INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: acute respiratory failure, mechanical ventilation, noninvasive ventilation, pressure support ventilation, proportional assist ventilation

Noninvasive positive pressure ventilation (NPPV) has gained increasing acceptance in recent years as a way to avoid endotracheal intubation and improve outcomes in selected patients with acute respiratory insufficiency (1, 2). Compared with conventional therapy, including intubation, NPPV can reduce intensive care unit (ICU) lengths of stay and lower complication and mortality rates in selected patients with chronic obstructive pulmonary disease (COPD) exacerbations (3) for whom it is now considered the ventilatory mode of first choice (8). Although enhanced patient compliance (9) and comfort (10) have been observed when NPPV was applied using pressure support ventilation (PSV) as compared with volume-limited ventilation, reported failure rates are similar for the various techniques, ranging from 13 to 42% (3). Most failures have been related to patient intolerance or inability to improve ventilation, and some may be associated with patient- ventilator asynchrony (11).

Proportional assist ventilation (PAV) was developed as a mode to enhance ventilator responsiveness to patient breathing effort (12, 13). Like PSV, PAV uses a sensitive inspiratory trigger. However, unlike PSV that uses a clinician preset inspiratory pressure, PAV provides inspiratory flow and pressure in proportion to the patient's spontaneous breathing effort as determined by instantaneous feedback from an in-line pneumotachometer (13). In addition, cycling from inspiration to expiration is not dependent on a predetermined reduction in inspiratory flow as is the case with PSV. Rather, PAV, when properly adjusted, terminates delivery of inspiratory assistance with cessation of inspiratory effort (14). In this way, PAV has the potential of enhancing patient-ventilator synchrony and comfort in comparison to PSV.

Recent studies have demonstrated the potential utility of PAV as a mode for delivery of NPPV. Short-term PAV improves gas exchange in patients with chronic respiratory failure resulting from restrictive thoracic disease or COPD (15). In patients with chronic respiratory failure, PAV reduces work of breathing as assessed by the pressure-time product (16). More recently, PAV has been shown to unload inspiratory muscles, improve gas exchange, and provide excellent patient-ventilator synchrony in patients with severe, stable COPD (17). In the only study yet reported on PAV delivered noninvasively to patients with acute respiratory failure, Patrick and coworkers (18) improved gas exchange and avoided intubation in eight of 11 patients. Although potential advantages of PAV over PSV have been demonstrated in invasively ventilated patients (19, 20), no controlled comparison studies between the two modes have yet been reported in patients with acute respiratory insufficiency treated noninvasively.

Different techniques for initiation of PAV have also been used. As originally described by Younes (12, 13), PAV settings were to be adjusted to match measured values for resistance and elastance of the respiratory system, but these measurements are impractical in patients ventilated noninvasively. An alternative technique is referred to as the "runaway method" (17) and consists of increasing the elastance and resistance settings separately. When these values exceed the actual ones, pressure and flow continue to increase or "runaway," even after the patient has begun to exhale. The elastance and resistance settings are then adjusted downward. A third method is to adjust PAV settings to patient comfort (18), increasing elastance and resistance settings until the patient's respiratory distress is alleviated, the respiratory rate declines, or signs of excess pressure delivery occur.

In the present study, we performed a pilot study to test the feasibility of treating acute respiratory insufficiency with PAV delivered noninvasively using a pragmatic, bedside initiation technique. The study was performed under the auspices of the U.S. Food and Drug Administration (FDA) which required a comparison with the Puritan-Bennett 7200ae as the predicate device to deliver PSV. Accordingly, we hypothesized that PAV and PSV administered using the Puritan-Bennett 7200ae would similarly improve gas exchange and avoid intubation. We also anticipated that mortality and complications associated with NPPV would occur at similar rates for PAV and PSV. Finally, by virtue of its enhanced ability to match patient breathing effort, we hypothesized that subjectively, patients would find PAV to be more acceptable and comfortable than the PSV mode on the 7200ae.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was conducted from June 1997 to April 1998 at three academic medical centers: Rhode Island Hospital, Providence, Rhode Island; Mayo Clinic, Rochester, Minnesota; and Massachusetts General Hospital, Boston, Massachusetts. The institutional review boards at each center approved the protocol, and written informed consent was obtained from all enrolled subjects or their next of kin.

Patients

Patients were offered enrollment into the study if they had acute respiratory insufficiency with worsening dyspnea and at least one of the following: (1) hypercapnia (Pa_CO2 > 45 mm Hg) with pH < 7.35 but > 7.10; (2) hypoxemia with Pa_O2 < 55 mm Hg on room air or a ratio of arterial oxygen pressure to fraction of inspired oxygen (Pa_O2:FI_O2)  250; (3) normal arterial blood gases with a respiratory rate > 25/min.

Patients were excluded if they met any of the following criteria: age < 18 yr; need for immediate intubation; pH < 7.10; inability to protect the upper airway; medical instability; agitation; severe underlying illness likely to be terminal within 3 mo; positive pregnancy test; inability to give informed consent; enrollment in another research project.

Study Design

Enrolled patients were randomized to receive either PAV delivered using the BiPAP Vision Ventilator (Respironics, Inc., Pittsburgh, PA) or PSV delivered using the 7200ae (Puritan-Bennett, Carlsbad, CA). The study was done as a pilot required by the FDA that specified the 7200ae as the current, standard device for delivery of NPPV in patients with acute respiratory insufficiency. Randomization was achieved using sealed envelopes.

Initiation of Assisted Ventilation

Patients were first fitted with a standard nasal mask (Contour Nasal Mask, Respironics, Inc). PAV settings included flow assist (FA), the integrated flow signal (volume assist [VA]), and a proportional adjustment of both signals (proportional assist). Standard initial settings (FA = 2 cm H₂O/L/s, VA = 5 cm H₂O/L, proportional assist = 100%, and end-expiratory pressure of 4 cm H₂O) were used for all patients receiving PAV therapy. PSV was begun at an inspiratory pressure of 10 cm H₂O and an expiratory pressure of 4 cm H₂O (pressure support = 6 cm H₂O). With both modes, settings were gradually adjusted upward as tolerated, FI_O2 was adjusted to maintain O₂ saturation  88%, and humidification was not routinely used. Aerosolized bronchodilator therapy was delivered by temporarily interrupting ventilatory assistance and using metered-dose inhalers with spacers or standard nebulizers.

Outcome Measures

The main outcome variables were avoidance of intubation and mortality. Successful therapy was defined as tolerance of therapy and avoidance of intubation with sustained improvements in vital signs and dyspnea. Failure occurred if the patient was intolerant of therapy despite repeated attempts at gaining acceptance, required intubation, or died. Secondary outcome measures included changes in serial vital signs, the intensity of accessory muscle use, arterial blood gases 1 to 2 h postinitiation, subjective responses including dyspnea and mask and air pressure comfort, the frequency of ventilator adjustments and encouragement, refusal of therapy, ICU and hospital lengths of stay, and complications.

Statistical Analysis

The significance of differences between the main outcome variables was determined by a two-sample test of equality of proportions, and by proportional hazards regression using Kaplan-Meier survival analyses. For secondary variables assessed by serial measures, analysis of variance for repeated measures was used to determine the significance of differences, with Tukey's exact test applied when F ratios indicated significance. Data are presented as mean ± SE, and differences were considered significant when p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Baseline Variables

Forty-four patients (21 PAV and 23 PSV) were enrolled in the study. As shown in Table 1, demographic and baseline physiologic and gas exchange variables did not differ significantly between the two groups at the time of enrollment, with the exception of pH, which was slightly lower in the PAV group. Further, the distribution of diagnoses was similar between the two groups (Table 2), although the PAV group tended to have fewer patients with COPD and more with pneumonia.

Major Outcome Variables

Kaplan-Meier curves for intubations (Figure 1, upper panel) and mortality (Figure 1, lower panel) were virtually identical between the PAV and PSV groups. A total of 11 patients required intubation. Two patients, one in each group, never received NPPV because they deteriorated rapidly after randomization and required prompt intubation. Two other patients in the PAV group and one in the PSV group were intubated after more than 6 h of ventilator use. Reasons for intubation included failure to improve gas exchange and progressive fatigue in the PAV patients, and excessive secretions in the PSV patient. Six additional patients (4 PSV, 2 PAV) were subsequently intubated during the hospitalization; between 1 and 5 d for the PSV and 13 h and 1 d after completion of the study for the PAV patients.

View larger version (13K): [in this window] [in a new window]   Figure 1.   Kaplan-Meier curves for occurrence of intubation for the first 6 d (upper panel ) and for mortality during the entire hospitalization. No intubations occurred after the first 6 d. No significant differences were observed.

One patient from the PAV group and none in the PSV group died during the period of ventilator use. The patient who died had initially improved on therapy but after 10 h of use, requested that it be withdrawn. His gas exchange and mental status subsequently deteriorated and because he had declined intubation, PAV was briefly resumed without success. After completion of the protocol, three patients in the PAV group and seven in the PSV group subsequently died during the hospitalization, including three of the intubated patients in the PSV group. Four had declined further treatment and elected a "do not resuscitate" status. Primary causes of death were cardiac complications in three (1 PAV, 2 PSV), progressive respiratory failure in three (1 PAV, 2 PSV), and multiple chronic processes in the other four patients (1 PAV, 3 PSV).

Secondary Outcome Variables

Vital signs and accessory muscle use. Although heart rate tended to drop slightly more in the PSV group than in the PAV group during the first 8 h of use (Figure 2, upper panel), the difference was not statistically significant (p = 0.17). Accessory muscle use also diminished at similar rates in the two groups (Figure 2, middle panel). On the other hand, the lower panel of Figure 2 shows that respiratory rate dropped more in the PAV than in the PSV group (p = 0.02). Serial measures were not analyzed beyond the first 6 h because nearly half of the patients had completed the protocol.

View larger version (19K): [in this window] [in a new window]   Figure 2.   Heart rate (HR) (upper panel ), accessory muscle activity (Access Musc Use) score (middle panel  ), and respiratory rate (RR) (lower panel ) during the first 360 min of therapy are shown. Data are mean ± SE. *p < 0.05, PSV versus PAV group.

Gas exchange. Figure 3 shows mean Pa_CO2 (upper panel) and Pa_O2/FI_O2 (lower panel) values at baseline, 30 min, and 2 h; analysis beyond 2 h was not feasible because of few data. Pa_CO2 fell more during the first 30 min of therapy in the PAV group than in the PSV group (6.7 versus + 0.4 mm Hg, respectively [p < 0.05]), although the absolute mean values did not differ between the groups at any time point. With regard to oxygenation, the Pa_O2/FI_O2 ratios in the two groups were similar throughout the first 2 h (Figure 3, lower panel).

View larger version (18K): [in this window] [in a new window]   Figure 3.   Mean values for PaCO2 (upper panel  ) and PaO2/FIO2 (lower panel ) during the first 2 h of therapy. Data are mean ± SE.

Subjective indices. Using a visual analog scale (1 least-10 most), patients in both groups registered dramatic relief in their sensation of dyspnea within 30 min of initiation, and there were no significant differences between groups (Figure 4, upper panel). Using the same analog scale, patients rated their level of comfort with air pressure as similar between the two groups, but they rated their comfort with the mask as significantly higher with PAV than with PSV (Figure 4, lower panel). Although all patients initially used nasal masks, excessive machine alarming associated with air leaks through the mouth necessitated switching to an oronasal mask in 15 of the 22 patients in the PSV group. Significantly fewer (5 of 20) patients in the PAV group required switching to a full-face mask (p = 0.005).

View larger version (19K): [in this window] [in a new window]   Figure 4.   Mean dyspnea analog scale (upper panel ) and mask comfort analog scale (lower panel ) values during the first 360 min of therapy. Visual analog scale is 1-10 with 10 being the most. Data are mean ± SE. *p < 0.05, PAV versus PSV.

Inspiratory and expiratory pressures and tidal volumes. Peak inspiratory pressures were significantly higher in the PSV group than in the PAV group after the first 30 min of ventilation, and expiratory pressures were slightly lower (Table 3). Inspiratory pressures were gradually titrated upward as tolerated in the PSV group and after 2 h of ventilation, the mean value was significantly higher than after 15 min. In the PAV group, peak inspiratory pressure remained steady for the first 2 h. As a consequence, at 2 h, the difference between the mean inspiratory and expiratory pressures was significantly greater in the PSV group than in the PAV group (8.8 ± 1.1 cm H₂O for PSV and 6.1 ± 1.2 cm H₂O for PAV [p < 0.05]). Despite this difference, estimated tidal volumes were similar in the two groups (Table 3). Associated with the increase in peak inspiratory pressure, mean tidal volumes in the PSV group were significantly greater after 2 h than after 15 min of assisted ventilation. Settings for FA, VA, and proportional assist did not change significantly during the first 2 h (Table 3).

Duration and tolerance of ventilation. Duration of ventilator use was similar in both groups (Table 4). Humidifiers were used by only one patient in each group. Complications occurred more frequently in the PSV group than in the PAV group, consisting mainly of nasal bridge ulcerations. The PAV mode was well tolerated and treatment refusals were more numerous in the PSV group than in the PAV group (Table 4). Ventilator adjustments in patients receiving PSV and PAV were similar in frequency (data not shown), but patients receiving PSV needed significantly more frequent encouragement (p = 0.04) to continue therapy than patients in the PAV group (Figure 5). Only five patients, four with COPD, encountered runaway, three after 2 h, two after 6 h, and none during initiation. This was easily managed by reducing the FA setting by 1 or 2 cm H₂O/L/s. Lengths of ICU and hospital stays were similar in the two groups (Table 4). In addition, after use of analysis of covariance to correct for pH, all statistically significant differences remained, including vitals signs, gas exchange, and subjective indices.

View larger version (14K): [in this window] [in a new window]   Figure 5.   Number of episodes of encouragement per time epoch during the first 360 min of therapy. PSV patients received significantly more frequent episodes of encouragement than PAV patients (p < 0.05).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We found that patients with acute respiratory insufficiency of various origins have similar intubation and mortality rates when treated with PAV as compared with PSV delivered using the Puritan-Bennett 7200ae. In addition, PAV supported gas exchange as well as PSV while achieving higher initial estimated tidal volumes with lower respiratory rates and peak inspiratory pressures. Further, administration of PAV using a practical application technique that relied on simple clinical indices of tolerance was well accepted by patients.

Previous studies have shown PSV to be an effective mode for delivery of noninvasive ventilation, improving gas exchange and avoiding intubation and its attendant complications in randomized controlled trials when compared with conventional therapy (3, 5). However, PSV has potential shortcomings as a mode to assist ventilation in acutely distressed patients. First, PSV may have difficulty responding to the onset of expiration in patients with COPD because of high inspiratory flow rates that are sustained for the brief duration of inspiration (23). This problem may be compounded during air leaking that is almost ubiquitous with NPPV, because the ventilator sustains a high inspiratory flow to compensate for the leak (24, 25). In addition, PSV at high inspiratory pressures may contribute to patient-ventilator asynchrony because the high tidal volume may suppress inspiratory effort and contribute to self-controlled positive end-expiratory pressure (auto-PEEP), causing a failure to trigger during subsequent breaths (26). Further, PSV cannot automatically adjust the level of assistance to changing patient ventilatory demands.

PAV offers the potential to circumvent many of these problems. First, PAV targets inspiratory flow as a surrogate of effort and reduces the problem of failure to sense the onset of expiration (13). In addition, PAV has been shown to effectively reduce the pressure-time product of the inspiratory muscles in patients with COPD (19), an effect that is potentiated by combination with positive end-expiratory pressure (PEEP) (27). Patient-ventilator synchrony is also enhanced relative to high levels of pressure support because patient effort determines the level of assistance and excessive inspiratory pressures are avoided (26). In addition, PAV is able to adjust to changing patient ventilatory demands (20, 28), relying less on operator vigilance to make needed adjustments.

Considering its potential advantages, we performed this feasibility study to determine how PAV performs as a mode to deliver NPPV in comparison to PSV, presently the consensus mode of choice for this application (8). The Vision ventilator provides PSV in the form of Bilevel Positive Airway Pressure (BiPAP), and comparison of PAV and PSV using the same ventilator would have been preferable to the use of two different ventilators, as was done in the current study. This would have permitted control for variability related to differences between devices and blinding of some of the caregivers. However, the U.S. FDA, under whose auspices the study was performed, required use of the Puritan-Bennett 7200ae as the standard device to deliver PSV, based on prior published experience with NPPV (29).

Patients randomized to receive PAV in our study were comparable to those in the PSV group in almost all respects, except that the pH was slightly and statistically significantly lower. Also, there were trends for slightly higher Pa_CO2 values, lower Pa_O2:FI_O2 ratios, fewer patients with COPD, and more patients with pneumonia in the PAV group. These differences suggest that patients randomized to receive PAV were slightly more acutely ill, at least with respect to blood gases, than those in the PSV group, and might have been expected to have worse outcomes (30, 31). However, our intent-to-treat analysis showed that intubation rates were very similar for PAV and PSV, respectively, both during study enrollment (14 versus 9%) and during the entire hospitalization (19 versus 22%). Further, correcting for the difference in pH did not change any of the results of the statistical analyses.

Although many of the secondary outcome variables, including heart rate, changes in Pa_CO2, oxygenation, accessory muscle use, sensation of dyspnea, and lengths of stay in the ICU and in the hospital were similar between the groups, a number of differences deserve further comment. First, respiratory rate, an important predictor of NPPV success (30) as well as an important determinant of work of breathing per unit time (32), was lower in the PAV group. In a prior study comparing PAV with PSV, intubated patients made acutely hypercapnic by adding dead space had smaller increases in respiratory rate, larger tidal volumes, lower esophageal pressure- time products, and less dyspnea when using PAV than when using PSV (20). The investigators surmised that by virtue of its ability to track changes in breathing pattern, PAV responded better than PSV to the increased demand posed by the hypercapnia. Likewise, the lower respiratory rate in the PAV group in our study may reflect the rapid response of PAV to the high demand for ventilatory assistance when patients first present with acute respiratory distress. The lower respiratory rate also suggests that PAV may have lowered work of breathing more rapidly than with PSV. However, lacking esophageal pressure measurements, we are unable to confirm this possibility. Also, the similarities in dyspnea scores and accessory muscle use suggest that any differences in work of breathing were small.

Other differences in favor of the PAV group included higher mask comfort ratings, fewer complications in the form of nasal bridge ulcerations, and fewer refusals of therapy, consistent with improved tolerance of the PAV mode. These differences were associated with lower peak inspiratory pressures with PAV, and greater use of oronasal masks with PSV. Lower peak airway pressures with PAV than PSV have previously been reported, presumably related to air pressure delivery that more closely matches the patient's spontaneous inspiratory flow pattern (26). The lower peak inspiratory pressure could have enhanced mask comfort by reducing the strap tension necessary to control air leaks. On the other hand, the differences in comfort, tolerance, and complications could also be related to the differences in nasal versus oronasal mask use between the ventilators. Air leaking through the mouth triggers alarms and interferes with ventilator triggering when nasal masks are used to deliver NPPV via the 7200ae, problems that are less often encountered with the Vision ventilator. Accordingly, the lower mask comfort ratings and higher complication and refusal rates for the PSV compared with the PAV group could have been related to the operating characteristics of the 7200ae ventilator that favored switching to an oronasal mask to help reduce interface-related air leaking. It remains possible that PSV modes on other ventilators that are more leak-tolerant and less likely to alarm may have been better tolerated.

Another potential advantage associated with use of PAV in the present study was the need for less encouragement to continue than with the PSV mode. Although this difference may have been related to the high proportion of patients in the PSV group who experienced intolerance of the nasal mask, the technique used for initiation of PAV in our study achieved patient comfort quickly and required little additional manipulation of ventilator settings. Using relatively low initial settings for volume and flow assist and gradually adjusting upward while monitoring relief of dyspnea and air pressure-related discomfort was an approach well tolerated by patients. In fact, increasing assist levels to the point where "runaway" occurred was unnecessary during initiation. "Runaway" did occur subsequently in a minority of our patients, mainly with COPD, probably related to improvements in lung mechanics and responded promptly to reduced FA settings.

A number of limitations should be borne in mind when interpreting our results. As noted earlier, ours was designed as a pilot study and the number of patients is not sufficient to eliminate the possibility of beta error. However, given the intubation rates during the study of 3/21 (14%) for PAV and 2/23 (9%) for PSV, we would have required 638 patients in each arm (total 1,276) to have a statistical power of 80% (alpha = 0.05) to detect the aforementioned difference with a two-sided test of significance. Also, the number of secondary hypotheses was large, introducing the possibility that some statistically significant differences occurred on the basis of chance. Further, our study was not blinded, raising concerns about investigator or clinician bias, and the use of different ventilators raises the possibility that differences in ventilator characteristics influenced the results, as previously discussed.

Despite these limitations, ours is the first randomized, prospectively designed study to demonstrate the feasibility of using PAV to deliver NPPV, administered using a practical initiation strategy based on simple clinical indices. Further, PAV compares favorably with PSV administered using the 7200ae ventilator with regard to the need for intubation and support of gas exchange. Finally, PAV may offer physiologic and subjective benefits over PSV, but these latter potential benefits should be confirmed in larger prospective trials using the same ventilator.