Continuous Positive Airway Pressure in New-generation Mechanical Ventilators: A Lung Model Study

A bellows-in-a-box lung model was used to simulate spontaneous breathing (fig. 1). The space between the rigid box and the bellows represented the pleural space. The box was connected to a low compliance T tube through which gas flow created negative pressure in the pleural space. Source gas (50 psi, oxygen) was connected to a pressure regulator (SMC AR200, 7–120 psi; SMC Co., Tokyo, Japan) and a proportional solenoid valve (SMC 315; SMC Co.). A function generator (EGC 2230; Kenwood, Tokyo, Japan) controlled the opening of the solenoid valve. Adjustment of the regulator allowed variation of the negative pressure in the pleural space, and the function generator simulated a variety of spontaneous breathing patterns. Thus, inspiratory flow demand, respiratory rate, and inspiratory time were controlled independently. A linear resister (7100 R-5; Hans-Rudolph Inc., Kansas City, MO) created an airway resistance of 8.2 cm H²O · l⁻¹· s⁻¹at a flow rate of 60 l/min. Compliance of the lung model (50 ml/cm H²O) was determined by a spring in the bellows.

The following seven ventilators were evaluated: Bear 1000 (Bear Medical Systems, Inc., Riverside, CA), Drager Evita 4 (Drager, Inc., Telford, PA), Hamilton Galileo (Hamilton Medical AG, Rhazuns, Switzerland), Nellcor Puritan-Bennett 740 and 840 (Nellcor Puritan-Bennett Corp., Carlsbad, CA), Siemens Servo 300A (Siemens-Elema AB, Solna, Sweden), and Tbird AVS (Bird Products Corp., Palm Springs, CA). Each ventilator was connected to the lung model using a standard heated wire circuit (Hudson Respiratory Care Inc., Temecula, CA) and a Conchatherm IV Humidifier (Hudson Respiratory Care Inc.). However, the humidifier was not operational during any evaluation. Abbreviations used and settings for each ventilator are summarized in table 1.

A pneumotachometer (model 3700A; Hans-Rudolph Inc.) was placed at the airway opening of the lung model. The pressure differential across the pneumotachometer was measured (45-14-871 ± 2 cm H²O; Validyne, Northridge, CA), amplified (model 8805C; Hewlett Packard, Waltham, MA), digitized, and converted to a flow signal using a computerized graphics program (WINDAQ; Dataq Instruments Inc., Akron, OH). The pneumotachometer was calibrated at 1 l/s flow delivered by a precision flowmeter (Brooks Instruments, Hatfield, PA). Pressure at the airway opening and in the simulated pleural space was measured with differential pressure transducers (45-32-871 ± 100 cm H²O; Validyne). The pressure transducers were calibrated simultaneously at 20 cm H²O using a water manometer. Pressure signals were amplified (8805 C; Hewlett Packard) and digitized (WINDAQ).

In this study, simulated spontaneous breaths were set as follows: respiratory rate, 12 breaths/min; inspiratory time, 1.0 s; lung model peak inspiratory flow rates before ventilator attachment, 40, 60, and 80 l/min. CPAP was set at 10 cm H²O in all seven ventilators. Each ventilator was tested at the three lung model peak inspiratory flows. If the ventilator could be set in both flow and pressure triggering, we evaluated both at all peak flows. Inspiratory trigger sensitivity, base flow (or continuous flow) during the expiratory phase, and expiratory trigger sensitivity (cycling criteria) of tested ventilators are listed in table 1. Inspiratory trigger sensitivity was set as sensitive as possible without causing autotriggering in both trigger modes. In EVITA, a fixed pressure trigger sensitivity of −0.2 cm H²O and base flow of 1 l/min were set by the manufacturer (flow triggering is not active during CPAP in this ventilator). In BEAR and T BIRD, we arbitrarily set base flow at 5 and 10 l/min, respectively. Base flow of GALILEO, PB840, and S300 was predetermined by the manufacturer. In PB740, there was no “constant” base flow even though this is a flow trigger ventilator. In GALILEO and PB840, we set expiratory trigger sensitivity at 25% of peak inspiratory flow rate to be consistent with the settings of other ventilators. Other ventilator's expiratory trigger sensitivity was predetermined by the manufacturer.

Airway pressure (Pao), intrapleural (between box and bellows) pressure (Ppl), and airway flow were measured. All signals were digitized at 100 Hz and recorded using WINDAQ. The WINDAQ playback software was used to analyze data. We analyzed three consecutive breaths at each inspiratory flow in each ventilator during both triggering modes.

The beginning and the end of the inspiratory phase (inspiratory time: TI) was determined by changes in intrapleural pressure (fig. 2). A negative deflation in pleural pressure indicated the start of the inspiratory phase and the subsequent positive deflection signaled the end of inspiration. Airway baseline pressure (P^BL) was defined as airway pressure measured at the end of expiration. From the beginning of inspiration, the time required to decrease airway pressure to the lowest pressure below P^BLwas recorded as the inspiratory trigger delay time (D^T). Inspiratory trigger pressure (P^T) was defined as the difference between P^BLand the lowest pressure below P^BLduring triggering of inspiration. The airway pressure time product required to trigger ventilation (PTP^T) prebreath was defined as the area of the airway pressure–time tracing below P^BLduring triggering. The airway pressure time product during inspiration (PTP^I) per breath was defined as the area of the airway pressure–time tracing throughout total inspiration

.

The ventilator's delivered peak inspiratory flow (VPF) and delivered inspiratory tidal volume (V^T) were also measured.

During the expiratory phase, the expiratory delay time (D^E) was defined as the time from the onset of expiration until airway pressure reached its peak level during the expiratory phase. Expiratory trigger pressure (P^Emax) was defined as airway pressure difference between P^BLand maximum Pao during exhalation. The airway pressure difference between minimum Pao during triggering and the maximum Pao during exhalation (dP^AW) was also calculated. The expiratory airway pressure time product (PTP^E) per breath was defined as the area of the Pao--time tracing above baseline from the onset of expiration to the return of airway pressure to baseline (fig. 2)

.

The integration of total pressure fluctuation from baseline pressure throughout both inspiration and expiration (AREA-T) per breath was calculated as follows:

Three consecutive breaths were analyzed for each experimental setting after stabilization. All values were reported as mean ± SD. Dependent variables evaluated were D^T, P^T, PTP^T, PTP^I, VPF, V^T, D^E, P^Emax, dP^AW, PTP^E, and AREA-T. For comparing the effect of ventilators (each trigger mode was considered a different ventilator in this comparison), one-way analysis of variance was performed. For comparing the effect of flow demand, two-way analysis of variance was performed for each ventilator and flow demand. For comparing the effect of triggering mode, two-way analysis of variance was performed for each ventilator and triggering method. Post hoc  analysis was performed with the Scheffé test if analysis of variance reached significance. Significance was set at P < 0.05.

Representative airway pressure waveforms for each ventilator at a peak lung model inspiratory flow of 80 l/min during pressure and flow triggering are shown in figure 3. None of the ventilators mistriggered, nor did any ventilator prematurely cycle to expiration. To simplify the comparison among ventilators, the flow-triggered S300 ventilator was used as a statistical reference for all parameters except dP^AWand P^Emax, for which the PB840 ventilator was used as a reference.

Inspiratory trigger delay time during flow-triggered BEAR, pressure-triggered EVITA, and flow- and pressure-triggered GALILEO were significantly longer than that of flow-triggered S300 (fig. 4A). P^Tof the BEAR and GALILEO (both flow and pressure triggering) was significantly larger than that of flow-triggered S300 (fig. 4B). The PTP^Tduring flow-triggered BEAR, pressure-triggered EVITA, and flow- and pressure-triggered GALILEO was significantly larger than that of flow-triggered S300 (fig. 4C). The PTP^Iof the BEAR (both flow and pressure triggering) and pressure-triggered EVITA were negative and significantly smaller than that of flow-triggered S300 (fig. 4D). A negative PTP^Iindicates that overall effort was performed by the lung model during the inspiratory phase. A positive PTP^Iindicates overall effort was performed by the ventilator. There was no statistical difference in VPF and V^Tamong ventilators (table 2).

Expiratory delay time during pressure-triggered EVITA and PB840 (both flow and pressure triggering) were significantly shorter than that of flow-triggered S300 (fig. 5A). On the other hand, D^Eduring flow-triggered GALILEO was significantly longer than that of flow-triggered S300 (fig. 5A). The PTP^Eof the GALILEO (both flow and pressure triggering) and flow-triggered T BIRD were significantly larger than that of flow-triggered S300 (fig. 5B). The dP^AWand P^Emaxof flow-triggered S300 was not different from other ventilators. However, compared with flow-triggered PB840, dP^AWduring pressure-triggered BEAR and GALILEO (both flow and pressure) were significantly larger (fig. 5C), and P^Emaxof the GALILEO (both flow and pressure) was significantly larger. The integration of total pressure fluctuation from P^BL(AREA-T) of the GALILEO (both flow and pressure triggering) and flow-triggered T BIRD were significantly larger than that of flow-triggered S300 (fig. 5D).

The effect of varying peak lung model flow on all evaluated variables in each ventilator is shown in tables 2 and 3. VPF, V^T, P^Emax, and dP^AWincreased significantly as flow demand increased in all ventilators. There was almost no significant effect of varying flow on D^Tand D^Ein any ventilator. P^Tand PTP^Tof BEAR, EVITA, and GALILEO increased significantly as peak flow increased. PTP^Iof EVITA, GALILEO, PB740, and S300 decreased significantly as peak flow increased. PTP^Eand AREA-T of T BIRD increased significantly as peak flow increased.

The effect of triggering method on all evaluated variables in each ventilator with both triggers is shown in figure 6. There was an interaction between ventilator and triggering method for D^T, PTP^E, and AREA-T. In the BEAR, post hoc  analysis revealed that PTP^Eand AREA-T of flow trigger were significantly smaller than those of pressure trigger (1.59 ± 0.61 vs.  3.64 ± 0.99 and 2.73 ± 0.64 vs.  4.81 ± 0.90 cm H²O · s;fig. 5). There was no interaction between ventilator and triggering method for P^T, PTP^T, PTP^I, VPF, V^T, D^E, P^Emax, and dP^AW. Between flow trigger and pressure trigger across all ventilators, only P^Tshowed a significant difference (3.74 ± 1.89, 4.48 ± 1.67 cm H²O;fig. 6).

The major findings of this study can be summarized as follows: (1) there is a large range of variability among the seven ventilators evaluated during CPAP; (2) trigger pressure during pressure and flow triggering was the only variable that differed among all ventilators; (3) most of the ventilators perform work in the form of low-level pressure support during CPAP; (4) all of the ventilators impose an expiratory work load; and (5) increasing peak inspiratory demand decreased the ability of all ventilators to maintain a constant airway pressure.

Continuous positive airway pressure by definition is the maintenance of airway pressure at a constant level above atmospheric during spontaneous breathing. 30The application of CPAP has been found useful in many clinical settings: management of cardiogenic pulmonary edema, 31treatment of atelectasis, 32maintenance of functional residual capacity during endotracheal intubation, 33reduction of the inspiratory load of auto positive end-expiratory pressure caused by dynamic airway obstruction, 34and during weaning of patients from ventilatory support. 14,15The use of the CPAP mode on the mechanical ventilator during spontaneous breathing trials avoids the set-up of additional equipment and enhances patient safety by monitoring and alarming respiratory rate and tidal volume. However, concern regarding the ability of ventilators to provide CPAP without imposing or performing work on patients has existed since CPAP was introduced as an adjunct to spontaneous breathing.

Although our data indicate that the performance of this newest generation of ventilators during CPAP is better than previous generations, 5,6,8,13these ventilators continue to impose a minor level of inspiratory effort and an expiratory workload. In addition, most of them provide a low level of inspiratory pressure support, performing some inspiratory work. Furthermore, the difference between flow and pressure triggering in this generation is much less than that described in previous generations of mechanical ventilators. 25,26 

Triggering of demand flow during CPAP is dependent on the rate at which the ventilator samples airway pressure or flow, the ventilator's mechanical response time, including microprocessor speed, and the type of triggering mechanism used. 11Based on these factors, it is reasonable to expect flow triggering in CPAP to be more efficient than pressure triggering because by design, on most ventilators, the demand valve must remain open to provide a base flow during flow triggering. An open demand valve should reduce the time for mechanical response of the ventilator in flow triggering. However, we did not observe a trigger time delay or an increase in trigger PTP when flow and pressure triggering were compared in any specific ventilator. We did observe an increase in trigger pressure, although small (3.74 ± 1.89 vs.  4.48 ± 1.67 cm H²O;P < 0.01), during pressure triggering in the four ventilators with both pressure and flow triggering. Considering the magnitude of this difference and the fact that no other variables evaluated differed significantly, we question if this difference is clinically relevant. Clearly, the difference we observed is much less than the difference observed by Sassoon and Gruer 11using a similar lung model with the Nellcor Puritan-Bennett 7200ae (Nellcor Puritan-Bennett Corp.) or differences observed by others in patients. 24,25 

We were surprised at the level of variability among ventilators in those variables that define inspiratory trigger: D^T, P^T, and PTP^T. Ideally, each of these variables should be zero. This is probably unachievable because of mechanical limitations, but the closer to zero the variables are, the more efficient the ventilator is in triggering demand flow. Without proprietary knowledge of the operation of each ventilator, it is impossible to specify why these differences existed among the ventilators. However, our impression is that differences are primarily a result of the mechanical response and the algorithm used to deliver flow during CPAP, since microprocessor time is assumed to be very short.

An ideal ventilator should deliver CPAP without performing inspiratory work or imposing expiratory resistive work. Only the BEAR and EVITA did not perform inspiratory work, but these two ventilators have a negative inspiratory PTP. In contrast, the GALILEO, PB740, PB840, S300, and T BIRD provided 0.5–2.0 cm H²O pressure support during CPAP. However, when the integration of total pressure fluctuation from baseline pressure (AREA-T) and the magnitude of the total airway pressure swing (dP^AW) are compared (we considered these factors to be an index of the ventilators’ ability to maintain the target pressure), the BEAR and EVITA ventilators perform similar to all others except the GALILEO and T BIRD, which demonstrated the large PTP^E.

Ventilators imposing small efforts may be less likely to result in false-positive spontaneous breathing trials, whereas ventilators performing low levels of inspiratory work on patients may be better tolerated for long-term CPAP. However, Esteban et al.  35demonstrated that readiness for ventilator discontinuance could be equally determined with T-piece trials or low-level pressure support spontaneous breathing trials. It is still unclear which of these characteristics is preferable, and additional clinical trials would be needed to provide the answer.

The most unexpected results of this study were the long D^E, the high P^Emax, and the impedance to exhalation. These factors produced the high PTP^Eand AREA-T observed in some ventilators. It can be argued that our inappropriate setting of sensitivity affected the inspiratory trigger variable. However, these expiratory variables during CPAP were only affected by the operation of the exhalation valve governed by the algorithm set by the manufacturer. The high P^Emaxobserved on many of these ventilators may partially be explained by the lung model square wave flow pattern. As a result, the ventilators end inspiratory flow was high, and a transient high-pressure spike was caused by the recoil of the lung model and the high flow from the ventilator. However, as shown in figure 2and table 3, P^Emaxdiffered greatly among ventilators. PTP^Eis primarily a result of the resistance of flow through the exhalation valve. Pressure spikes during both triggering and cycling of inspiration may not be observed to the same magnitude in patients because of the location of pressure measurement in many ventilators. Most ventilators measure pressure internally on the inspiratory or expiratory limb. As a result, pressure spikes are dampened. These results raise concerns regarding the expiratory retard and expiratory workload imposed by some of these ventilators, which could be clinically significant in patients with chronic obstructive pulmonary disease.

As expected, peak flow rate and tidal volume increased as peak demand flow of the lung model increased. In some ventilators, trigger and inspiratory variables such as P^T, PTP^T, and PTP^Iwere affected by peak flow rate. Increasing inspiratory peak flow reduced the ability of CPAP to maintain a constant airway pressure. However, peak flow had less of an impact on these ventilators than that observed in previous-generation mechanical ventilators. 32Interestingly, D^Twas not affected in any ventilator. This indicates that once the trigger signal is identified, these ventilators are able to respond in equivalent time regardless of the peak flow. P^Emaxwas affected by the peak flow. This was probably a result of the larger tidal volume with higher peak flow rate. However, on most ventilators, expiratory variables such as D^Eand PTP^Ewere not affected by the demand flow rate.

In some critically ill patients, inspiratory demand may exceed that evaluated in this study. In fact, some critically ill hypoxemic patients may require peak inspiratory flow much greater than 100 l/min. However, we were unable to evaluate the performance of these ventilators at these flows because of the limitations of the lung model. Maximum peak flow generated by the model was 86 l/min. We also attempted to evaluate high inspiratory demand using a Michigan Instrument TLL (Grand Rapids, Michigan) lung model drive by a PB840, but could only generate 88 l/min peak inspiratory demand. As a result we can only rely on the trends observed as flow demand increased from 40 to 80 l/min to speculate on the effects of higher inspiratory demands. Specifically, performance of all ventilators can be expected to decrease at higher peak inspiratory flows. Trigger pressure and PTP^Ias well as P^Emax, dP^aw, and AREA-T can all be expected to be greater at higher peak inspiratory flows.

When these ventilators are compared with data on previous-generation intensive care unit ventilators, it is clear that the overall operation during CPAP has dramatically improved. 1,8,9,13Older-generation ventilators not only impose greater work loads, but also demonstrate marked differences between flow and pressure triggering favoring flow triggering during CPAP. Our data indicate that on most ventilators during CPAP, little difference between flow and pressure triggering exist, but flow triggering appears to perform slightly better.

Although these data were not collected on patients, the low levels of imposed effort favors the use of most of these ventilators for spontaneous breathing trials, but the low level of work performed may still be a concern. However, our data clearly indicate which ventilators impose and perform the most effort–work and provides direction for ventilator selection. Clinical trials are needed before any of these ventilators can be unequivocally recommended for use during spontaneous breathing trials. However, our impression is that the amount of work imposed or performed by these ventilators is clinically insignificant for most patients.

Controversy exists over the benefit of pressure versus  flow triggering during assisted ventilation. 19–25Some studies 19,21,22found no difference between the two, whereas others favored flow triggering. 23–25Methods used to set triggering and modes evaluated may partially explain these difference.

Our data on the Servo 300 is consistent with data from Aslanian et al.  25and Tütüncüet al.  22During pressure support in 16 adults, Tütüncüet al.  found no difference between pressure and flow triggering with the pressure trigger set at 1.0 cm H²O and the flow trigger at 0.7– 2.0 l/min. During a lung model evaluation, Aslanian et al.  also found little difference during pressure support when the flow trigger was set at 1 l/min and pressure trigger at 0.5 cm H²O, which in our experience are equivalent settings. 21However, during an evaluation on patients, Aslanian et al.  set flow triggering at 2 l/min and pressure triggering at 2 cm H²O. These results favored flow triggering. Ideally, both should be set as we set sensitivity, at the most sensitive setting not causing self triggering. Unfortunately, we did not evaluate assisted ventilation. However, considering the differences between pressure and flow triggering during CPAP on older-generation ventilators and the improvement observed on the newer ventilators, we would expect less response difference during assisted ventilation than that observed in some comparisons. 23–25 

The most important limitation of this study is the fact that it was performed on a lung model instead of patients. It is possible that performance on patients may differ greatly from that demonstrated here. Specifically, the “pleural space” of our model is always open to atmosphere, unlike the normal human pleural cavity. However, this model allows the determination of the onset of inspiration by deflection in pleural pressure. Peak inspiratory flows of only 40–80 l/min were evaluated on all ventilators because of the limitations of the lung model. In critically ill hypoxemic patients, peak inspiratory flows may greatly exceed those evaluated. It is thus difficult to predict the performance of these ventilators outside the flow range studied. Water in the ventilator circuit may cause autotriggering. The fact that we did not operate the ventilator humidifier may have affected ventilator response. In addition, only one level of CPAP in a single ventilator of each type was evaluated. Our pilot data indicated that CPAP level did not affect the evaluated variables, but we cannot comment on the variability in ventilators from a single manufacturer. We also did not evaluate the effect of pressure and flow triggering during assisted ventilation or the impact of auto–positive end-expiratory pressure on the response of these ventilators.

In conclusion, this generation of mechanical ventilators demonstrates a large range of variability during CPAP. However, very little difference exists between pressure and flow triggering. Most of the ventilators evaluated provide low levels of inspiratory pressure support, and all ventilators impose an expiratory work load. Increasing inspiratory peak flow does reduce the ability of CPAP to maintain a constant airway pressure.

All ventilators were specifically provided by the manufacturer for evaluation (Bear Medical Systems Inc., Riverside, CA; Drager Inc., Telford, PA; Nelcor Puritan-Bennett Corp., Carlsbad, CA; Bird Products Corp., Palm Springs, CA; Hamilton Medical AG, Rhazuns, Switzerland; Seimens-Elema AB, Solna, Sweden).