INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There is ample experimental evidence that mechanical ventilation of the lungs to volumes exceeding total lung capacity (TLC) produces diffuse alveolar damage (reviewed in Reference 1). Proposed mechanisms involve endothelial and epithelial stretch injury, capillary stress failure, tissue disruption, alveolar flooding, and alterations in the physicochemical properties of pulmonary surfactant (2). There is also considerable evidence that the application of positive end-expiratory pressure (PEEP) protects the lung from ventilator-induced injury (VILI), at least to some extent (2, 5, 6). Most attribute the protective action of PEEP to a reduction in tissue shear stresses caused by the repeated opening and collapse of unstable lung units. There are two lines of experimental evidence in support of such a mechanism. Using computer tomography (CT) Gattinoni and colleagues (9, 10) examined the topographic distribution of gas in injured lungs of patients and concluded that dependent lung units are collapsed. Consequently, it has been reasoned that during mechanical ventilation without PEEP, some of these units are subject to within-breath recruitment and derecruitment. Several groups have examined the morphologic and functional consequences of cyclic airway opening and collapse (7, 11). When normal lungs are forced to deflate to residual volume on every breath, epithelial "abrasions" appear in small airways. These can develop in conjunction with proteinaceous alveolar edema and are associated with the release of inflammatory mediators.

The interpretation of the CT data is based on the assumption that lung water (edema) is uniformly distributed and that the higher CT gray-scale (density) in dependent regions is, therefore, evidence of alveolar collapse. We set out to test this conclusion by directly measuring regional lung volumes and lobar expansion in oleic-acid-injured dogs. To this end, we used the parenchymal marker technique, which provides absolute measures of regional tissue dimensions, as opposed to relative measures of regional air to liquid content (12). Our goal was to test the hypotheses that (1) oleic acid (OA) injury leads to the collapse of dependent lung units at FRC, (2) OA injury steepens the vertical gradient in regional lung volumes at FRC, and (3) during mechanical ventilation without PEEP, dependent regions of OA-injured lungs undergo cyclic reopening and collapse. On the basis of our findings, we reject all three hypotheses and propose an alternative mechanism for the topographic variability in regional impedance and lung expansion, namely, liquid and foam in conducting airways.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Parenchymal Marker Technique

To measure regional lung expansion during mechanical ventilation, the parenchymal marker technique was used. It has been described in detail previously (12). Briefly, 1-mm metal beads were implanted transthoracically into the caudal lobe of anesthetized dogs. Nine weeks later, experiments were conducted during which biplane fluoroscopic images of the thorax were recorded on videotape. The orthogonal projection images of each bead were sampled at a rate between 1.5 and 5 Hz (depending on the experimental condition) using an operator-interactive computer tracking system. The three-dimensional marker locations were derived from these data. Four markers define a volume (tetrahedron) that contains air, tissue, edema fluid, and blood. Thus, as a final output, the method provides the description of regional volume behavior in time (i.e., regional spirograms). An average of 79 ± 51 (SD) tetrahedra were created for each animal. Their volume at TLC (defined as the volume at an airway opening pressure of 35 cm H₂O) before injury averaged 3.4 ± 0.8 cm³.

Animal Preparation and Data Acquisition

All techniques and procedures were approved by the Institutional Animal Care and Use Committee of the Mayo Foundation, and the care and handling of the animals was in accordance with National Institutes of Health guidelines. Nine adult beagle dogs of either sex (9.7 ± 1.5 kg) were studied. The animals were anesthetized with pentobarbital sodium (30 mg/kg, with supplemental doses of 100 mg/h), a tracheostomy was performed, and a 9-mm ID glass canula was inserted into the tracheal lumen. The dogs were mechanically ventilated (Siemens Servo 900C; Siemens, Solna, Sweden) with a fractional oxygen concentration of 1.0, a tidal volume (VT) of 230 to 250 ml, a respiratory rate of 20 cycles/min, an inspiratory time fraction of 0.33, and a positive end-expiratory pressure (PEEP) set between 3 and 5 cm H₂O. Airway opening pressure (Pao) was measured at the proximal end of the endotracheal tube (24A1 Microswitch; Honeywell, Freeport, IL). Distal esophageal pressure (Pes) was monitored by means of a balloon catheter attached to a pressure transducer (24A1, Honeywell Microswitch). The optimal position of the catheter was determined fluoroscopically and was verified using the balloon occlusion method. Transpulmonary pressure was calculated by subtracting Pes from Pao. Gas flow rates were measured using a pneumotachograph and transducer (163PC Honeywell Microswitch) attached to the Y-piece of the ventilator circuit. Flow was integrated to monitor the tidal volume. Gas exchange was assessed by periodic analysis of blood samples drawn from a femoral arterial line (blood gas analyzer IL-1304; Instrumentation Laboratories, Lexington, MA). A pediatric Swan-Ganz pulmonary artery catheter was inserted through the right internal jugular vein for monitoring of pulmonary artery pressure, cardiac output (thermodilution method, Spectromed Hemopro 1; Spectromed, Eden Prairie, MN), and pulmonary artery occlusion pressure. Systemic blood pressure was monitored from the femoral arterial line with a pressure transducer (Statham P37A; Statham, Oxnard, CA). Three leads attached to the paws provided a single electrocardiographic tracing to monitor the heart rate. Rectal body temperature was monitored (YSI Tele-thermometer 44TF; Yellow Springs Instrument Co., Yellow Springs, OH) and kept constant at 37° C using heat lamps. A 16-f canula was placed in a peripheral vein for administration of normal saline and an anesthetic. In six dogs intravenous fluid infusion was aimed at keeping a constant occlusion pressure between 7 and 10 mm Hg. In the remaining three animals, normal saline was administered only to prevent hemodynamic instability. Airway opening pressure, esophageal pressure, gas flow, tidal volume, pulmonary artery pressures, systemic blood pressures, EKG tracing, and notes taken during the experiment were all stored in a digital form (LabVIEW; National Instruments, Austin, TX).

Experimental Protocol

Seven animals were studied in the supine and two in the prone posture. They were paralyzed intravenously with pancuronium bromide, 3 mg, supplemented with 1 mg as needed. The inspiratory capacity was determined by inflating the lungs to an airway pressure of 35 cm H₂O. Measurements (including biplane thoracic images) were then taken during closed-circuit sinusoidal oscillations of the respiratory system using a locally engineered precision-calibrated piston pump. The pump was set to deliver a tidal volume of 200 ml at rates of 2.5 and 20 cycles/min.

After baseline measurements, all animals were turned prone and injected with 0.09 ml/kg OA in three aliquots via the right atrial port of the pulmonary artery catheter. The prone posture has been shown to minimize topographic gradients in pulmonary blood flow (13). Animals in the supine group were repositioned 5 min after the last OA injection. Three of seven supine animals were filmed 30, 60, and 90 min after OA administration while being mechanically ventilated without PEEP. Starting 90 min after the OA injection, all measurements were repeated, including whole lung mechanics, gas exchange, and regional volume expansion while the lungs were sinusoidally oscillated with and without PEEP.

At the end of the study, animals were killed with an overdose of pentobarbital sodium. The lungs were excised, weighed, and air-dried at a transpulmonary pressure of 25 cm H₂O, after which they were weighed again to determine the wet-to-dry ratio and the dry-to-body weight ratio.

Data Analysis

Data for animals studied supine are reported as means ± standard deviations. The small number of dogs in the prone group precludes a statistical analysis of posture effects. When appropriate, two-tailed paired t-tests and linear regression methods were used to examine the effects of OA injury on regional lung function. Statistical significance was assumed at a probability  0.05.

Respiratory system mechanics. The mechanical properties of lungs, chest wall, caudal lobe, and regions within a lobe were measured during sinusoidal oscillations of the closed respiratory system at the rate of 20 cycles/min. Pressure components in phase and out of phase with mouth volume were computed; the appropriate elastances and resistances were derived by fit of a linear model to the pressure-volume data (14). Volume of lobes and regions within a lobe are presented as fractions of their preinjury TLC volumes. The apparent regional elastances and resistances are referenced to the overall transpulmonary pressure, i.e., any variance between them reflects the variance in regional volumes and flows.

Regional lung expansion. Each region's FRC, mean volume, and oscillation amplitude (regional VT) were computed from its spirogram. The average of all regions in a given dog was taken as a representative of the entire lobe. The vertical positions of the centroids of the tetrahedra were computed and from them gradients in regional lung function were determined.

Analysis of cyclic opening and collapse of regions during sinusoidal oscillation. Regional spirograms were inspected for significant departures from sine wave in order to detect temporal heterogeneity from the delayed "popping open" of regions during lung inflation. Regional volume versus mean lobar volume and mouth volume plots were inspected for nonlinearities and hysteresis. The temporal heterogeneity in lobar expansion was later quantified using a cross-correlation approach. Each region's spirogram was correlated in time with the mean lobar spirogram, yielding a correlation coefficient R. Because R is a function of phase-matching between the two time series (spirograms), this process was repeated after displacing spirograms by 0.25 or 0.05 s (depending on cycling rate) relative to each other. Lead or lag times of regions relative to the lobe correspond to the time shifts required to maximize the correlation between the two signals. The 95% confidence intervals in time shifts of all regions within a lobe were taken as measures of temporal heterogeneity in lobar expansion. Among all dogs, an average of 3.5% of all regions examined showed no volume change at the oscillation frequencies of 2.5 and 20 cycles/min. These regions were excluded from cross-correlation analysis.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Hemodynamics, Gas Exchange, and Lung Weight

The effects of injury 90 min after OA administration on hemodynamics and gas exchange and the results of the wet-to-dry and dry-to-body weight ratio measurements are summarized in Table 1. Results from prone and supine dogs are shown separately. Over the course of the experiment (during which the animals received between 0.25 and 5.5 L of normal saline), OA injury produced alveolar edema. This became manifest radiographically within 30 min. Foamy thin secretions appeared in the central airways towards the end of the study. Wet-to-dry lung tissue ratios ranged from 7.5 to 16.0, averaging 10.7 ± 2.7 in the nine dogs. Injury caused a marked deterioration in pulmonary gas exchange accompanied by respiratory as well as metabolic acidosis. Blood pressure was generally well maintained, but most animals became oliguric during the latter stages of the study.

Respiratory System Mechanics

The effects of OA injury on elastance and resistance of the lungs and the chest wall are shown in Figure 1. In supine animals, OA caused an increase in lung elastance from 26 ± 10 to 138 ± 50 cm H₂O/L (p < 0.01) and lung resistance from 2.7 ± 0.7 to 20.2 ± 15.2 cm H₂O/L/s (p = 0.02). However, neither the elastance nor the resistance of the chest wall was affected by OA injury. The OA-associated increase in respiratory impedance appeared to be smaller in the prone posture than in the supine position.

View larger version (18K): [in this window] [in a new window]   Figure 1.   Effect of oleic acid (OA) injury on lung and chest wall mechanics. Data were collected at the beginning of the experiment and 90 min after OA injection during sinusoidal oscillation of the respiratory system at 20 cycles/min with zero PEEP. Probability p refers to injury effects.

Lobar and Regional FRC

Because of the location of the caudal lobe in the thorax, in animals studied supine, most parenchymal markers were in a dependent part of the lung. In contrast, in prone animals, most markers were in a nondependent part of the lung. The effects of injury on a representative regional spirogram of one supine and one prone dog are illustrated in Figure 2. Before injury each region expanded sinusoidally. Consistent with the known effects of posture on regional lung mechanics, the FRC of the nondependent region in the prone dog was greater than that of the dependent region in the supine dog. After injury the dependent region of the supine dog ceased to expand during lung inflation, whereas the oscillation amplitude of the nondependent region in the prone dog increased. Because the overall volume of gas delivered to the lungs was kept constant, these changes reflect a redistribution of regional ventilation with injury. Of interest is the observation that the FRC of the derecruited region increased after injury even though it presumably had to bear the weight of the superimposed edematous lung.

View larger version (18K): [in this window] [in a new window]   Figure 2.   Representative examples illustrating the effect of oleic acid (OA) injury on regional volume expansion. Data were obtained from one prone and one supine dog representing a nondependent and a dependent lung region, respectively. Measurements were made before and 90 min after OA injection during sinusoidal oscillation of the lungs at 20 cycles/min in the absence of PEEP. Regional volumes are expressed as the fraction of each tetrahedron's volume measured at baseline TLC. Changes in regional volumes with injury reflect net changes in regional gas and liquid volume (consisting of intra-alveolar and extra-alveolar blood and edema).

To emphasize this observation further, the effects of OA injury on lobar FRCs in all nine dogs are shown in Figure 3. The two prone dogs experienced a slight fall in lobar volumes from 0.54 to 0.5 (expressed as fractions of baseline TLC), whereas the lobar FRCs of the seven supine animals did not change (0.38 versus 0.39, p < 0.53).

View larger version (13K): [in this window] [in a new window]   Figure 3.   Effect of oleic acid (OA) injection on lobar FRC. Lobar volume is an average behavior of all regions (i.e., tetrahedrons) within each lobe. Data were collected before and 90 min after OA injection during sinusoidal oscillation of the lungs at 20 cycles/min without PEEP. Lobar FRC is expressed as the fraction of lobar volume measured at baseline TLC.

The lobar FRCs of three supine dogs measured 0, 30, 60, and 90 min after OA administration are shown in Figure 4. During the evolution of OA injury, lobar FRCs remained near constant at 0.38 ± 0.09, 0.37 ± 0.07, 0.37 ± 0.07, and 0.38 ± 0.06, respectively. Of note, during this time the dogs received 0.25 to 1 L of normal saline (0.50 ± 0.43 L) and voided 0.02 to 0.16 L of urine (0.07 ± 0.08 L).

View larger version (12K): [in this window] [in a new window]   Figure 4.   Changes in lobar FRC following OA injection. Data in three supine dogs were collected before OA as well as 30, 60, and 90 min after OA lung injury during mechanical ventilation at 20 cycles/min without PEEP. Lobar FRC is expressed as a fraction of lobar volume measured at baseline TLC.

The FRCs of all regions as a function of their vertical position within the thorax are shown in Figure 5. Ranges in vertical positions of tetrahedra averaged 1.81 ± 0.59 cm. Before OA injury, the average vertical gradient in regional FRC among supine dogs was 3.3 ± 4.3 per cm difference in vertical height. Differences in regional FRC between nondependent and dependent portions of the caudal lobe were observed in five of seven supine dogs. Gravitational gradients in regional FRC were not altered after OA injury, changing by less than 0.5 per cm difference in vertical height in either posture (p < 0.82).

View larger version (20K): [in this window] [in a new window]   Figure 5.   Regional FRC as a function of vertical position. FRC measurements were made during sinusoidal oscillation of the lungs at 20 cycles/ min in the absence of PEEP, before, and 90 min after oleic acid (OA) injection. Regional FRCs are referenced to the spatial coordinate values of each region's (i.e., tetrahedron's) centroid. Regional FRC is expressed as the fraction of each region's volume measured at baseline TLC.

Amplitude of Regional Expansion during Sinusoidal Oscillation

The effects of OA injury on the frequency distributions in regional lung expansion (regional VT) during sinusoidal oscillations from FRC are illustrated in Figure 6. After injury the expansion of nondependent lung regions of prone animals increased on average from 0.16 to 0.19, whereas that of dependent regions of supine dogs fell from 0.14 ± 0.02 to 0.06 ± 0.02 (p < 0.0001). Regional tidal volumes were approximately normally distributed before as well as after OA injury. There were no consistent gravitational gradients in regional VT in either posture (data not shown).

View larger version (29K): [in this window] [in a new window]   Figure 6.   Effects of oleic acid (OA) injury on the frequency distributions in regional lung expansion (regional VT). Measurements were made before and 90 min after OA injection, during sinusoidal oscillation of the lungs at 20 cycles/min without PEEP. Regional VT is expressed as the fraction of each region's volume measured at baseline TLC.

Temporal Heterogeneity in Regional Expansion

The results of the cross-correlation analyses for two oscillation rates, 2.5 and 20 cycles/min, are summarized in Figure 7. Before injury, when cycled at a rate of 2.5/min, 95% of the regions oscillated within 1.0 ± 0.8 s of each other. OA injury had no significant effect on the temporal heterogeneity of lobar expansion, raising it to 1.5 ± 1.7 s (p = 0.27). The same was true for cycling rates of 20/min, during which 95% of all regions oscillated within 0.08 ± 0.04 before injury and 0.11 ± 0.05 s after injury (p = 0.37). The fact that after OA injury 95% of all injured regions expanded within 23 and 13 degrees of each other at oscillation frequencies of 2.5 and 20 cycles/ min, respectively, indicates relatively homogeneous behavior of the regions sampled.

View larger version (14K): [in this window] [in a new window]   Figure 7.   Effects of oleic acid (OA) on the temporal heterogeneity in regional lung expansion at two cycling rates (2.5 and 20 cycles/ min). Temporal heterogeneity in regional lung expansion refers to the 95% confidence interval of lead and lag times of regions (in seconds) relative to the lobar mean. Data were obtained during sinusoidal oscillation of the lungs without PEEP before and 90 min after OA injection. The average ± SD of the 95% confidence intervals refers to supine animals only, at both cycling rates.

Effect of PEEP on Regional Expansion

A representative example of the effects of PEEP on volume and ventilation of a dependent region is shown in Figure 8. OA injury causes derecruitment of this region during sinusoidal oscillation of the lungs without PEEP. The application of 7.5 cm H₂O PEEP for 20 min was associated with the reappearance of low amplitude volume oscillations. Regional FRC was not altered by 7.5 cm H₂O PEEP. The subsequent application of 15 cm H₂O PEEP for 20 min reestablished normal tidal oscillations and raised the region's FRC from 0.36 to 0.43. 

View larger version (25K): [in this window] [in a new window]   Figure 8.   Representative example of the effects of oleic acid (OA) injury and PEEP on airway pressure and the spirogram of a dependent region. Data were collected during sinusoidal oscillation of the lungs at 20 cycles/min before OA without PEEP as well as after OA injury at PEEP settings of 0, 7.5, and 15 cm H2O.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main findings of our study can be summarized as follows. (1) The gravitational distribution of volume at FRC is not affected by OA injury. In particular, OA injury is not associated with decreased parenchymal volume of dependent regions. (2) The temporal inhomogeneity of regional tidal expansion does not increase with OA injury. These two results are contrary to the current paradigm of VILI, according to which a gravitational gradient in "superimposed pressure" produces compression atelectasis of dependent lung that in turn produces shear injury from cyclic recruitment and collapse. (3) Although regional volume is not affected by OA, regional tidal expansion is altered; regional tidal expansion of dependent regions is significantly reduced after OA injury. (4) Regional tidal expansion of dependent regions is restored by the application of PEEP. These results are consistent with other published data, but the mechanisms that cause the changes in regional tidal volume cannot be those invoked in the current paradigm, namely, alveolar collapse and reopening.

Before we return to discuss the central findings of this study and relate them to the shear injury hypothesis we will address potential methodologic and experimental limitations of our study.

The Canine Oleic Acid Injury Model

The OA injury model is commonly used as an animal model of acute lung injury (15). OA is an unsaturated fatty acid that binds to cell membranes at low concentrations and ultimately produces cell death. The resulting lesion consists of hemorrhagic edema and diffuse alveolar damage with injury to both endothelial and alveolar epithelial cells. When OA is injected into a central vein it damages the pulmonary microvasculature in a perfusion distribution-dependent manner. In the hope of producing uniform injury, all animals in this study were injected prone because perfusion tends to be most uniform in that posture (13). However, seven of nine dogs were repositioned supine shortly after OA injection. In them, topographic gradients in blood flow and microvascular hydrostatic pressure probably explain the hemorrhagic appearance of the paraspinal portions of the caudal lobe at autopsy. Perhaps the modest amounts of PEEP over long periods also contributed to dependent lung injury in supine animals. Posture dependent differences in the topographic distribution of edema and inflammation have been described in high tidal volume injury models of OA preinjured dogs (16).

Notwithstanding the popularity of the OA model in lung injury research, the findings of this study may have limited relevance for human disease. At least in our hands, OA injection in dogs produces more hemorrhagic edema than is typically seen in patients with the acute respiratory distress syndrome (ARDS). Moreover, atelectasis and cyclic opening and collapse cannot be dismissed as possible injury mechanisms in neonatal respiratory distress syndrome and in surfactant depletion models (2, 7, 8). Nevertheless, the findings of this study motivated us to take another look at the strength of the evidence upon which currently accepted mechanisms of regional lung function in ARDS are based.

Parenchymal Marker Technique

The parenchymal marker technique resolves volume changes of tetrahedra (lung regions) at a scale of  1 cm³ with a temporal resolution of  30 Hz. In contrast to CT, the technique is well suited to the measurement of regional parenchymal expansion because it directly provides data that follow the behavior of parenchymal elements as the lung moves. CT images of the thorax are density maps from which the topographic distribution of air per unit tissue volume may be estimated. The images do not define tissue state, and they give no information on tissue extension or strain. It is usually not possible to track a specific structure such as an alveolus or a small airway bifurcation across multiple CT images in time. Therefore, it is not possible to measure the amplitudes of regional lung expansion unless one is willing to ignore errors resulting from the misalignment of parenchymal structures across images obtained at different phases of the respiratory cycle.

The parenchymal marker technique has different limitations. The technique is labor intensive, it is often difficult to distribute markers uniformly within a lobe, it is usually not possible to sample more than two lobes per animal, and regional volumes reflect tissue expansion rather than regional air content. Considering these limitations in the context of the three hypotheses that were stated at the outset, the between-dog variability in topographic gradients of volume and ventilation deserves particular scrutiny.

The range of vertical centroid distribution at FRC varied among dogs, from 0.9 to 2.7 cm. The variability in vertical gradients that we observed may have been a result of the limited vertical range of the data. It should also be noted that many tetrahedra were overlapping, i.e., the central portions of the caudal lobe were oversampled relative to its periphery. Therefore, the lack of apparent effect of injury on the topographic distribution of regional volumes should carry the least weight in our consideration of mechanisms.

Effects of Injury on Lung Impedance and Topographic Distribution of FRC

The most striking finding of this study was the preservation of regional lung dimensions after OA injury. In supine animals, injury led to a derecruitment of tidal expansion of dependent lung regions but not to their collapse. Dependent regions may have been airless, underventilated, or not ventilated at all, but they were not atelectatic. Quite to the contrary, dependent lung units were slightly expanded. In all likelihood the alveolar spaces were filled with blood and edema.

There is a considerable body of literature on the mechanical properties of edematous lungs. Notwithstanding differences between species and experimental models, injury and edema raise both pulmonary elastance and resistance (10, 15, 17). Our results depicted in Figure 1 are consistent with these reports. Numerous mechanisms have been proposed to explain the change in lung mechanics in the presence of edema. These include increased minimal surface tension caused by surfactant inactivation (for review see reference 3), airway block caused by air-liquid interfaces and bubble formation in small airways (17, 21), reflex bronchoconstriction (24), pneumoconstriction reflecting inflammatory mediator release (25), and peribronchial edema (22). Knowing that dependent lung tissue of OA-injured dogs is not collapsed and that there is massive edema and foam in conducting airways, it is hard to ignore air-liquid interfacial tensions as an important determinant of increased regional and whole lung impedance. Flooding of dependent lung is compatible with the "baby lung" concept (10), according to which the increased lung elastance reflects in part the reduced number of recruitable lung units (10, 26).

Current concepts about regional lung expansion in ARDS rest on static density measurements of the lungs with CT (9, 10, 27). Accordingly, the injured lung is thought to have three compartments: a nondependent compartment of air-filled and recruited lung units, a "middle" compartment of "compressed" but potentially recruitable lung units, and a dependent compartment of nonrecruitable lung units. The separation of the compartments along gravitational lines has been considered evidence that increased lung density and density gradients account for "compression" and derecruitment of dependent lung regions. However, this interpretation rests on the assumption that lung water (including alveolar edema) is uniformly distributed. In light of the gravitational gradients in pulmonary capillary flow and perfusion pressure, this assumption seems difficult to uphold (13). Furthermore (at least under disease-free conditions), the hypothesis that the effect of gravity on the lung is the principal determinant of regional volume and stress distribution has been largely rejected (28).

Certainly, the increased weight of the edematous lung would be expected to alter the lung surface pressure distribution, raising the vertical gradient in pleural pressure toward 1.0 cm H₂O/cm vertical height. Models in which it is assumed that gravity acting on the lung is the main determinant of regional lung expansion predict a corresponding volume gradient (and conclude that the dependent lung must be compressed). However, there may also be a hydrostatic gradient in alveolar pressure as a result of the accumulation of fluid in dependent contiguous air spaces. The preservation of regional lung dimensions in the absence of a significant redistribution of regional FRCs is consistent with this possibility.

It remains to be established whether the lung dimensions at FRC are preserved in other models of ARDS. Although the term atelectasis is used liberally in reports describing injured lungs, we have not found convincing morphometric evidence of alveolar collapse (as opposed to flooding) in the published literature. In part this reflects the difficulties in preserving the in situ lung architecture during fixation and/or removal of the lungs from the chest (31). It can be argued that extensive atelectasis should readily develop in lungs with abnormal or reduced surfactant (3). However, this need not be the case if surfactant dysfunction occurs in conjunction with increased capillary permeability. As surface tension rises, the collapse of an alveolus is opposed by interdependence mechanisms (32). The resulting fall in interstitial pressure promotes fluid egress from alveolar corner vessels. In the presence of impaired lymphatic drainage and damage to the alveolar epithelial cell layer, fluid is likely to enter the alveolar space. Once the alveolus begins to flood and the mean radius of curvature of the air-liquid interface in it falls, local micromechanics dictate that a decreasing alveolar liquid pressure draws additional fluid into the space until the air-liquid meniscus moves to the alveolar entrance. Given this interplay between surface tension, interdependence, and pulmonary capillary filtration coefficient, the probability of alveolar collapse as opposed to flooding should decrease with increasing vascular injury.

Effects of Injury on Regional Ventilation

Consistent with the "baby lung concept" in the supine dogs, OA injury caused a marked reduction in the amplitude of volume oscillation in the dependent caudal lobe (Figures 2 and 6). In contrast, in the prone posture the oscillation amplitude of the nondependent caudal lobe remained the same or increased. Because the parenchymal marker technique cannot resolve gas from blood or edema fluid, it is important to remember that volume changes of dependent lung regions may have been caused by blood and edema fluid shifts as opposed to ventilation per se. Irrespective of this uncertainty, regions that by virtue of their small oscillation amplitudes appeared derecruited in the absence of PEEP regained normal oscillation amplitudes when PEEP of 15 cm H₂O was applied for 20 min (Figure 8).

The model that depicts dependent lung as compressed by the weight of edematous superimposed parenchyma predicts that there is a gravitational gradient in regional opening pressures. During mechanical ventilation, gas is first directed toward nondependent aerated lung units. Later in the breath, some compressed regions begin to fill as their "superimposed hydrostatic pressure" is counterbalanced by a rising airway pressure. Consequently, the model predicts sequential regional filling along a vertical axis, namely, temporal heterogeneity in lobar expansion (27). According to the hydrostatic model, these mechanisms are thought to determine the shape of the inflation pressure-volume (P-V) curve and explain the cyclic recruitment and collapse of units during breathing, which promotes shear injury.

Our observations on OA-injured dogs are not consistent with this model. As already discussed, dependent lung tissue is expanded not collapsed. Furthermore, there was not a great deal of asynchrony in lobar expansion during sinusoidal oscillations of the lungs. In trying to find evidence of cyclic reopening and collapse during mechanical ventilation, we undertook three analyses. The first two were largely semiquantitative. Regional spirograms were screened for departures from sine waves. Regions with discernible oscillation amplitudes in the respiratory frequency range were examined for delayed filling. None were identified. Using similar reasoning, we plotted regional volumes against mean lobar volume and volume delivered to the airway. Delayed filling and/or reopening and collapse should have produced nonlinearities and hysteresis in these relationships. Again, none were found. To quantify temporal heterogeneity less subjectively, we chose a cross-correlation approach, as its validity does not depend on the choice of the statistical model with which to fit the time series data. Using 95% of the distribution in regional lead and lag times relative to the lobar mean as a measure of temporal heterogeneity, we arrived at the same conclusion: lobar expansion was remarkably synchronous and did not change substantially following OA injury in either posture (Figure 7). This was true irrespective of cycling rate.

This result may seem surprising because interrupter mechanics measurements of patients with ARDS display nonlinear dynamics (18, 26, 33). Experimental data indicate that both airway and tissue components of the total respiratory system resistance are increased, that airway resistance decreases with tidal volume but not with flow, and that tissue resistance has a greater than normal flow and volume dependence. However, these observations do not prove any one mechanism to the exclusion of others. Flow and, thus, frequency dependent "tissue resistance" could just as easily reflect a dissipation of pressure across foam and air-liquid interfaces in conducting airways as it could indicate small-scale heterogeneity in airway dimensions, local surface tension, and/or the degree of pneumoconstriction.

Proposed Mechanisms of Impaired Regional Ventilation after OA Injury and Its Implications

Although the findings of this study cannot be reconciled with a model in which dependent lung is compressed by its own weight, they are consistent with alveolar trapping by liquid and foam in conducting airways. For small applied forces, foam is an elastic material. That is, for small driving pressures, foam in an airway would deform slightly but would not flow. Foam yields and begins to flow at a critical shear stress. Typical values of critical stresses in foam are of the order of a few tenths of a cm H₂O (34). However, flow of foam through an airway branch requires simultaneous yield on more than one surface, and the pressure required to drive foam through a branch may be many times the yield stress. The theory of plasticity shows that the pressure required to drive flow of a perfectly plastic material is independent of the scale of the flow. Thus, the pressure drop across a branch is the same for all sizes of tubes. Therefore, the pressure drop required to drive flow through a self-similar branching network of n generations is n times the pressure drop required to drive flow across a single branch, and the pressure drop that would be required to drive flow through the bronchial tree may be quite large.

Considering the mechanics of injured lungs and their interactions with positive pressure ventilation from the perspective of moving air-liquid interfaces through conducting airways rather than forcing collapsed tubes open generates alternative hypotheses. For one, energy dissipated in foam as opposed to tissue is less apt to produce lung damage and inflammation. The two studies in which tissue shear has been most directly implicated as a cause of VILI have imposed experimental conditions that may not prevail in a clinical setting (7, 11). For example, Muscedere and colleagues (7) showed that mechanical ventilation of isolated unperfused rodent lungs produces epithelial abrasions in small airways. Allowing the lungs to repeatedly collapse ex vivo to residual volume and minimizing the chance for alveolar flooding by eliminating perfusion raises questions about the relevance of the histologic findings. Although there is strong experimental evidence that PEEP has a protective effect on the injured lung, the mechanisms need not involve the prevention of shear stresses (2, 5, 6, 8). It may be that PEEP-related improvements in alveolar O₂ tension or the reestablishment of surface tension in alveoli has beneficial effects on local cell function and injury repair processes.

Another hallmark of injured lungs, increased P-V hysteresis, has been attributed to alveolar recruitment and derecruitment, which fits well within the conceptual framework underlying "best PEEP" trials (5, 20). However, the shape of the whole lung P-V curve cannot be attributed to a single mechanism. One need only examine the P-V curve of newborn animals as they take their first breath to realize that displacing air-liquid interfaces along the tracheo-bronchial tree results in a P-V curve with a prominent inflection point (35).

In summary, we conclude that the experimental evidence on the determinants of regional mechanics of injured lungs (be it based on CT imaging, lung histology, morphometrics, or P-V behavior) is consistent with a mechanism that centers around alveolar flooding and liquid movement in airways. The evidence we have collected on OA-injured dogs is also consistent with this perspective. However it is not consistent with a paradigm that views the dependent lung as being compressed by its own weight.