. Nevertheless, in this instance of decreased ventilation FDPAP would reduce P_n to the level strictly necessary at all times during the breathing cycle.

The FDPAP is based on applying a nasal pressure with a constant component and a component proportional to flow (Equation 7). From the technical viewpoint of pressure generation, FDPAP is similar to proportional assist ventilation (PAV) (21) with settings for only flow support and with added positive end-expiratory pressure (PEEP). The only difference would be that FDPAP is applied both during inspiration and expiration and PAV is intended to apply flow-proportional pressure only during inspiration. Nevertheless, despite the technical similarity between FDPAP and PAV, it is worth noting that the rationale for using these two approaches is completely different. Indeed, with PAV P_n is aimed at providing a pressure in proportion to the patient's muscle effort in order to compensate for a respiratory system with increased impedance and/ or weakened inspiratory muscles. The addition of PEEP is addressed to compensate for dynamic hyperinflation (22). By contrast, FDPAP is aimed at maintaining airway patency by applying the nasal pressure required to avoid static and dynamic airway obstruction at any time during the breathing cycle (Equation 4). Consequently, the constant term P₀ corresponds to the critical pressure of the supper airway and k represents only a fraction of the total respiratory resistance (R_n: upstream the collapsible segment). Accordingly, we defined a process (Figure 3) for setting the ventilation parameters (P₀, k) which took into account the particular mechanical problem to solve in the application of pressure support in SAHS.

The suitability of FDPAP for reducing the pressure support (P_n,mean) and the breathing work (W) required to allow normal breathing was confirmed when applied to a variety of analog airway models characterized by different P_crit, R_n, and airway wall compliance (Figure 6, Table 1) and in patients (Figures 8 and 9). In fact, greater reductions in P_n,mean and W could be obtained if the minimal P_n = 3 cm H₂O, which we fixed to avoid rebreathing in our setup, were not necessary, for instance by using a ventilator with separate inspiratory and expiratory lines. When applied in patients FDPAP allowed us to reduce the magnitude of nasal support for a P_n,mean of 9.1 ± 1.2 cm H₂O to 6.6 ± 1.2 cm H₂O (Figure 9). The estimated value of the critical pressure (P₀) was on average 5.9 cm H₂O (Table 2), which is in keeping with the values measured conventionally in patients with severe SAHS (12, 13). Moreover, the value of the upper airway resistance estimated by applying FDPAP (k = 5.2 ± 1.8 cm H₂O · s/L was consistent with the values of lung (23, 24) and respiratory (25) resistance measured by other methods in the absence of flow limitation, which was the case in our patients during application of FDPAP. In the model study, FDPAP was particularly performant with models M₁-M₄ simulating an almost pure Starling resistor. Moreover, the designed mode of support was also useful when applied to more complex and realistic airway analogs (M₅, M₆) exhibiting nonlinear R_n (10, 26) or with a collapsible segment which does not behave as a pure Starling resistor (8). In these instances, the setting procedure resulted in values of P₀ (6.5 and 8.0 cm H₂O for M₅ and M₆, respectively) which were slightly greater than the effective P_crit (5.0 and  6 cm H₂O for M₅ and M₆ [15], respectively). The values of estimated R_n (k) were also greater than expected: for M₅ the value of k = 16.4 cm H₂O · s/L corresponded to the resistance of the nonlinear R_n at flow 0.5 L/s, and for M₆ k was 56% greater than the actual R_n. In this regard, it should be noted that FDPAP is conceptually based on the assumption that P_crit and R_n remain constant (Starling model in Figure 1). This means that the collapsible tube presents only two possible states (open or closed) according to the transluminal pressure. However, the upper airway in patients with SAHS does not exactly meet this assumption owing to nonlinearities in the upper airway resistance (10, 26), to hysteresis phenomena in the critical pressure (19), and to the static elastic properties of the airway wall (27). In the case of nonlinear R_n or in a collapsible tube with relatively stiff wall (and hence with a resistance depending on transluminal pressure), the settings of P₀ and k are not expected to exactly coincide with P_crit and R_n. However, the procedure (FDPAP) of applying a nasal pressure with a static (P₀) and a dynamic (k · ) component is expected to be effective in reducing P_n,mean and W, as occurred in models M₅ and M₆. This suggests that comparison of the estimates of P_crit and R_n obtained from FDPAP and from other methods used in the literature (4, 10, 12, 13) could allow us to test the adequacy of the models used to interpret airway mechanics in SAHS and, hence, to better characterize the pathophysiology of the collapsible upper airway.

Practical application of FDPAP in patients may be affected by a possible air leak in the mask. Indeed, with such a leak the flow measured by the pneumotachograph is greater than actual flow and consequently, the applied P_n is higher than required. However, when assessing the importance of this problem we found that even a considerable level of air leak (0.25 L/s) resulted in a modest nasal overpressure (2.4 cm H₂O). Moreover, the presence of air leaks high enough to produce a significant misestimation of actual breathing flow can be automatically detected and quantified from the pneumotachograph flow signal. A second concern which may arise from the application of FDPAP in patients is related to the fact that this mode of support assumes that the estimates (P₀, k) of P_crit and R_n obtained during the setting procedure remain constant. This, however, may not be the case because P_crit and R_n change in the short time (posture and sleep phase [28], alcohol ingestion [29]) or over a longer time period (change in body weight [30] or adaptation to treatment [31, 32]). To solve this practical shortcoming, which is in common with CPAP titration, it is possible to implement the concept of autotitration (33, 34) to FDPAP. To this end, the ventilator could be programmed for periodically repeating the setting procedure to update P₀ and k according to the actual changes in P_crit and R_n.

The application of FDPAP to our patients with SAHS during sleep showed that this support procedure was well tolerated and that it allowed the patients to sleep normally with a mean pressure level significantly lower than in the case of CPAP. These preliminary results suggest that FDPAP has a potential interest both in optimizing the treatment and in better understanding the mechanisms determining upper airway obstruction in SAHS. However, future clinical work should be aimed at ascertaining whether FDPAP is effectively useful in maintaining airway patency while reducing intrathoracic pressure and improving patient compliance.