INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Mechanical ventilation can result in ventilator-induced lung injury (VILI) (1, 2). Strategies aimed at reducing lung stretch, and thereby reducing VILI, have been proposed (3). The idea that ventilation strategy may have an impact on patient outcome received its strongest initial support from a clinical study of protective ventilatory strategy (4). Those findings have recently been strongly endorsed by the release of the preliminary results of the National Institutes of Health (NIH)-sponsored "high stretch-low stretch" trial, demonstrating that "low stretch" ventilation is associated with significantly reduced mortality. (The preliminary report is available at http://hedwig.mgh. harvard.edu/.)

Although the pathophysiologic mechanisms linking ventilatory strategy and patient outcome have not been defined, two lines of investigation have emerged. First, injurious ventilatory strategies have been associated with increased production of lung-derived cytokines in nonperfused (5), perfused (6), and in vivo (7) lung models, in addition to recent human data (8). These findings have been synthesized into a unifying hypothesis relating ventilatory strategy to VILI, multisystem organ dysfunction, and adverse patient outcome (9). Second, bacterial translocationfrom the lungs into the systemic circulationhas been described in preinjured lungs (10, 11). Extending this idea, it has been shown that such bacterial translocation can be caused by adverse ventilation strategy (12, 13). However, given that the organisms responsible for clinical ventilator-associated pneumonia (VAP) are not usually detectable in the systemic circulation (14), it seems unlikely that bacterial translocation can explain the adverse outcome related to mechanical ventilation.

Nonetheless, VAP is important in mechanically ventilated patients with acute respiratory distress syndrome (ARDS) because it is associated with significant attributable morbidity and mortality (15). It is most frequently caused by gram-negative organisms (16), and such organisms produce large quantities of endotoxin (lipopolysaccharide [LPS]) that can result in powerful cytokine stimulation (17). It is likely that endotoxin is clinically important in the pathophysiology of cytokine- mediated systemic inflammatory responses contributing to multisystem organ failure (17, 18), and elevated levels of circulating endotoxin are associated with poor outcome (19), and may predict development of ARDS (20). In patients with VAP, bronchoalveolar lavage fluid (BALF) contains elevated levels of endotoxin (21), and interventions such as diagnostic bronchoalveolar lavage (BAL) precipitate elevations in circulating levels (22). Furthermore, in patients with VAP, elevated levels of circulating cytokines enable identification of those with probable poor outcome (23).

Nevertheless, the role of endotoxin in VILI or VAP has not been defined. This may be because the gastrointestinal tract has traditionally been considered to be the source of systemic endotoxin appearing in the circulation of critically ill patients (24). However, accumulating evidence suggests that translocation of gut-derived endotoxin may not be as clinically important as previously thought (25, 26). Thus, although ventilatory strategy has been shown to govern the appearance of lung- derived bacteria (12) and cytokines (7, 8) into the circulation, the possibility that ventilatory strategy may determine pulmonary-to-systemic endotoxin translocation could be important. This may be especially important in patients with ARDS in whom VAP coexists.

We therefore hypothesized that in a surfactant-depleted rabbit model of hypoxemic VILI, endotoxin instilled into the trachea would translocate from the lung into the systemic circulation, and would be associated with elevated levels of circulating tumor necrosis factor-alpha (TNF-). We further hypothesized that the magnitude of such translocation would be greatest with the use of an injurious ventilatory strategy, and least in the context of a protective strategy. We therefore investigated the effects of a protective versus a nonprotective ventilatory strategy on the sequential plasma levels of endotoxin and TNF- after intratracheal instillation of endotoxin solution in the in vivo saline lavaged rabbit lung model of hypoxemic respiratory failure.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Male New Zealand White rabbits, 3.5 to 4.0 kg, were used in all experiments. All experimental work conformed to the guidelines of the Canadian Committee for Animal Care, and was approved by the Animal Care Committee of the University Health Network.

Anesthesia

Premedication was with intramuscular ketamine (85 mg · kg¹). Anesthesia was induced with intravenous pentobarbital sodium (range, 15 to 25 mg · kg¹), and gentamicin (5 mg · kg¹) was administered. Incremental bolus doses of pentobarbital (5 mg · kg¹) were administered as required. Sterile technique was used during all manipulations. A tracheotomy was performed, and a 4-mm (internal diameter) endotracheal tube (ETT) was inserted to a depth of 1 cm, and secured in place. Pancuronium (1 mg intravenously) was administered after depth of anesthesia was confirmed by absence of response to paw compression. The lungs were ventilated using a small animal ventilator (Model 683; Harvard Apparatus, Holliston, MA) with fraction of inspired oxygen (FI_O2) 1.0, rate 25 min¹, tidal volume (VT) 10 ml · min¹, and 1 cm H₂O positive end-expiratory pressure (PEEP). Under aseptic conditions the carotid artery was cannulated for arterial pressure (Pa) measurement and blood sampling. Anesthesia was maintained with a combination of fentanyl (40 to 90 µg · h¹) and midazolam (0.2 to 0.5 mg · h¹). Additional monitors included a pulse oximeter and a rectal temperature probe. Baseline arterial blood samples were taken for blood gas measurement (Radiometer ABL 300, Copenhagen). Blood samples for endotoxin and cytokine assay were centrifuged and the supernatant stored at 20° C. Animals were excluded if baseline Pa_O2 was less than 300 mm Hg. Lactated Ringer's crystalloid (5 ml · kg¹ · h¹) was administered.

Saline Lavage

Surfactant depletion was induced by lavaging the lungs with warmed sterile saline 0.9% (15 ml · kg¹, temperature 37° C), using a modification of previously described techniques (27). Saline was infused into the trachea from a height of 30 cm. The abdomen and chest were massaged to assist intrapulmonary distribution. The effluent was then drained by gravity. After the first lavage, ventilation was recommenced but with PEEP = 5 cm H₂O and the other ventilatory parameters remained as before. This lavage procedure was repeated a further 3 to 6 times until Pa_O2was less than 100 mm Hg. After this target Pa_O2 was achieved, ventilation was continued for 30 min, and Pa_O2 was measured again. If the Pa_O2 was more than 100 mm Hg, additional lavage was performed. Throughout this period, Pa_CO2 was maintained in the range 35 to 45 mm Hg by adjusting the respiratory rate.

Lung Recruitment

In rabbits in which the Pa_O2 was sustained less than 100 mm Hg at 3 min after lavage, a recruitment maneuver was performed (before randomization). Continuous positive airway pressure (CPAP) at 25 mm Hg was administered for 30 s. After this, the ETT was clamped and reconnected to the ventilator circuit where PEEP was set at 10 cm H₂O. The ETT was unclamped during expiration, and ventilation was continued for 10 min. Pa_O2 measurement was repeated, and recruitment considered successful where Pa_O2 was greater than 200 mm Hg. If Pa_O2 was less than 200 mm Hg, the recruitment maneuver was repeated, but was followed with ventilation as previously described, where PEEP was set at 12.5 cm H₂O. If at this stage Pa_O2 was less than 200 mm Hg, the preparation was excluded from further study.

Group Allocation

Preparations were then randomly allocated to one of four treatment groups, defined by the type of ventilation strategy (protective versus nonprotective) and intratracheal instillation (endotoxin versus saline), as follows: nonprotective ventilation + intratracheal instillation of saline (NS); nonprotective ventilation + intratracheal instillation of endotoxin (NE); protective ventilation + intratracheal instillation of saline (PS); or protective ventilation + intratracheal instillation of endotoxin (PE).

Ventilatory Strategy

Nonprotective ventilatory strategy consisted of: VT = 12 ml · kg¹, with PEEP = 0 cm H₂O. Protective ventilatory strategy consisted of: VT = 5 ml · kg¹, with PEEP ranging from 10 to 12.5 cm H₂O.

Intratracheal Instillation

Intratracheal instillation consisted of administrationnot blinded of either endotoxin (Escherichia coli lipopolysaccharide; Sigma Inc., St. Louis, MO) or sterile saline (0.9%), via a catheter inserted through the ETT to a distance of 1 cm distal to the ETT tip. Ventilation was interrupted, and preparations randomized to endotoxin administration received endotoxin (500 µg) diluted in 2 ml of sterile saline (0.9%), followed by a flush with sterile saline (0.9%, 2 ml). Preparations randomized to saline received sterile saline (2 ml, 0.9%), followed by an additional flush with sterile saline (0.9%, 2 ml). A recruitment maneuver was performed in order to reexpand the lungs, and ventilation commenced according to the randomized ventilatory strategy.

Monitoring and Parameter Targets

Animals were ventilated for 3 h. Mean arterial blood pressure (), peak airway pressure, and temperature were recorded at baseline values before lung lavage, and at 20, 40, 60, 120, and 180 min after group allocation. Arterial blood samples were drawn at these times for blood gas measurement, as well as cytokine and endotoxin assay, as previously described. less than 40 mm Hg was treated with up to three additional boluses of lactated Ringer's solution (5 ml · kg¹), and thereafter with intravenous phenylephrine (10 µg · kg¹) boluses. FI_O2 was 1.0 throughout the experiment. After randomization, Pa_CO2 was targeted to a range of 50 to 60 mm Hg. This was achieved in groups with high VT (NS, NE), by adjusting the ventilator circuit dead space; and, in the groups with low VT (PS, PE), by alteration of the ventilator frequency. In groups randomized to PEEP (low VT: PS or PE), Pa_O2 was targeted to > 100 mm Hg by titration of PEEP (10 to 12.5 cm H₂O). Sodium bicarbonate (0.5 mmol · L¹ · kg¹) was given where arterial HCO₃ was less than 18 mmol · L¹.

Ventilation continued for 3 h, at which time the animals were killed by pentobarbital overdose.

Endotoxin Assay

Analysis of the serum endotoxin was carried out in triplicate and in a blinded fashion using a commercially available Limulus amebocyte lysate (LAL) Pyrochrome diazo-coupling assay kit (Associates of Cape Cod, Falmouth, MA), using standardized methodology (30).

TNF- Assay

Analysis of serum TNF- was carried out in triplicate and in a blinded fashion with a commercially available polyclonal sandwich antibody ELISA kit (PharMingen, Mississauga, ON, Canada), using standardized methodology (31).

Statistical Analysis

Data were analyzed using SigmaStat (Version 2.0; Jandel Corporation, San Rafael, CA). Group comparisons of parametric data and nonparametric data were performed with 2-way repeated measures analysis of variance (ANOVA) or ANOVA on ranks, respectively, followed by one-way ANOVA followed by Student-Newman-Keuls tests. Comparisons of plasma concentrations between ultimate survivors versus nonsurvivors was by t test or Mann-Whitney tests. Categorical data were compared using ² or Fisher exact test. Animals with detectable levels of LPS or TNF- at baseline values were deleted from analyses relating to plasma concentrations of LPS and TNF- respectively. Linear regression was used to assess the correlation between Pa_O2 and plasma endotoxin concentration. Significance was set at p < 0.05. Results are expressed as mean ± SEM.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Baseline Characteristics

Eight animals per group completed the protocol. Baseline characteristics were comparable among the groups (Table 1, Figures 1-6). Two rabbits were excluded before randomization.

View larger version (17K): [in this window] [in a new window]   Figure 1.   Plasma endotoxin concentration versus time. In the interval 40 to 180 min after randomization, endotoxin concentration was significantly higher in NE versus all other groups (*p < 0.05). Concentration was unchanged throughout the experiment in all other groups.

View larger version (22K): [in this window] [in a new window]   Figure 2.   Plasma endotoxin concentration was significantly higher in eventual nonsurvivors than survivors, at 40, 60, and 120 min after randomization (*p < 0.05).

View larger version (19K): [in this window] [in a new window]   Figure 3.   Plasma TNF- versus time. At 120 min after randomization, TNF- concentration was significantly higher in NE compared with all other groups (*p < 0.05). Concentrations were not different among any of the other groups at any time.

View larger version (18K): [in this window] [in a new window]   Figure 4.   Pa over time. In the interval 40 to 180 min after randomization, Pa was significantly lower in NE versus all other groups (*p < 0.05). Pa remained unchanged in PS, but fell significantly during the course of the experiment in all other groups (*p < 0.05).

View larger version (18K): [in this window] [in a new window]   Figure 5.   Peak airway pressure (Paw) over time. Paw was similar in groups PE and PS, and in groups NE and NS. At 120 and 180 min, Paw was significantly greater in both NE and NS groups than in either PE or PS groups (p < 0.05).

View larger version (18K): [in this window] [in a new window]   Figure 6.   Arterial bicarbonate concentration over time. Bicarbonate concentration was unchanged over time in the PS group. In the interval 40 to 180 min after randomization, bicarbonate was significantly lower in NE than in all other groups (p < 0.05).

Survival

All animals in the groups PS (8/8) and PE (8/8) completed the protocol. In the NS and NE groups the survival was 7 of 8 and 5 of 8 animals respectively (NS: ² test).

Plasma Endotoxin

The plasma endotoxin levels were negligibleand similar among the groups at baseline values and 20 min after randomization (Figure 1). From 40 min until completion of the experiment (180 min), the endotoxin levels were markedly greater in the NE group versus all other groups. Furthermore, there were no differences in endotoxin concentrations among the other groups. At 40, 60, and 120 min, plasma endotoxin levels were significantly greater in animals that did not survive compared with those that survived (Figure 2). Elevations in plasma endotoxin concentration were not associated with the degree of hypoxemia (linear regression: adjusted coefficient of determination [R²] = 0.119, p = 0.55). Furthermore, from 60 to 180 min inclusive, there was no association between Pa_O2 less than median (45.1 mm Hg) and endotoxin values greater than median (4.7 endotoxin units [EU] · ml¹) (p = 0.22, Fisher exact test).

TNF-

There was significant elevation in plasma levels of TNF- over time in all groups (Figure 3). At 120 min, the concentration in the NE group was significantly greater than in the other groups.

was comparable over time in the PS, PE, and NS groups (Figure 4). By 40 min and thereafter, the was significantly lower in the NE group versus all the other groups (Figure 4). Administration of phenylephrine was required in more animals in the NE group (5/8) versus the NS (0/8), PE (0/8), or PS (1/8) groups (p < 0.05, ² test).

Airway Pressure (Paw)

Paw was comparable among the groups at baseline values (Figure 5). After 40 min, Paw was similar in the NS and NE groups, and was significantly greater in these groups versus the PE and PS groups until the completion of the experiment.

Arterial Oxygenation

Pa_O2 was similar in all groups at baseline (Table 1). From 20 to 60 min after randomization, Pa_O2 was comparable in PS and PE, and was significantly higher in these versus NS or NE. By 180 min, Pa_O2 was comparable between PE versus PS, and between NE versus NS. At that stage, Pa_O2 was significantly greater in PS and PE versus NS and NE.

Pa_CO2 and Acid Base

Pa_CO2 values at baseline and after randomization are presented (Table 1). In addition to intravenous crystalloid, sodium bicarbonate was administered to maintain plasma HCO₃ above the target level 18 mmol · L¹. The number of animals that received intravenous sodium bicarbonate was significantly greater in the NE group (7/8) versus the NS (0/8), PE (2/8), or PS (0/8) groups (p < 0.05, ² test). The estimated HCO₃ concentration fell progressively in the NE group after randomization (Figure 6), paralleling the changes in pH (Table 1). The HCO₃ concentration in the PS group remained unchanged throughout, whereas there were minor reductions in the concentrations in the PE and NS groups after the 60 min.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The findings of the current study confirm our hypothesis that pulmonary-to-systemic translocation of endotoxin can occur. We further confirmed that in this model, nonprotective ventilatory strategy causes translocation; that translocation in the context of adverse strategy is associated with maximal levels of circulating TNF-; and finally, that plasma levels of translocated endotoxin are far higher in nonsurvivors compared with survivors. Although there are many nonclinical features of the current laboratory model, the VT used are within clinically relevant ranges (4). (The preliminary report is available at http://hedwig.mgh.harvard.edu/ardsnet/nih.html.)

The results are corroborated by the findings that the group in which translocation developed had a significantly worse outcome in terms of hypotension, metabolic acidosis, and requirements for resuscitation with pressor support and intravenous sodium bicarbonate. Arterial oxygenation and Paw were affected by ventilatory strategy, and not by administration of intratracheal endotoxin alone. Therefore in this model, nonprotective ventilatory strategy (high VT and zero PEEP) in the absence of endotoxin administration was associated with adverse oxygenation and elevated airway pressures. However, protective ventilatory strategy (low VT, high level of PEEP) was associated with negligible plasma levels of endotoxin and good systemic outcome, regardless of whether endotoxin was (PE group) or was not (PS group) administered. Therefore, protective ventilation prevented endotoxin translocation, and the associated adverse physiologic effects.

Translocation and Lung Infection

Translocation of bacteria from the lung has been described in a variety of conditions (10). However, the current findings offer an additional perspective for several reasons. First, bacterial translocation from the lung to the systemic circulation has been described in canine (12) and rat (13) models. However, that demonstration of translocation involved extremely large VT. Second, the diagnostic isolation of pathogens from the bloodstream of patients with VAP is extremely infrequent (14), suggesting that bacterial translocation is not a common occurrence. Nevertheless, VAP does have an attributable morbidity (15) that could be contributed to by multisystem dysfunction (23). Given that diagnostic BAL can result in the detection of circulating endotoxin in patients with VAP (22), it is possible therefore in the clinical context, that translocation of endotoxin may be responsible for some of the morbidity arising from VAP.

It is becoming progressively clearer in multiple studies that adoption of ventilatory strategies that use smaller VT is associated with improved clinical outcome in ARDS (4). Although several mechanisms have been proposed to account for the improved outcome associated with low stretch ventilatory strategies, such as decreased release of lung-derived cytokines (6) or reduced translocation of bacteria (12) into the systemic circulation, the exact explanation is unclear. It is possible, however, that translocation of endotoxin occurs in patients with ARDS, especially those in whom VAP is present, and particularly where a nonprotective ventilatory strategy is being used.

Mechanisms of Translocation

Although much is known about the effects of parenterally administered endotoxin, less is known about the effects after intratracheal administration. It is clear, however, that the endotoxin translocation in the current study was related to the ventilatory strategy used. However, the precise of contribution of PEEP versus VT was not explored in the current study. The strategy used suggests that the translocation occurred because of stretch of nonrecruited (zero PEEP) lung. However, in the absence of studies of the effects of altered VT over a broad range of lung recruitment, this remains conjectural. It is not possible, from the current data, to separate the specific effects of PEEP versus VT in the causation of translocation. Nonetheless, at the microvascular level, capillary stress failure has been described in the context of alterations in inflation stretch (32) and elevations in capillary hydrostatic pressure (33). Given the VT used (with zero PEEP) in the current study, it is possible that alveolar capillary stress failure occurred, allowing passage of endotoxin from the alveolus to the pulmonary circulation. Our data are consistent with previous findings (34) that endotoxin administered into the airspace is not detectable in the intravascular compartment, in the absence of lung injury.

It is conceivable that hypoxemia played a role in the endotoxin translocation, given the finding that exposure to hypoxemia over 24 h is associated with development of pulmonary edema following viral respiratory infections in the rat (35). However, in the current study, no correlation was demonstrable between endotoxin concentration and degree of hypoxemia. Furthermore, categorical analysis exploring possible association between high endotoxin concentrations (above median endotoxin) with severe hypoxemia (below median Pa_O2), and between lower endotoxin concentrations (below median endotoxin) and less severe hypoxemia (above median Pa_O2), revealed no significant association.

Mechanisms of Injury

The higher endotoxin concentrations in the nonsurvivors suggests that the translocation may be a cause, or at least a marker, of poor outcome in this model. Maximal plasma endotoxin levels and TNF- levels occurred in the NE group. The worse systemic hypotension and metabolic acidosis in this group, suggests that adverse strategy and the presence of lung endotoxin together result in elevated circulating levels of endotoxin and TNF-, and these elements may play a role in outcome.

Although the endotoxin appears to be derived from the lung, the source of the TNF- is not clear. Adverse ventilation has been associated with release of TNF- into the airspace (5) and into the systemic circulation (7) in rodent models. However, it is quite possible that systemically released endotoxin resulted in systemically generated TNF-, through stimulation of bloodborne macrophages. Alternatively, the TNF- may be derived from the lung, as has been clearly demonstrated in an isolated perfused murine lung model (6). This possibility extends the previous data demonstrating that pulmonary versus circulating TNF- profiles resulting from endotoxin stimulation depend critically on the locus of endotoxin application (34). Specifically, airway delivery of endotoxin was associated with significant recovery of TNF- in the alveolar compartment, with little detection in the circulating pulmonary perfusate (34). Furthermore, these investigators reported that airspace endotoxin administration was not associated with development of physiologic lung injury (34). The in vivo findings from the current study that airspace endotoxin administration in the context of protective ventilatory strategy was associated with neither significant systemic endotoxin detection, greater levels of TNF-, nor the development of overt lung injury, extend and support those earlier data (34).

A novel possibility relates to the potential for lung "priming" of pulmonary or systemic inflammatory responses, where prior exposure to endotoxin greatly potentiates the severity of subsequent lung injury (36, 37). In fact, a similar response has been described after preadministration of intravascular endotoxin (38). These findings may impact upon the current study in that a component of VILI may result in accentuated lung and/or systemic injury in the context of airspaceor translocated intravascularendotoxin. Such a two-hit phenomenon may have important clinical implications.

Limitations of Current Findings

There are several limitations of the current study. We have not proven conclusively that the endotoxin was derived from the lung. However, given the negligible concentrations in the other three groups, use of aseptic technique, and pretreatment with gentamicin, it seems highly unlikely that there was contamination or an alternative source of endotoxin. In addition, it is possible that BALF may have demonstrated differential cytokines concentrations (5) or that different cytokine profiles may have been detected if tested. These issues, however, were not the primary focus of interest for the study.

It is possible that the recruitment maneuvers had adverse effects; however, all groups were subjected to the same recruitment strategies before randomization and after intratracheal instillation. Although the VT studied are within clinically relevant ranges, there are several issues that limit direct extrapolation of the current findings to the clinical situation. Because rabbits are more prone to stress failure compared, for example, with dogs (39), it is possible that significant species variability exists, and that caution must therefore be exercised with extrapolation from the current model. The impact of alveolar endotoxin on stress fracture susceptibility is unknown. The experimental protocol was of short duration, and was not associated with bacterial sepsis. Furthermore, the model VILI in the context of surfactant depletionmore reflects neonatal respiratory distress syndrome associated with prematurity, rather than adult ARDS. A final limitation of the model reflects the inability to precisely determine the nature of the epithelial or microvascular disruption responsible for the passage of endotoxin.

Significance and Potential Application

We speculate that the findings from the current study may be of significance relating to a number of important areas. First, pulmonary-to-systemic endotoxin translocation advances our knowledge of events occurring as a result of adverse ventilatory strategy. Second, such translocation, and the accompanying cytokine profile, may also explain some of the morbidity or mortality associated with nonprotective ventilatory strategies. This might be especially applicable in the setting of VAP. Of course, combining these two points suggests, hypothetically, that protective ventilatory strategy might be of greatest significance in the context of VAP, because of the ready source of endotoxin and the potential for translocation. If this is demonstrated clinically, then protective ventilatory strategies may prove to be of greatest importance in patients with coexisting ARDS and VAP. Third, the demonstration of endotoxin translocation points to the potential for use of rapid detection of plasma endotoxina technology already well described (40)for the early diagnosis and treatment of VAP, as has previously been suggested for measurement of endotoxin in BAL samples (21). Finally, the findings suggest that measurement of plasma endotoxinor any comparable moleculeafter intratracheal administration might serve as a marker of lung injury in experimental models.