INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

AMONG THE DIFFERENT FORMS of partial ventilatory support, pressure support (PS) ventilation is the most commonly used, both in the early phase of acute respiratory failure and during weaning from mechanical ventilation (7, 9, 23, 27). During PS, each spontaneous breath is assisted by a constant positive pressure applied by the ventilator all along each inspiration. In the early study by Kacmarek (21) of intensive care PS devices, some of the devices failed to reach the prescribed pressure before the end of inspiration or failed to maintain the PS level near the inspiratory peak flow (_peak), i.e., at onset of inspiration. To minimize pressure instability during inspiration, some manufacturers have developed devices in which pressure is servo controlled to obtain a square pressure wave. Usually, the pressure-measuring site used to servo control the PS device is inside the ventilator. Because the inspiratory line of the breathing circuit and the respiratory system generate substantial resistance to flow, a pressure gradient exists between the ventilator, the airway opening, and the alveoli (19). Therefore, when a square pressure signal is generated inside the ventilator, the pressure signal in the airways is far from being constant throughout inspiration and is even highly flow dependent at the alveolar level. We hypothesized that a PS device capable of providing a square pressure signal at the airway opening would reduce the work of breathing compared with a conventional PS device producing a square pressure signal in the ventilator and that this effect would be further increased if the PS device maintains the square pressure signal at the level of the alveoli.

To test these hypotheses, we have built a ventilator in which the location of the inspiratory positive-pressure plateau can be moved from the ventilator to the alveoli. The site where the pressure is controlled and regulated as a plateau can be inside the ventilator, at the airway opening, or at the alveolar level. One of these three sites is selected after estimation of the overall airway pressure (Paw) drop. The effect of PS generated at the three different levels of the ventilator-patient system (PS_vent, PS_aw, and PS_A) was physiologically evaluated by measuring ventilatory parameters and the esophageal pressure (Pes)-time product in normal subjects under normal conditions (air breathing) and during breathing of a gas mixture enriched with 5% CO₂ to enhance ventilatory demand.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

Apparatus Tested

1

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Basic setup consisting of a 1-m-long inspiratory line, a 1-way valve, and an expiratory valve that separated expiration from inspiration. To this basic setup, a device that created positive inspiratory pressure was added. However, when pressure support (PS) device was not activated, total resistance of inspiratory apparatus was ~2.5 cmH2O · l1 · s and entire setup could be used to control measurements. To obtain a positive inspiratory pressure, a jet of air or CO2-enriched gas mixture from a pressurized cylinder was blown into an injector, where it entrained a flow of gas from a balloon (100 liters). Pressure was regulated by an electrovalve upstream from jet injector. This electrovalve opened when inspiratory flow rose to >17 ml/s and remained open until inspiratory flow fell to <150 ml/s. Level of electrovalve opening was servo-controlled to keep pressure constant at different levels of ventilator-patient circuit. Site of pressure stability was located inside ventilator (PSvent regulation) when pressure was measured at this site, at airway opening (PSaw regulation) when pressure was measured near mouthpiece, or at level of alveoli (PSA regulation) when a formula was used to estimate alveolar pressure (PA) from opening airway pressure (Paw), airway resistance (Raw), and inspiratory flow (): PA = Paw  ( · Raw).

To maintain the pressure plateau in the various modes of PS tested, a servo valve was placed upstream of the jet injector, instead of the simple on-off electrovalve used in the previous system. The servo valve was programmed to open when rose above 17 ml/s (1 l/min) and to remain open until fell below 150 ml/s, as in conventional inspiratory PS systems. Contrary to these conventional systems, however, servo valve opening was commanded by a computer that compared, at intervals of 2 ms, the measured pressure [corrected or not corrected for airway resistance (Raw)] with the desired plateau pressure. The difference between these two pressures was minimized by using a second-order control loop that was modified to take into account the constant delay due to dry friction in the servo valve and to the self-inductance of the magnetic coil. The loop regulation of the pressure is described in the APPENDIX (11, 15). In practice, after a cycle, the servo valve opening was progressively increased to compensate for 1) the pressure drop through the pressure generator in the PS_vent mode, 2) the pressure drop through the PS device inspiratory line in the PS_aw mode, and 3) a minimal estimated airway pressure drop in the PS_A mode. With our computer-controlled PS device, the pressure plateau was generated at the site where the reference pressure was measured (or estimated), i.e., 1) inside the ventilator (PS_vent), 2) at the airway opening (PS_aw) when pressure was measured near the mouthpiece, or 3) at the alveolar level when alveolar pressure (PA) was estimated (PS_A regulation). PA was estimated on the basis of the Paw, Raw, and according to the following formula Raw was taken to be constant and equal to 80% of the resistance of the respiratory system (Rrs). Rrs was measured in each subject with use of the forced oscillation method (10).

Before using the above-described modified PS device in our study subjects, we tested it in vitro using a lung model simulating patient inspiratory effort (Fig. 2, top). The lung model was a two-chamber Michigan test lung. One chamber was connected to and powered by a motor ventilator with a flow-controlled mode in such way that the effort generated was not influenced by the connected system (Cesar, Taema, Antony, France) (driving chamber), whereas the other chamber (PS-pressurized chamber) was connected to the PS ventilator under test. The two chambers were physically connected to each other by a small metal piece that allowed the driving chamber to lift the PS-pressurized chamber, thus mimicking the patient's contribution to inspiration. With this system, the generation of positive pressure in the driving chamber lowered the pressure in the PS-pressurized chamber to subatmospheric levels, just as inspiratory muscle contraction produces negative PA in vivo. This effect was detected by the triggering system of the PS ventilators under test. Because the metal component was not secured to the PS-pressurized chamber, the latter, once effectively pressurized, could rise above the driving chamber within a time interval dependent on the mechanical properties (resistance and compliance) of the PS-pressurized chamber. The compliance of this chamber was set at 80 ml/cmH₂O. Positive end-expiratory pressure (PEEP) was applied to the driving chamber at a level that ensured synchronized motions between the two chambers at onset of inspiration. A resistance of 4 cmH₂O at 1 l/s, connected between the device and the lung model, was used to mimic the patient's Rrs. The motor ventilator was set to obtain a predetermined respiratory frequency of 15 cycles/min, an inspiratory time (TI) of 1.2 s, and a decelerating flow with a _peak of 1.1 l/s.

View larger version (39K): [in this window] [in a new window]   Fig. 2.   Top: experimental setup. Circuits of tested PS devices were connected to a 2-chamber Michigan test lung. One chamber (driving chamber) of test lung was attached to and powered by a ventilator; the other chamber (PS-pressurized chamber) was connected to PS device under study. Driving chamber was able to lift PS-pressurized chamber and induced an initial negative pressure until PS device contributed to expansion of freely moving PS-pressurized chamber. Pressure and flow were measured at end of PS device circuit. C, compliance. Bottom: pressures obtained inside ventilator (Pvent), at mouth (Paw), and inside pressurized chamber (PA) during simulated spontaneous breathing with PS device disconnected (SB) and during each condition of PS regulation: PSvent regulation, PSaw regulation, and PSA regulation.

A Fleisch no. 2 pneumotachograph (Gould Electronic, Longjumeau, France) was inserted between the lung model and the circuit of the tested device. Pressure was measured inside the ventilator (P_vent), at the airway opening (Paw), or inside the pressurized chamber (PA) with a differential pressure transducer (model MP45, ±70 cmH₂O, Validyne, Northridge, CA) used at each pressure site. Figure 2, bottom, shows the P_vent, Paw, and PA signals for a PS set at 7 cmH₂O during simulated spontaneous breathing (SB) with the PS device disconnected and with each of the three PS regulation conditions tested; the data clearly demonstrate the efficiency of the pressure-regulated device used in our study.

Clinical Study

Measurements. The subjects were seated, wore a noseclip, and breathed via a mouthpiece. Flow was measured using a pneumotachometer (Fleisch no. 2) connected to a pressure transducer (model MP45, ±2 cmH₂O, Validyne) and integrated to yield tidal volume (VT). Paw (model MP45, ±50 cmH₂O, Validyne) and PCO₂ in respiratory air [end-tidal PCO₂ (PET_CO2); infrared analyzer, Gould, Ballainvilliers, France] were measured in the breathing tube close to the lips. Pes and gastric pressure (Pga) were recorded using a catheter-mounted transducer (Gaeltec, Dunvegan, Isle of Skye, UK). The validity of the Pes measurements was checked by analyzing the shape of the Pes curve after swallowing and by using the occlusion technique (5).

Experimental protocol. Two sessions were performed in each subject with two successive inspired CO₂ fractions: 0% CO₂ and 5% CO₂ to stress the respiratory system. Within each session, four 10-min measuring periods were performed, each with a different ventilatory mode: SB, PS_vent regulation, PS_aw regulation, and PS_A regulation. The order of the different ventilatory modalities was different from one subject to another and was determined using a randomization table. For each PS ventilation mode, the level of plateau pressure was set at 7 cmH₂O.

Data analysis. The variables were recorded, after stabilization, from the 7th to the 10th min of each period. The following variables were read breath by breath: TI as the onset of to the onset of the expiratory flow, TE as the remainder of the total breath duration, VT from the calibrated integrated flow signal, PET_CO2 as the peak of the airway CO₂ record, Pes, Pga, Paw, and _peak.

Statistical analysis. Values are means ± SD. Statistical differences between the three assisted ventilatory modes within each session were tested using the nonparametric Friedman test because of the small number of subjects. The 5% level was chosen as significant. When a significant difference was observed, bilateral comparisons were performed using the Wilcoxon signed-rank test. The Wilcoxon test was also used to assess bilateral comparisons between air and 5% CO₂ breathing.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

Effects of PS Mode on Breathing Pattern

During air breathing, VT was largest during PS_A, followed by PS_aw, PS_vent, and SB. The differences between each of the three PS modes and SB were significant. No significant differences were found during 5% CO₂ breathing.

During air breathing, E was largest during PS_A, followed by PS_aw, PS_vent, and SB. No significant differences were observed during 5% CO₂ breathing. Similarly, we found a trend to a decrease in PET_CO2 from PS_vent to PS_aw and PS_A (P = 0.07) during air breathing, whereas PET_CO2 was identical during 5% CO₂ breathing with all three PS modes. There was a tendency to a reduction in TI during air breathing, with no significant variation in TI/TT. During 5% CO₂ breathing, TI and TI/TT decreased significantly from PS_vent to PS_aw and PS_A.

Effects of PS Mode on Pes

The mean level of PEEP_dyn,i was <1 cmH₂O with all ventilator modes during air and 5% CO₂ breathing. There were no significant differences among the four ventilatory modes. Also, the shape of the Pga curves did not suggest any significant abdominal expiratory activity (22).

Modifications in Paw, VT/TI, and _peak

View this table: [in this window] [in a new window]   Table 3.   Mean inspiratory airway pressure, VT/TI, and peak

View larger version (75K): [in this window] [in a new window]   Fig. 3.   Tracing obtained in a representative patient during air breathing and 5% CO2 breathing and during 4 ventilatory modes: SB, PSvent regulation, PSaw regulation, and PSA regulation. Pes, esophageal pressure; FICO2, fraction of inspired CO2.

Responsiveness to CO₂

View larger version (18K): [in this window] [in a new window]   Fig. 4.   Relationship between Pes-time product per breath (PTP/b), which represents inspiratory effort, and Paw-time product during inspiration per breath (PawTP/b), which represents assistance by machine during air breathing () and 5% CO2 breathing ().

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES APPENDIX

PS ventilation is a pressure-targeted mode in which pressure is delivered in a square-wave pattern that begins with the patient's inspiratory effort and terminates when a threshold of decreasing is reached. However, a discrepancy between the expected inspiratory plateau of pressure and the observed pressure waveform is frequently observed in clinical practice (21). This discrepancy is related to high , high resistive properties of the open-valve demand system and of the tubing between the pressure generator and the patient, and/or differences in pressure rise profiles. In extreme situations the result can be a lack of positive-pressure assistance at the beginning of inspiration, which is the time when resistive WOB is greatest.

Our laboratory has developed a regulation system that takes into account Paw and Raw. We used this system to evaluate the physiological effects of moving the pressure plateau from the generator to the mouth and presumably to the alveolar level. The regulation system was incorporated into a ventilator designed for noninvasive PS ventilation in intensive care and currently used in clinical practice with good results (9, 23). The efficacy of our regulation system was first checked using a lung model, with satisfactory results.

Once a given inspiration is initiated, the PS ventilation device delivers a high , which decreases gradually throughout the inspiration. Clearly, our servo-regulated system allows adjustment of the initial flow to the value needed to reach the appropriate preset pressure level at different sites in the airway between the ventilator and the alveoli. As expected, we found that when the system was set to produce a pressure plateau in the airways rather than in the ventilator, the initial ventilator-delivered pressure and flow increased to overcome the resistance of the ventilator circuit. When the system was set to produce a pressure plateau in the alveoli, further increases in ventilator-delivered flow and pressure were observed; this effect would theoretically improve the rate of alveolar pressurization without increasing peak PA.

WOB includes a predominant elastic component and a flow-resistive component. Breathing through circuit tubing, connectors, a humidifier, internal pneumatic circuitry, and ventilator valves necessarily increases the resistive component. This may be crucial at the beginning of inspiration, at the time where the prescribed inspiratory plateau pressure has not yet been reached and the inspiratory flow demand is the highest. Therefore, the pressure delivered by a conventional PS ventilator (PS_vent) at the beginning of inspiration may undercompensate the resistive loading imposed by the circuitry and airway even when the level of prescribed PS is high. This may result in a zero or negative PA at the beginning of inspiration, and therefore the elastic and resistive load imposed by the respiratory system on the inspiratory muscles may not be reduced by PS during a crucial part of inspiration. This phenomenon can theoretically be minimized by prescribing higher levels of PS. However, these high levels can cause respiratory system overdistension and barotrauma as a result of high pressures being delivered at the end of inspiration (17). Also, high levels of PS can produce desynchronization between the patient's SB and the ventilator (12). Alternatively, regulating the pressure at the alveolar level offers two advantages: 1) compensation for the resistive load due to the circuitry and airway at any level of ventilatory demand, resulting in positive PA from the very beginning of inspiration, and 2) minimization of the risk of barotrauma or patient-ventilator desynchronization, since in theory there is no increase in peak PA.

Before discussing the possible advantages of this approach, we will address several potential limitations or negative consequences.

The proposed approach, in which we apply a constant pressure at the alveolar level, produces an abnormal flow pattern with an elevated early peak flow. This very rapid flow acceleration is not necessarily well tolerated at a subjective level, and the impact of this very high sheer force at the beginning of inspiration may be significant when the Rrs and/or the level of plateau pressure is high. Moreover, if the total Raw is completely compensated during PS_A, with the assumption that gas inertance is negligible, the time constant of the respiratory system would decrease to zero. In this case, unless the subject is making a substantial crescendo effort to maintain flow, flow will essentially consist of a very brief spike, and the cycle will be immediately terminated. This pattern, which should clearly not be an acceptable outcome, was not observed in our study. In fact, extreme shortening of TI did not occur in our study, possibly because we have compensated for only 80% of the Rrs during PS_A. Furthermore, the forced oscillation resistance is estimated in the linear range and tends to underestimate the actual resistance that is flow dependent during breathing.

Another limitation is the increase in the peak pressure at the upper airway level, induced by regulating the pressure at the alveolar level, which may facilitate a possible gastric inflation. To prevent this gastric inflation during noninvasive ventilation, we voluntarily limited the maximal Paw induced by our apparatus to 25 cmH₂O. Thus our ventilator should fail to maintain a Paw or PA if the Rrs and/or the setting of the level of plateau pressure is too high. In addition, we do not know the effect of PS_A in patients with airflow obstruction in which pulmonary heterogeneities exist, but it is probable that regional PA differed with differences in regional time constants.

Therefore, the findings from this study cannot be generalized, and the results may have been different at a higher level of the Rrs, at a higher setting of the plateau pressure, and/or when pulmonary heterogeneities predominate.

Previous studies have demonstrated that the site of pressure regulation during continuous positive airway pressure (CPAP) influences the imposed WOB (1, 3, 4, 13, 28). To eliminate the WOB imposed by the breathing circuit and ventilator, many CPAP devices have been designed with the goal of controlling the pressure level at the Y piece (1). More recently, attempts have been made to overcome the work imposed by the endotracheal tube, which is an additional resistive component of the breathing apparatus, by relocating the site of pressure regulation to the tracheal end of the endotracheal tube (3, 4, 28) or by estimating the tracheal pressure from the Paw and the endotracheal tube resistance (13). These studies consistently found that for a given CPAP level the WOB decreased when the pressure-controlling site was physically moved closer to the ventilatory muscles. In keeping with these studies, we observed that during PS ventilation inspiratory activity decreased when the site of the preset pressure was moved from the ventilator to the alveoli.

Varying the initial flow rate may interact with the breathing pattern and/or the patient's effort. The respiratory control system has options: 1) it can maintain the central respiratory output, utilizing the additional unloading to produce more ventilation, or 2) it can maintain ventilation at the baseline level and use the assist to reduce respiratory muscle work. Any combination of these two extremes is also possible, and the result may vary with the ventilatory demand.

During air breathing in the resting condition, PS ventilation is known to induce hyperventilation and hypocapnia as a result of an increase in VT and also to decrease the inspiratory drive (2, 18, 24, 30). Our results confirmed this phenomenon and demonstrated that it was accentuated when the initial flow rate was further increased by moving the site of regulation of the appropriate preset pressure from the ventilator to the alveolar level.

The literature suggests that in acute conditions the additional unloading is used primarily to reduce inspiratory activity (6, 23). In a study of intensive care unit patients, Bonmarchand et al. (6) found that an initial flow rate increase during PS was associated with a decrease in diaphragmatic activity, although arterial PCO₂ and minute ventilation remained unchanged over a broad range of flow rate variation. Similarly, we recently compared several PS devices in critically ill patients undergoing weaning from mechanical ventilation. We did not observe any differences in arterial PCO₂ and minute ventilation, but we found that the inspiratory PS device that provided the highest initial flow rate was more efficient in reducing inspiratory activity than the inspiratory PS device that provided the lowest initial flow rate (23). These findings suggest that intubated patients with respiratory failure used the additional unloading afforded by the increase in the initial flow rate to reduce respiratory muscle work rather than to produce more ventilation.

Noncontradictory results were observed in normal subjects when inspiratory activity was constrained by CO₂ inhalation. During PS ventilation and CO₂ inhalation, subjects have full control over the onset of inspiration. TI and VT were dependent on the level of PS and the activity of inspiratory muscles. TE was determined to be the time when the subject triggered the next inspiration. Therefore, a respiratory response to inhaled CO₂ is possible during conventional PS, and we previously found that PS shifted to the left the relationship between E and PET_CO2 and reduced inspiratory activity at any given PET_CO2 (14, 18, 30). Therefore, these studies suggested that conventional PS induced a nonchemical inhibition of inspiratory activity. Our study confirms the findings of these previous studies and suggests that the nonchemical inhibition is influenced by the shape of the initial flow rate induced by the ventilator, since we observed that moving the site of regulation of the appropriate preset pressure from the ventilator to the alveoli was associated with a significant decrease in the inspiratory effort indexes with no statistically significant differences for E and PET_CO2. It is noteworthy that the reduction of inspiratory activity is not only due to the reduction in duration of inspiratory effort, since significant reductions were also observed with WOB indexes, which do not depend on TI.

Previous studies have shown correlations between the Pes-time product and measurements of O₂ utilization and CO₂ production by the contracting respiratory muscles (8, 29). This led us to expect a decrease in CO₂ production with a reduction in arterial PCO₂ at a given level of alveolar ventilation during PS_aw and PS_A compared with PS_vent. However, although E was similar during CO₂ and air breathing, PET_CO2 values were virtually identical under all the PS ventilation conditions. This is probably ascribable to the fact that a number of phenomena have opposite effects on arterial PCO₂. First, although a diminution in the work of breathing can lead to a diminution in CO₂ production, the magnitude of this effect is probably limited. Second, the increase in peak inspiratory pressure in the upper airways that occurs when the site of pressure regulation is moved from the ventilator toward the alveoli (Fig. 3) may increase the distension of the upper airways, leading to an increase in the anatomic dead space. Third, variations in the physiological dead space related to variations in TI/TT have been reported during controlled mechanical ventilation in different circumstances, although their mechanism remains unclear (16, 25, 26). Rebreathing can probably be excluded on the basis of the continuous analysis of the PET_CO2 records (Fig. 3). Finally, some variation in E was observed among the three PS modes; although the differences were not statistically significant in such a small group of subjects, we cannot rule out variations in alveolar ventilation.

Despite similarities between CO₂ stimulation in our study and that observed in most patients in clinical practice in tonus of elevated inspiratory drive, major differences exist because of differences in respiratory resistance and compliance. For example, Jubran et al. (20) observed that although PS ventilation reduced respiratory frequency in patients with chronic obstructive pulmonary disease, several of these patients displayed expiratory activity. The patients with expiratory activity were those with long TI, which facilitated expiratory muscle activity during lung inflation or shortened the time available for lung emptying. Because our servo-regulated system allows VT/TI to increase by moving the site of preset pressure regulation from the ventilator toward the alveoli, it also reduces TI (as observed in our study) and may therefore reduce the occurrence and/or the importance of expiratory activity in these patients. Unfortunately, we were unable to demonstrate this potential beneficial effect, because we did not logically observe in our normal subjects any expiratory activity as detected from the shape of Pga recordings in all studied conditions.

In conclusion, we have built a PS device allowing inspiratory pressure regulation to be moved from the ventilator toward the alveolar level. The efficacy of this system has been confirmed during experiments first in a lung model, then in humans at rest and during stimulation with a 5% CO₂-enriched mixture. We observed a noticeable decrease in inspiratory activity when the site of the preset pressure was moved from the ventilator to the alveoli. This beneficial effect was particularly marked when ventilation was stimulated by CO₂. These data established in normal humans warrant a clinical evaluation.