INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Despite a number of advances in the techniques and application of conventional mechanical ventilation, acute respiratory distress syndrome (ARDS) remains a significant problem in critical-care units. It has a multifactorial etiology, complex pathophysiology, and significant morbidity and mortality. Several computed tomographic (CT) scan studies have demonstrated that the lungs of mechanically ventilated patients with ARDS are not uniformly affected by this condition (1). Dense lesions tend to concentrate in the dependent part of the lung, with gas ventilation being directed predominantly to the nondependent regions. Repeated opening and closing of alveoli in the dependent region during mechanical ventilation has been suggested as an important contributor to maintaining and aggravating the lung injury in ARDS (4). Current ventilation practices therefore aim at minimizing baro-/volutrauma by limiting peak airway pressures, using positive end-expiratory pressure (PEEP) to maintain an open lung, and limiting tidal volumes. Many animal studies have demonstrated that this ventilation approach minimizes lung injury with acceptable gas exchange (7). Recruitment of the atelectatic dependent zone is obviously important in the evolution and outcome of acute lung injury. Are there other ways of recruiting this region and keeping it open?

Since Clark and Gollan (10) first demonstrated that mammals could survive submersion in perfluorocarbons (PFCs) equilibrated with oxygen, numerous studies have shown that liquid ventilation is a feasible method of obtaining adequate gas exchange in both the normal and the injured lung (11). The liquid would act as a "bottled PEEP," recruiting atelectatic regions and maintaining them in this state, as well as reducing the surface tension associated with surfactant deficiency and dysfunction seen in various forms of lung injury (14). Initial interest focused on total (tidal) liquid ventilation. Researchers demonstrated beneficial changes in lung mechanics and gas exchange with minimal effects on pulmonary vascular resistance and cardiac output in a lung filled with PFC in different models of lung injury (15). However, liquid ventilators were complicated and cumbersome.

In 1991, Fuhrman and colleagues (18) described a simpler approach to liquid breathing: perfluorocarbon-associated gas exchange (PAGE), more recently called partial liquid ventilation (PLV), in which conventional mechanical gas ventilation is imposed on a lung filled to FRC with PFC. PLV has now been shown to result in improved lung mechanics and oxygenation in a variety of animal models of lung injury (19). What has not been clearly delineated, however, is the optimal ventilation strategy, or whether a full FRC of PFC is necessary for efficient gas exchange. Current ventilator strategies in PLV have tended to use larger tidal volumes with the acceptance of higher peak inspiratory pressures. This has been supported by recent work done by Fuhrman's group, which has demonstrated a correlation between VT and oxygenation (24).

The encouraging results of the aforementioned work stimulated initial Phase I trials in humans, in which similar improvements in oxygenation and pulmonary mechanics were found. This has led to a U.S. Food and Drug Administration (FDA)-approved, industry sponsored Phase II/III trial of PLV in both adult and pediatric populations.

Two recent reports of the use of PLV in association with extracorporeal life support in humans have described improvements in compliance and oxygenation with a decrease in extracorporeal pump flow (25, 26). However, both reports also described a significant incidence of air leak during the period of PLV. It is unclear whether this was a result of instilling the large volume of liquid into the lung and then ventilating with a relatively large VT, or was secondary to the underlying lung disease. Given the current trend of ventilating with small tidal volumes at a higher PEEP, we set out to investigate the relative contributions of VT and liquid volume to the incidence of barotrauma, as well as to determine whether adequate gas exchange could be achieved with alternative ventilation strategies. In addition, we investigated whether a lower dose of PFC would be sufficient to maintain adequate gas exchange.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our study was conducted after being approved by the animal care committee of the Research Institute at the Hospital for Sick Children, and according to guidelines of the National Institute for Health (USA).

Male adult New Zealand rabbits (2,200 to 3,800-g body weight) were premedicated with acepromazine (0.5 mg/kg intramuscularly) and anesthetized with sodium pentobarbital (10 to 20 mg/kg intravenously). After supine positioning of the animals, a tracheotomy was performed and a 3.5- or 4.0-mm endotracheal tube was inserted and secured to prevent a leak. Anesthesia was maintained with sodium pentobarbital (6 mg/kg/h), and paralysis was initially achieved with pancuronium bromide (0.2 mg/kg intravenously), followed by a continuous intravenous infusion of 0.2 mg/kg/h. Maintenance fluid consisting of 5% dextrose/0.45% saline was administered at 7.5 ml/kg/h. A peripheral arterial line was inserted for continuous blood pressure monitoring (Model 1280; Hewlett-Packard, Waltham, MA) and blood sampling for blood-gas measurements. A central venous catheter was inserted into the jugular vein. After tracheotomy, the animals were ventilated with a fixed VT of 10 ml/kg at a rate of 30 breaths/min, with PEEP maintained at 1 to 2 cm H₂O and an FI_O2 of 1.0 for a control period of 30 min.

Respiratory failure was then induced by repeated lung lavage with aliquots of 25 ml/kg of warmed normal saline. Lavage was considered to have been adequate if the Pa_O2 was reduced to below 60 mm Hg at 15 min after lavage with the following ventilator settings: VT = 10 ml/ kg, PEEP = 5 cm H₂O, rate = 30 breaths/min, and FI_O2 = 1.0. Animals were then ventilated for a period of 1 h at these settings to establish the lung injury. This lung model has been used previously and is considered a reliable model of severe RDS, with similar histologic and pathophysiologic changes (27).

After the lung injury was established, animals were randomized into four groups as follows:

Group 1. (Low dose, moderate VT, high PEEP). A dose of 6 ml/kg of warmed PFC (perfluorooctyl bromide; Alliance Pharmaceutical Corporation, San Diego, CA) was introduced through the endotracheal tube. VT was set at 9 ml/kg, with a rate adjusted to maintain the Pa_CO2 between 35 and 45 mm Hg. In this group, PEEP was fixed at 9 cm H₂O to maintain the animals for the duration of the experiment (animals died of hypoxia at a PEEP of 5 cm H₂O). FI_O2 was 1.0.

Group 2. (FRC dose, large VT, low PEEP). PFC was instilled in 6-ml aliquots until a meniscus was seen in the endotracheal tube, or to a maximum of 18 ml/kg. VT was adjusted to and maintained at 15 ml/ kg, with the PEEP set at 5 cm H₂O. The ventilator rate and FI_O2 were set as in Group 1.

Group 3. (FRC dose, moderate VT, low PEEP). PFC was instilled as in Group 2. VT was adjusted to and maintained at 9 ml/kg, with the PEEP set at 5 cm H₂O. The ventilator rate and FI_O2 were set as in Group 1.

Group 4. (Low dose, Large VT, low PEEP). A dose of 6 ml/kg of PFC was instilled as in Group 1. VT was adjusted to and maintained at 15 ml/kg. PEEP was 5 cm H₂O, and the ventilator rate and FI_O2 were as in Group 1.

All groups were ventilated for a period of 2 h following the instillation of PFC. Arterial blood-gas and pulmonary-mechanics measurements were made as illustrated in Table 1. Chest radiographs (posterior and cross-table lateral) were made immediately prior to instilling PFC and at the end of the experiment. Additionally, all animals were autopsied, with care being taken to open the chest cavity under water and with the lungs inflated to 15 cm H₂O to examine for the presence of an air leak.

Controlled mechanical ventilation was provided by a Humming V infant ventilator (Medtran, Oomia City, Japan). Arterial blood gases were intermittently and VT was continuously measured throughout all experiments. Blood gases were measured with an ABL-330 (Radiometer; Copenhagen, Denmark). VT was measured with a low-dead-space (1.3 ml) hot-wire pneumotachograph set in the inspiratory-volume mode (NVM-1; Bear Medical Systems, Riverside, CA) inserted between the endotracheal tube and the ventilator adapter. The flow signal was digitally integrated to volume. Airway pressures were recorded with a Validyne MP 45 pressure transducer (Validyne Engineering Corp., Northridge, CA) connected to the side port of the endotracheal tube adapter. Both flow and pressure signals were recorded with an on-line IBM computer (Armonk, NY), using analog-to-digital (AD) conversion at a sampling rate of 100 Hz (DT2801A converter; Data Translation Inc., Marlboro, MA) and the ANADAT/ LABDAT software package (McGill University, Montreal, Canada).

For dynamic pressure-volume loops, 10-s periods of flow and pressure data were collected. After digital integration of flow to volume, the dynamic compliance was computed, through linear regression, as the slope of the pressure-volume relationship. During each period of dynamic pressure-volume loop recording, the endotracheal tube was briefly occluded via a three-way rotary valve at end expiration, and then opened to the atmosphere to allow deflation to FRC. The measured expired volume from end-expiration (end-expiratory lung volume; EELV) allows one to set the dynamic PV loops on the volume axis above the FRC.

Statistical Analysis

Physiologic variables are expressed as mean ± SD, and the corresponding data were analyzed by analysis of variance (ANOVA). A significant difference was accepted at p < 0.05. Air-leak data were analyzed with multiple logistic regression. To determine the significance of the individual factors contributing to air leak (PFC volume and VT, respectively), the change in log-likelihood function was assessed by analysis of deviance (28, 29).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Thirty-five rabbits were studied (Table 2): seven in Group 1 (VT = 9 ml/kg, PFC = 6 ml/kg, PEEP = 9 cm H₂O), 13 in Group 2 (VT = 15 ml/kg, FRC filled with PFC, PEEP = 5 cm H₂O; five animals in this group died prior to completion of the experiment); seven in Group 3 (VT = 9 ml/kg, FRC filled with PFC, PEEP = 5 cm H₂O); and eight in Group 4 (VT = 15 ml/kg, PFC = 6 ml/kg, PEEP = 5 cm H₂O; one animal in this group died prior to completion of the experiment).

Physiologic data for the measured variables of animals surviving for the duration of the experiment are given in Table 1. Baseline and pretreatment values were comparable in all groups. Pa_O2 data are illustrated in Figure 1. Pa_O2 values in Groups 1 and 2 were similar. Significant differences in Pa_O2 between groups are listed in Table 1. Of particular note was that Pa_O2 in Group 4 tended to decrease throughout the duration of the experiment. Pa_CO2 was controlled and comparable in all groups, as were hemodynamic variables. VT and PEEP were controlled, and their effect on PImax is illustrated in Figure 2. Of interest was that dynamic compliance in Group 3 (VT = 9 ml/ kg, PEEP = 5 cm H₂O, FRC-filled) was significantly different from that in other groups at all time intervals after baseline (Figure 3). Static compliance varied as a function of PEEP.

View larger version (21K): [in this window] [in a new window]   Figure 1.   Arterial oxygenation over time. Note the initial significant improvement in PaO2 in all groups. With the exception of Group 4, this improved oxygenation was well maintained for the duration of the experiment. Differences between groups are given in Table 1. We suspect that the declining PaO2 in Group 4 was secondary to evaporation of PFC.

View larger version (17K): [in this window] [in a new window]   Figure 2.   Peak inspiratory pressure as a function of time. This is a controlled variable, as it is affected by both the set PEEP and the selected VT. All pressures significantly increased from baseline. Statistical differences between groups are given in Table 1.

View larger version (22K): [in this window] [in a new window]   Figure 3.   Dynamic compliance. Significant improvement was noted in all groups of animals from baseline. Group 3 was significantly different from other groups at all times after baseline. Compliance in Group 4 decreased over time, possibly because of evaporation of PFC.

Air-leak data are given in Table 2. Animals in Group 1 had neither radiologic nor autopsy evidence of air leak, and all survived the experiment. Of the 13 animals in Group 2, five died before completion of the experiment. Of these, three had postmortem chest radiographs, all of which demonstrated bilateral pneumothoraces. All five animals had autopsy evidence of air leak. Of the eight animals in Group 2 that survived for the duration of the experiment, two had radiologic evidence of pneumothoraces (Figure 4) and a further three had air leak found at autopsy. No animals in Group 3 died during the experiment, and none had evidence of air leak. Only one of the eight animals in Group 4 died (with both radiologic and autopsy evidence of air leak). Analysis of these data show that the combination of an FRC dose of PFC and a large VT would be 6.2 times more likely to be associated with an air leak (95% CI: 1.7 to 21.9) than the combination of a low dose of PFC and a large VT. Given that no animals in Groups 1 or 3 had air leak, the relative risk of an FRC dose of PFC and a large VT, when compared with these other strategies, would be even higher. Air leak was further analyzed through multiple logistic regression, with PFC volume and VT being the variables entered. An FRC dose of PFC was found to be significant at p < 0.003 when compared with low-dose PFC. A large VT was significant at p < 0.0001 when compared with a low VT. In an attempt to determine the univariate risk of air leak for either VT or PFC singly, there is a risk of 15.6 (95% CI: 6.56 to 37.56) for VT (a large VT is 15.6 times more likely to cause a pneumothorax than a low VT) and 7.5 (95% CI: 1.1 to 52.4) for PFC (i.e., FRC filling is 7.5 times more likley to be associated with air leak than partial filling).

View larger version (146K): [in this window] [in a new window]   Figure 4.   Anteroposterior chest radiograph of a rabbit that had died during the course of the experiment, with a PaO2 of 548 mm Hg. Note the significant pneumothorax.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

These results show that both oxygenation and lung mechanics are improved with all of the ventilatory strategies used in our experiment. However, we discovered quite accidentally and early in our work a pneumothorax in an animal whose final Pa_O2 was 548 mm Hg. During this particular study there had been a brief decrease in the Pa_O2 shortly after instillation of the PFC, with no sign of cardiovascular instability or change in compliance. Pneumothoraces during gas ventilation in this model of lung injury are invariably dramatic and usually fatal. We hypothesized that when the lung was filled with PFC, the liquid splinted the lung in an open state, thus allowing excellent gas exchange and unimpaired cardiovascular function, so that it was easy to completely miss the presence of an air leak. We subsequently took anteroposterior and cross-table lateral radiographs after the lavage, prior to PFC instillation, and again after 2 h of partial liquid ventilation. We also did a very careful autopsy of the lungs under water. As our results show, 10 of the 13 animals in Group 2 had evidence of air leak. Five of the eight animals in Group 2 that survived for the duration of the experiment had pneumothoraces. In two, these were large and visible on chest radiographs, and were not associated with cardiovascular compromise. However, the remaining three animals had either trivial signs or no sign of when the pneumothorax occurred. Moreover, all animals that died prematurely did so of cardiovascular shock, and had evidence of air leak. Only one other animal (in Group 4) had a pneumothorax. This animal also died prematurely. The site of lung rupture was usually the edge of the superior part of the upper lobes of the lungs (the nondependent zone). Some of these pneumothoraces were small, but this is far from reassuring. First, it provides clear evidence of barotrauma. Second, the animals were only ventilated for a period of 2 h. Third, as the PFC evaporates and the lung becomes more gas filled, we predict that the more classic manifestations of a pneumothorax would appear. We conclude that the combination of filling of the lung with PFC and using a large VT is inherently unsafe, and that previous animal studies that used this approach may well have overlooked the possibility of pneumothoraces. This complication is neither model- nor animal-specific, as we made similar preliminary findings in an acid-aspiration model in 25-kg pigs. ARDS is characteristically a nonhomogeneous disease, with atelectatic, consolidated areas being found predominantly in the dependent zones of the lung. If one of the primary objectives of liquid ventilation is to recruit the dependent alveoli that are difficult to recruit with gas ventilation, completely filling the lung with liquid seems redundant. How much liquid is required is, however, very difficult to determine. Tütüncü and colleagues (30, 31) have previously shown dramatic improvement in pulmonary mechanics with as little as 3 ml/kg of PFC. Furthermore, at a dose of 6 ml/kg there was improvement (although not maximal) in oxygenation. We arbitrarily chose this latter dose for our experiment. A PEEP of 9 cm H₂O was chosen to increase the gas-liquid interface. This was based on unpublished data collected in our laboratory for the distribution of liquid at various PEEP levels and for pulmonary mechanics measurements. A VT of 9 ml/kg was selected as being somewhat between levels allowing significant permissive hypercapnia and the large-VT group. The most striking difference was that none of the animals in either of the two small-VT groups had autopsy or radiographic evidence of air leak. The significant improvement in both oxygenation and compliance observed initially was not maintained in all groups for the duration of the experiment. We postulate that some of the PFC may have evaporated and the beneficial effects of the low-surface-tension liquid might have been lost. The rate at which to replace this evaporated liquid is the subject of ongoing research.

The use of large tidal volumes with relatively low PEEP levels runs counter to an increasing body of evidence supporting the use of small tidal volumes, a higher PEEP, and a "lung-protective strategy." This is particularly true if the lung has previously been filled with a large volume of incompressible liquid, in which case a delivered volume of gas would preferentially go to the nondependent zone. The scenario would be akin to ventilating the "baby lung" so well described by Gattinonni (32). Given the density of PFC (2 g/cm³), pressures approximating twice the vertical height of the lung would be required to drive gas to the dependent alveoli in the lung filled to FRC with PFC and thus achieve ventilation of this dependent zone. Indeed, in a liquid-filled lung, it is questionable whether it is necessary to get gas into these alveoli.

The goal of mechanical ventilation in the critical-care setting is to optimize gas exchange and oxygen delivery without aggravating underlying lung disease nor damaging healthier regions of the lung. In addition to being excellent solvents for respiratory gases, PFCs offer the unique combination of a substance that is able to recruit the dependent zones of the lung (through its physical characteristics) and improve pulmonary mechanics (through its relatively low surface tension). However, given the findings in this study, we are concerned that the use of a large-VT ventilation strategy in a PFC-filled lung may expose patients to the risk of concealed but potentially life-threatening barotrauma. This work demonstrates that one can have the benefits of PLV without the disadvantage of air leaks by translating a lung-protective strategy to the liquid ventilation paradigm. Further work is necessary to define dosing strategies in PLV.