INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In large animals, lung injury induced by mechanical ventilation at high transalveolar pressures distributes preferentially to the dependent regions (1, 2). Although the mechanisms responsible for this uneven distribution of injury along the vertical axis are not yet elucidated, vertical gradients of blood flow, vascular pressure, and pleural pressure (3) may contribute. The higher (more positive) pleural pressure that surrounds the dependent lung regions tends to decrease regional volume (4) and to promote dependent collapse/reopening during the tidal cycle. Such mechanisms could account for the dorsal (dependent) predominance of ventilator-induced lung injury found in supine animals (1, 2, 5, 6). Regional hydrostatic and hemodynamic differences may also help to explain the dorsal predominance of injury. Blood flow distributes predominantly to the dependent lung regions in the supine position (7, 8). The amplifying effect of cardiac output on lung water when ventilator-induced lung injury (9) or acute lung injury (10) is present suggests that increased regional flow could exacerbate lung injury; exudation of protein-rich fluid may inactivate surfactant and further alter membrane permeability as a result of increased surface tension and radial traction on pulmonary microvessels (11). The potential contribution of capillary pressurewhich tends to increase in parallel with flowto the development or extension of ventilator-induced lung injury was recently demonstrated by West and colleagues (12, 13).

The present study tested the hypothesis that lung perfusion may alter the extent of alveolar damage induced by mechanical ventilation. Two sets of isolated rabbit lungs were perfused at either low-flow or high-flow rates and ventilated with the same injurious ventilatory pattern. A third (control) group was ventilated at lower peak pressure to verify that the ventilatory pattern used in the study groups had produced lung injury (Table 1). Compared with the low-flow group, the high-flow group experienced more overall lung damage, as assessed by histology and by changes in lung compliance (CL), weight, and ultrafiltration coefficient (K_f). The control animals exhibited the least alterations.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animal Instrumentation and Ex Vivo Heart-Lung Preparation

All techniques and procedures were approved by the Animal Care and Use Committee of Regions Hospital. Young New Zealand white rabbits weighing 2.9 ± 0.4 kg (mean ± SD) were anesthetized with an intramuscular injection (1 ml/kg) of a ketamine/acepromazine solution (10 ml of ketamine [100 mg/ml] + 2 ml of acepromazine [10 mg/ ml]). An endotracheal tube with 4.0 mm internal diameter was inserted via a tracheostomy. The right carotid artery was cannulated for heparin administration (700 U/kg).

The heart-lung block was excised through a midline sternotomy. Exsanguination was performed and 50 ml of blood were collected and later added to the perfusate (vide infra). The heart-lung block and the attached endotracheal tube were weighed together and then at the end of the experiment separately, to calculate the initial lung weight. Perfusion cannulas were then placed into the pulmonary artery and left atrium. Perfusion was started at a constant flow of 25 ml/min. The ischemic time for the lungs was recorded (mean ischemic time: 16.2 ± 3.7 min). The preparation was suspended from a counterbalanced force transducer (FT03; Grass Instruments, Quincy, MA).

Circuit Description

The perfusate in the circuit consisted of 300 ml of Krebs-Henseleit solution (buffered to pH 7.4) to which 5% bovine serum albumin and 50 ml of autologous blood were added to serve as a marker of vascular damage when assessed histologically. The resulting hematocrit of the perfusate was approximately 3%. A cyclooxygenase inhibitor (ketorolac, 10 mg) was added to the blood to prevent thromboxane-mediated pulmonary hypertension secondary to ischemia and reperfusion (14). The closed perfusion circuit consisted sequentially of a venous reservoir, a heat exchanger that maintained the lung temperature at approximately 33° C (33.5 ± 0.8), a digital rotary pump (Masterflex 07550; Cole-Parmer, Niles, IL), two parallel blood transfusion filters (SQ40S/SQ40ST; Pall Biomedical Products, East Hills, NY), a bubble trap, arterial tubing leading to the pulmonary artery cannula and the lung, and venous tubing leading from the left atrial cannula back to the venous reservoir through a free distal end tube (open to atmosphere) positioned above the venous reservoir. The free distal end of the tubing was raised or lowered to adjust effective left atrial pressure. The left atrial and pulmonary artery cannulae housed pressure taps (PE 50) which were connected to pressure transducers and to a monitor (Sirecust 404, Siemens, Germany) from which signals were sent to a chart recorder (Model 95000; Astro-Med, West Warwick, RI).

Determination of Lung Weight, Ultrafiltration Coefficient, Hemodynamic and Ventilatory Parameters Measurements

The counterbalanced force transducer (FT03; Grass Instruments) was placed in series with a balance (Cent-O-Gram triple beam balance; Ohaus Corporation, Florham Park, NJ) to act as differential force transducer. The output from the force transducer was amplified (P122-A; Grass Instruments) and fed to the chart recorder and to digital tape (RD-111T; TEAC, Tokyo, Japan) for later analysis. Two calibrations (one for measurement of the ultrafiltration coefficient [K_f] and one for weight gain) were performed by adding 1 and 20 g on the suspended heart-lung preparation to verify signal transmission under the actual experimental conditions.

Airway opening pressure was measured (Validyne MP, 45 ± 100 cm H₂O) from a side tap connected to common limb of the ventilator circuit. CO₂ (~ 0.1 L/min) was blended into the inspired gas (fraction of inspired oxygen [FI_O2] 40%) at a fraction of inspired carbon dioxide (FI_CO2) of approximately 5% during ventilation to help maintain a physiologic pH of the perfusate (7.28 ± 0.07) (Corning 168 blood gas analyzer; Corning Glass, Medfield, MA) during the course of the experiment. Airway pressure, vascular pressure, and weight gain signals were read from the chart recorder and simultaneously stored on digital tape (RD-111T; TEAC, Tokyo, Japan) for later determination of K_f (see below).

Initial Hemodynamic and Ventilatory Settings

The heart-lung preparation was connected to the ventilator (Veolar; Hamilton, Reno, NV). At continuous positive airway pressure (CPAP) of 5 cm H₂O, perfusion was increased to 300 ml/min and vascular pressures were referenced to the bottom of the lungs. The free extremity of the left atrial cannula was positioned so that the left atrial pressure was 10 mm Hg, ensuring West zone III conditions along the vertical axis of the lungs ( 8 cm height) at CPAP 5 cm H₂O. To let the preparation equilibrate, the lungs were ventilated for 20 min with pressure control ventilation (PCV 15 cm H₂O, positive end-expiratory pressure [PEEP] 5 cm H₂O, frequency 20/min, inspiratory time fraction 0.33, FI_O2 40%, and FI_CO2 5%).

Baseline K_f and Hemodynamic Measurements

The stability of the temperature and weight (isogravimetry: no weight gain nor loss during CPAP [5 cm H₂O]) of the heart-lung preparation was confirmed. The free end of the atrial cannula was then rapidly raised by 3 cm. Simultaneously, the weight of the heart-lung preparation was recorded continuously for at least 4 min. Perfusion of the lung under West zone III conditions was documented by simultaneously recording the rise in pulmonary artery pressure associated with the step increment in left atrial pressure (15). Ultrafiltration coefficients (at baseline and at the end of the experiment) were calculated from the slope of the segment of slow steady rate of weight gain (representing transcapillary fluid filtration) (16) between 2 to 4 min after the step increase in left atrial pressure. This slope was determined by computerized least squares curve fitting of the weight signal recorded on digital tape during this interval. Left atrial pressure was then returned to the baseline value (10 mm Hg). Pulmonary artery, left atrial, and capillary pressures were recorded. Capillary pressures were determined by the simultaneous double arterial/venous occlusion method, as described elsewhere (17, 18).

Ventilation/Perfusion Protocol

Each preparation was randomized to be perfused with a constant flow of either 300 ml/min (low flow group: n = 7) or 900 ml/min (high flow group: n = 7). To confirm the stability of the ex vivo model over the time course of the experiment, as well as to demonstrate that ventilation and perfusion of the model at near physiologic inflation pressure and rate of perfusion does not cause severe injury, a third group was perfused with a 500 ml/min flow (control group: n = 7) at lower tidal pressures (PCV 15 cm H₂O, PEEP 5 cm H₂O, frequency 20/min). Left atrial pressures were adjusted until the capillary pressures were identical (10 mm Hg) in all groups. After isogravimetry was verified, the lungs of the high- and low-flow groups were ventilated for 35 min with identical ventilatory settings expected to induce lung injury in this model: PCV 30 cm H₂O, external PEEP 0 cm H₂O, frequency 20/min, inspiratory time fraction 0.33, FI_O2 40%, and FI_CO2 5%).

Within 5 min after the start and within 5 min before the end of the ventilation/perfusion protocol, the following parameters were recorded: peak static airway opening pressure (inflation pressure at zero flow), end-expiratory pressure (including autoPEEP), and pulmonary artery pressures (peak value during inspiration and nadir value during expiration). Simultaneously, VT was read from the ventilator display to track changes in lung compliance from the beginning to the end of the ventilation/perfusion protocol.

Postventilation/Perfusion Protocol Measurements and Tissue Sampling Method

At the end of the protocol, CPAP 5 cm H₂O was resumed, and perfusion rate and left atrial pressure were returned to the same values under which baseline K_f was measured. Pulmonary arterial, left atrial, and capillary pressures were recorded. As soon as the heart-lung preparation became isogravimetric, K_f was determined as described earlier for comparison with the baseline value.

Perfusion was then interrupted and the perfusate drained from the vascular compartment. The left main bronchus was isolated from the airway by placing a ligature around the left hilum. Only the right lung was fixed for histology with 10% buffered formalin (30 ml) injected into the airways kept open by CPAP of 20 cm H₂O. The left lung was used for a separate study. Three histologic samples per lung were obtained according to a predetermined template to minimize sampling bias. Samples (3-mm coronal sections, ~ 2 cm in greatest dimension) were systematically taken perpendicular to the cranial-caudal axis from the upper, middle, and lower lung at mid-distance from the ventral, cranial, and caudal boundaries of each lobe, regardless of the gross pathologic appearance.

Histology

Coded slides prepared from each sample were examined by a pathologist (DO) blinded to the experimental protocol, study group, and region of sampling. Each slide was initially screened by light microscopy under low power (×25 and ×100) in order to visualize the most damaged area present. In the area of most severe damage, the percentage of alveoli containing erythrocytes was determined by examining 100 sequential alveoli (extent of alveolar hemorrhage). The same area was examined at high power (×400), and the number of red blood cells in the alveoli with the most intensity of intra-alveolar hemorrhage was counted (intensity of alveolar hemorrhage). In practice the number of cells could be easily approximated when less than 50 red blood cells were present (e.g., Figure 1A). In areas of extensive hemorrhage, the number of alveoli with more than 50 erythrocytes in one high-power field was counted (e.g., Figure 1B). For example, in areas of extensive hemorrhage, one high-power field often contained numerous alveoli, each with more than 50 erythrocytes. Each slide was also examined at ×25 and ×100 to determine the number of extra-alveolar vessels that were surrounded by prominent perivascular hemorrhage (extent of extra-alveolar hemorrhage) (Figure 1C and D). The mean values of each index of hemorrhage were calculated for each lung (3 slides per lung) and the groups were compared for each index. In order to assess the overall intensity of hemorrhage, each of these histologic indices (extent of alveolar hemorrhage, intensity of alveolar hemorrhage, and number of vessels with perivascular hemorrhage) was equally scored from 0 to 5. The hemorrhage score of each lung (histologic score of lung hemorrhage) was calculated as the sum of the scores assigned to each distinct index of hemorrhage (Table 2). The accuracy and consistency of the grading system were validated by comparing the histologic scores assigned by two separate pathologists, who independently examined and graded 33 out of a total of 66 histologic slides in a blinded fashion. The strength of correlation between the histologic scores of lung hemorrhage and weight gain was also examined.

View larger version (94K): [in this window] [in a new window]   Figure 1.   Representative histologic findings. Panels A and B: alveoli containing erythrocytes. Panels C and D: perivascular hemorrhage.

Statistics

All values are expressed as mean ± SD. Between-group comparisons were performed using one-way analysis of variance (ANOVA) or Kruskal-Wallis one-way ANOVA on ranks, as appropriate. Tukey's correction (for ANOVA) and Student-Newman-Keuls' adjustment (for Kruskal-Wallis one-way ANOVA on ranks) were performed for multiple comparisons. Within-group comparisons of variables obtained at the beginning and at the end of the protocol were analyzed with either paired t test or Wilcoxon's signed rank test, as appropriate. Correlations were examined using the Pearson product-moment coefficient correlation and linear regression, as appropriate. An  < 0.05 was considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Initial Parameters

The initial weights of lungs and baseline pulmonary artery pressures, capillary pressures, and initial K_f values did not differ statistically between groups (Table 3).

Ventilation Parameters

The level of autoPEEP was similar in the high- and in the low-flow groups. Tidal volumes (VT) were, however, slightly higher in the high-flow group (p = 0.048) (Table 3). The difference was, however, not significant when VT was normalized to initial lung weight (a surrogate of lung size: high-flow group 8.4 ± 1.4 ml/g; low-flow group 7.3 ± 1.5 ml/g; p = 0.16). In addition, there was no statistically significant correlation between the VT generated by a given transpulmonary pressure and the different indices of lung injury in the study groups. During the ventilation/perfusion protocol, VT remained unchanged in the low-flow group (start 96 ± 12, end 98 ± 12 ml) and declined significantly only in the high-flow group (start 107 ± 6, end 95 ± 9 ml) (p < 0.05). Indeed, the decrease in static lung compliance of the high-flow group differed significantly from those of the other two groups (low-flow versus high-flow; p = 0.007) (Figure 2).

View larger version (18K): [in this window] [in a new window]   Figure 2.   Changes in static lung compliance, ultrafiltration coefficient (Kf) and lung weight. **p < 0.05 for the comparison with the control and low-flow groups. *p < 0.05 for the comparison with the control group.

Hemodynamic Parameters

Prior to the beginning of ventilation, perfusion at different flows did not result in markedly different pulmonary artery pressures between groups (Table 3). After beginning ventilation, measured pulmonary artery pressures remained similar between groups during expiration but differed during inflation (significantly higher levels in the high-flow group; Figure 3). A regression analysis indicated that a statistically significant and strong relationship ties the change in vascular pressure resulting from tidal ventilation and the indices of lung injury used in this study (Figure 3). The pulmonary arterial pressures, measured in all preparations under the same conditions (perfusion 300 ml/min, left atrial pressure 10 mm Hg, CPAP 5 cm H₂O) before and after the ventilation/perfusion protocol, rose significantly in the high-flow (from 20.3 ± 5 to 24.5 ± 4.4 mm Hg) (p < 0.05) but not in the other groups (data not shown).

View larger version (35K): [in this window] [in a new window]   Figure 3.   Correlations between the changes in pulmonary artery pressures during ventilation and the indices of lung injury. **p < 0.05 for the comparison with the control and low-flow groups. The measures in pulmonary artery pressure were obtained at the start of the ventilation/perfusion protocol.

Pulmonary Edema and K_f

Lung weight, a surrogate of pulmonary edema formation measured continuously during the ventilation/perfusion protocol, increased markedly more in the high-flow group (16.2 ± 9.2 g) than in the low-flow (8.7 ± 3.4 g) (low-flow versus high-flow; p = 0.038) and control groups (1.6 ± 1.0 g) (Figure 2). Large weight gains were consistently associated with a decreasing VT due to the concomitant reduction in lung compliance (Figure 2). At the end of the protocol, the K_f values (g · min¹ · cm H₂O¹ · 100 g¹) of the high-flow (0.50 ± 0.24), low-flow (0.51 ± 0.28), and control groups (0.30 ± 0.19) did not differ statistically but tended to be lower in the control group. The high-flow group, however, appeared to have experienced the highest absolute and relative K_f (0.26 ± 0.20 g · min¹ · cm H₂O¹ · 100 g¹ ~ 115%) whereas K_f of the control group did not increase (K_f 0.05 ± 0.22 g · min¹ · cm H₂O¹ · 100 g¹ ~ 8%) (p < 0.05 high-flow versus control group). In the low-flow group, K_f (0.17 ± 0.10 g · min¹ · cm H₂O¹ · 100 g¹ ~ 60%) was less marked than in the high-flow group, but not significantly so (p = 0.33 low-flow versus high-flow) (Figure 2).

Histology

The individual components of the combined histologic score of lung hemorrhage correlated significantly with weight gain: percentage of alveoli containing red blood cells, r = 0.63, p < 0.01; score of intensity of alveolar hemorrhage, r = 0.53, p = 0.01; number of extra-alveolar vessels with perivascular hemorrhage, r = 0.62, p < 0.01. The combined histologic score of lung hemorrhage, however, correlated better with weight gain than any of its individual components (r = 0.74, p < 0.01). The combined lung hemorrhage score also correlated well (r = 0.73, p < 0.01) between the two pathologists, who separately examined the samples in a blinded and independent fashion.

Lung hemorrhage was more intense and extensive in the high-flow group than in the low-flow group (Figure 4: histologic scores of lung hemorrhage, p < 0.05 for all pairwise comparisons; p = 0.0023 for low-flow versus high-flow). The lungs of the high-flow group had more red cells per alveolar space, more alveoli with red blood cells, and more extra-alveolar vessels with perivascular hemorrhage than the low-flow and control groups (Figure 4). Interestingly, perivascular hemorrhages were exclusively found in the lungs ventilated at high transalveolar pressure, and not in the control group (Figure 4).

View larger version (26K): [in this window] [in a new window]   Figure 4.   Histologic indices of lung hemorrhage. **p < 0.05 for the comparison with the control and low-flow groups. *p < 0.05 for the comparison with the control group. RBC = red blood cells.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our main finding is that in this model of ventilator-induced lung injury, the intensity of perfusion contributed to the reduced lung compliance resulting from an adverse ventilatory pattern and promoted lung edema (weight gain) and hemorrhage (histology). The intensity of perfusion per se did not significantly increase permeability (K_f tended to be higher in the high-flow than in the low-flow group but not significantly so). We found, however, a strong correlation between the vascular pressure changes resulting from the interaction between ventilation and perfusion and indices of lung injury, including K_f. Before discussing the possible mechanisms and implications of these findings, several methodological issues must be addressed.

Critique of the Study

Study design and ventilator-induced lung injury. Ventilation at high transpulmonary pressure (30 cm H₂O) was associated with alveolar hemorrhage, edema formation, decreased lung compliance, and altered permeability. To conclude with certainty that these changes were mainly caused by mechanical ventilation, a control group was ventilated at a lower transpulmonary pressure (15 cm H₂O). The control group manifested fewer alterations than either study group despite being perfused at higher flow (70% increase) than the low-flow group and lower flow (40% decrease) than the high-flow group. Compared with the control group, the more extensive lung alterations observed in the study groups were, therefore, mainly related to the injurious ventilatory pattern used in those groups, and not to the differences in the intensity of perfusion between the two study groups and the control group. This is not surprising, since 15 (control group) and 30 cm H₂O (study groups) transpulmonary pressures correspond to inflation pressures about 10 cm H₂O lower and 5 cm H₂O higher (respectively) than those required to reach total lung capacity of adult rabbit lungs (19).

In the study groups, the higher rate of perfusion contributed significantly to reduced lung compliance and tended to promote edema and hemorrhage. Although the study design did not include control groups perfused at flows identical to those of the study groups, our data strongly suggest that the contribution of the perfusion to lung injury varies with the ventilatory pattern used. As the lungs are inflated above FRC, pulmonary vascular resistance rises, and for a given tidal inflation, pulmonary artery pressure increases in parallel to the rate of perfusion, as predicted from Poiseuille's law. It should be stressed that the pulmonary arterial pressures were measured in the trunk of the pulmonary artery, before it divides and enters into the lungs. Any tidal change in pulmonary artery pressure thus reflects change in vascular resistance (flow being constant) and not airway pressure change per se. The magnitude of the inspiratory rise in pulmonary artery pressure varies with the intensity of the perfusion and the size of the tidal breath since lung volume is an important determinant of vascular resistance. Combining the data from all groups, we found good correlations between the rise in pulmonary artery pressure caused by tidal inflation and weight gain, lung hemorrhage, and K_f (Figure 3). These correlations indicate that lungs perfused at rates identical to those used in the study groups (300 or 900 ml/min) but ventilated at low peak transalveolar pressures ( control group) should experience fewer lung alterations. More importantly, however, the absence of a significant difference in K_f between the high- and low-flow group, and the significant correlation between the change in vascular pressure and K_f strongly suggest that it is not the intensity of perfusion per se that is responsible but rather the tidal vascular pressure changes that result from the interaction between perfusion and ventilation that has the greatest impact on vascular integrity in this model. In other words, the potential for the rate of perfusion to promote lung injury may depend on the amplitude of the inflation, and vice versa.

Limitations of the isolated perfused lung model. We used an isolated perfused lung model to assess whether pulmonary blood flow affects the severity of ventilator-induced lung injury for two important reasons: precise control of flow and ability to continuously track both edema formation and the induction of altered permeability. Certain characteristics of this model, however, prevent direct extrapolation of our results to the intact animal.

The range of the flows studied was somewhat arbitrary. These perfusion rates, however, approximate 0.5 (300 ml/ min), 0.8 (500 ml/min) and 1.5 (900 ml/min) times the normal cardiac output of a 3-kg rabbit (20) and, percentagewise, are well within the range of variation in cardiac output encountered in critically ill patients. In vivo, cardiac activity generates a pulsatile flow profile. Theoretically, a pulsatile flow would be expected to increase the difference in vascular pressure observed during inflation with different flow intensities (21). Thus, if anything, the use of nonpulsatile flow here should bias the study against the hypothesis we tested. Venous return and cardiac output vary cyclically during positive pressure ventilation, due to complex cardiopulmonary interactions (22). In contrast, our pump maintained forward flow constant. Lung isolation tends to abolish lymphatic drainage and to amplify the consequences of edema formation. The addition of a prostacyclin inhibitor to the perfusate has the potential to alter the lung microcirculation and inflammatory response in vivo but, we believe, is unlikely to explain the differences observed between our study groups. Finally, although we used larger VT/ kg than those used in the clinical arena, in the setting of severe adult respiratory distress syndrome (ARDS), only a limited number of lung units ("baby lung") may receive the bulk of the delivered tidal volume, and the transpulmonary pressures used are quite comparable to those applied in this study. Using an isolated perfused lung, we demonstrated that increased flow enhances ventilator-induced lung injury; to determine whether increased cardiac output enhances ventilator-induced lung injury would require an intact animal model.

Assessment of tissue injury. Lung injury is not easily defined or quantified (23). Individual indices of injury have inherent limitations; for instance, weight gain does not distinguish between hydrostatic and permeability pulmonary edema, and lung compliance can be influenced by alterations in lung capacity as well as by tissue damage (24). We opted to assess lung injury using four different indexes of injury to strengthen any conclusion regarding the effect of flow on lung injury. Measurements of K_f have limitations (16) but allow the accurate assessment of permeability alterations (fluid conductance), provided that the pre- and postinjury measurements of K_f are obtained at an identical surface area for filtration, which tends to vary with lung volume and vascular recruitment. At CPAP of 5 cm H₂O, lung volume and vascular recruitment were likely to have been lower after injury than before, at least when lung compliance decreased significantly and hemorrhage was extensive (e.g., in the high-flow group). This may have caused underestimation of K_f in the latter group, which would not change the conclusions of the study. The primary purpose of the histology assessment was to determine if substantial capillary disruption (manifested by extravasation of red blood cells) resulted from mechanical ventilation. The fact that the high-flow group had significantly larger K_f and more extensive alveolar and extra-alveolar hemorrhage than the control group is consistent with this interpretation and makes random contamination of alveoli by red blood cells during tissue sampling very unlikely. The significant correlation between the scores of lung hemorrhage and weight gain also argues strongly against this hypothesis. It is thus very likely that overall lung hemorrhage was caused by vascular injury associated with ventilator-induced lung injury. The changes in vascular resistance resulting from inflation with positive pressure take place predominantly in the compartment located in series between large arterial and venous pulmonary vesselsthe so-called "intermediate segment" which includes the capillaries (25). It follows that although capillary pressures measured by double occlusion were identical in the high- and low-flow groups just prior to the beginning of ventilation, intramural capillary pressures were probably higher in the high-flow group than in the low-flow group during tidal inflation. This may be important, because transmural filtration and vascular breaks associated with tidal inflation occur in this same segment (13, 16). Longitudinal traction on capillaries during tidal inflation, in conjunction with increased capillary pressure at higher flow, can result in a greater vascular insult (13) and could explain the tight correlation found between the tidal changes in vascular pressures and the indices of lung injury. It remains possible but unlikely, we believe, that higher flow resulted in greater driving forces or a more extensive surface for red blood extravasation, regardless of the severity of capillary alterations per se.

Conclusions: Possible Relevance of the Study

The results of this study suggest that the dependent distribution of ventilator-induced lung injury previously observed in supine large animals might be, at least partially, explained by regional differences in blood flow and vascular pressure. Similarly, the more homogeneously distributed blood flow along the vertical axis in the prone position (8) could also help explain the more uniform distribution of lung water and histologic damage associated with ventilator-induced lung injury (2). Our finding that hemodynamics may interact with the airway pressure profile in the development of ventilator-induced lung injury suggests that differences among ventilatory patterns with regard to ventilator-induced lung injury may be due, at least partially, to differences in the hemodynamics. Our findings also have potential clinical implications. For instance, vasoactive drugs that alter the regional distribution of lung blood flow may either exert protective or detrimental effects on ventilator- induced lung injury, depending on how blood flow is redistributed. Similarly, medical interventions to increase blood flow to supraphysiological values may be ill advised when high transalveolar pressures are employed concomitantly. Until definitive clinical data are available, further experimental studies are needed to assess the interactions between vascular pressures, regional flow, cardiac output, and the airway pressure profile in generating ventilator-induced lung injury.