INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Inverse ratio ventilation (IRV) has been proposed for the treatment of patients with acute respiratory distress syndrome (ARDS) because it may improve gas exchange over that achieved with conventional ventilatory modes (1). However, more recent reports on gas exchange during IRV in both experimental (5) and clinical studies (8) have yielded variable results. The cause of improvement occurring with IRV, and the variable results in different studies, are not fully understood. One mechanism by which IRV improves gas exchange is so-called intrinsic positive end-expiratory pressure (PEEP_i), which occurs if the expiration time is too short for a complete exhalation (12). PEEP_i increases the end-expiratory lung volume (EELV) (13) and may thereby prevent or reduce end-expiratory lung collapse. However, PEEP_i is not uniformly distributed throughout the lungs, since it is affected by the regional compliance and resistance (time constant) of lung units. Consequently, in lungs with markedly different time constants, aeration may be more heterogenous with PEEP_i than with the use of extrinsic PEEP (PEEP_e), which could explain some of the conflicting results obtained with IRV.

In addition, airway pressure transmitted to the pulmonary vascular bed has been shown to affect the distribution of pulmonary blood flow (14). This may affect gas exchange, such as if blood flow is preferentially redistributed toward poorly aerated or collapsed lung regions, and such an effect may differ in IRV and ventilation with more conventional settings of the inspiratory-to-expiratory (I:E) ratio.

In the present study we investigated whether any differences in gas exchange with conventional mechanical ventilation with PEEP_e as opposed to IRV with PEEP_i could be attributed to differences in the distribution of aeration of the lung and/or differences in the distribution of lung blood flow, and what effect such differences might have on the ventilation-perfusion (A/) relationship.

Knowing from previous studies that lung collapse during expiration was more or less avoided with PEEP > 20 cm H₂O or with an expiratory time (TE) interval of < 0.6 s regardless of PEEP (15), we compared ventilation with a PEEP_e of 20 cm H₂O and to that with IRV with an TE interval of 0.5 s. We hypothesized that these settings should lead to similar oxygenation if IRV and conventional mechanical ventilation have similar effects on aeration (16) and pulmonary blood flow distribution. In contrast, a TE interval of 1 s without application of PEEP should result in greater end-expiratory lung collapse and worse oxygenation than with either of the ventilatory patterns mentioned previously, and we also investigated this in our study to further test our hypothesis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Protocol

After approval of the study by the local animal ethics committee, we anesthetized and mechanically ventilated 26 healthy pigs belonging to a mixed breed of the Hampshire, Yorkshire, and Swedish country breeds and weighing 30 ± 4 kg (mean ± SD). Six animals died during ventilation with the control mode after induction of lung injury (see the subsequent discussion), leaving 20 pigs that completed the study. The sample size was based on the findings of a previous study. A minimum of 15 animals were needed to detect a difference in Pa_O2 of 50 mm Hg after induction of lung injury, with an  error of  5% and a power of 95%.

Baseline values of hemodynamic and ventilatory parameters were obtained after a 30-min stabilization period, after which lung injury was induced with oleic acid injected via a central venous catheter. After a further stabilization period of 120 min, two or three different ventilatory modes (see the subsequent discussion) were applied sequentially in random order for 1 h each. A 30-min control period of conventional mechanical ventilation without PEEP_e (used during induction of lung injury) followed each ventilatory mode, in order to produce a consistent volume history. Hemodynamic and ventilatory measurements were made at the end of each study period. In 15 pigs the A/ relationship was studied through the multiple inert gas elimination technique (MIGET) during ventilation with PEEPe and with two different modes of IRV in random order. After application of the third ventilatory mode, the pigs were transferred to the radiology department and transverse computed tomographic scans of the chest were obtained during ventilation with that specific mode. This resulted in a total study period of approximately 8 h.

In another five pigs, the spatial blood flow distribution was studied with perfusion scintigraphy during ventilation without PEEP, with PEEP = 20 cm H₂O, and with one mode of IRV (TE interval of 0.5 s; see the subsequent discussion). For technical reasons it was not possible to conduct both computed tomography (CT) and to make scintigraphic measurements, or to do more than three scintigrams, in the same pig.

The hemodynamic and respiratory investigations were performed in the experimental laboratories of the Department of Clinical Physiology and in the Department of Diagnostic Radiology at the University Hospital of Uppsala.

Anesthesia

Forty milligrams of azaperonum (Stresnil; Janssen, Beerse Belgium) were given intramuscularly as premedication before transport of animals from the farm to the laboratory. General anesthesia was induced with atropine (0.04 mg/kg), tiletamin/zolazepam (Zoletil; Reading, Carros, France) (6 mg/kg), and xylazin (Rompun; Bayer, Leverkusen, Germany) (2.2 mg/kg), given intramuscularly and followed by a constant infusion of 400 mg/h clomethiazole (Heminevrin; Astra, Södertäje, Sweden), 150 µg/h fentanyl, and 2.5 mg/h pancuronium bromide for muscle relaxation. Additional fentanyl and pancuronium bromide were given as needed. The animals were endotracheally intubated and ventilated through a cuffed tube.

Before making baseline measurements, we infused 1,000 ml Ringer's acetate (Pharmacia AB, Stockholm, Sweden) at body temperature; thereafter, fluid replacement was directed at maintaining a constant hemoglobin value and a stable systemic arterial blood pressure. This resulted in an average infusion rate of 40 ml/h/kg after induction of lung injury.

Ventilation

Control mode. Mechanical ventilation was initiated in the pressure-controlled mode (pressure controlled ventilation; PCV) (Servo 300; Siemens Elema, Lund, Sweden) at a respiratory rate (RR) of 20 breaths/min, an I:E ratio of 1:2, fraction of inspired oxygen (FI_O2) of 1.0, PEEP of 0 cm H₂O, and inspiratory airway pressure (Paw_high) that resulted in a tidal volume (VT) of approximately 15 ml/kg. Paw_high was then adjusted to achieve normocapnia (Pa_CO2 = 35 to 45 mm Hg) under guidance with end-tidal CO₂ monitoring (Capnomac Ultima; Datex Instrumentation Corp., Helsinki, Finland) and intermittently obtained arterial blood samples.

After induction of lung injury, Paw_high was increased to the extent that VT was raised by a maximum of 20% above baseline in order to maintain normocapnia. If hypercapnia developed despite this increase in VT, a Pa_CO2 of 60 mm Hg was allowed before Paw_high was increased further.

Following this, IRV and conventional PCV were matched for:

1. Similar degrees of atelectasis. A TE of 4 s with PEEP = 20 cm H₂O (PEEP₂₀, see the subsequent discussion) and TE of 0.5 s regardless of PEEP (IRV₃₀, as subsequently discussed) have been shown to almost completely prevent end-expiratory lung collapse (15).
2. Aveolar ventilation. Normoventilation was achieved by adjusting Paw_high.

Accordingly, preselected factors were the use of PCV, an FI_O2 of 1.0, RR, I:E ratio, and PEEP level. The settings used were:

1. PEEP = 20 cm H₂O (PEEP₂₀), with an I:E-ratio of 1:2, an RR of 20 breaths/min, and a Paw_high that resulted initially in a VT of 15 ml/ kg. Paw_high was then adjusted to achieve normocapnia.
2. IRV with an RR of 30 breaths/min (IRV₃₀). The RR with IRV was 30 breaths/min and the I:E ratio was 3:1. This resulted in an inspiratory time (TI) and TE of 1.5 and 0.5 s, respectively. In order to adjust for the higher RR with IRV as compared with PEEP₂₀, a Paw_high that would result in a VT of 10 ml/kg was initially chosen, and was then adjusted to achieve normocapnia (Pa_CO2  40 mm Hg). No extrinsic PEEP was applied.
3. IRV with an RR of 15 breaths/min (IRV₁₅). The RR was 15 breaths/min and the I:E ratio was 3:1. This resulted in a TI and TE of 3 s and 1 s, respectively. This respiratory pattern was chosen because more lung collapse was expected to occur during the longer expiration than that in the second ventilatory protocol (15), although the same I:E ratio was used. Paw_high was adjusted to achieve normocapnia. No extrinsic PEEP was applied.

A block design was used for randomization of animals to the three ventilatory modes, so that a balanced distribution was obtained for the mode that was applied first, second, and third. This was necessary because CT was performed only with the last ventilatory mode used for each animal, for technical reasons.

Lung Injury

Oleic acid (Apoteksbolaget, Göterborg, Sweden), suspended in 20 ml isotonic saline, was injected slowly (over a period of 20 min) in a volume of 0.1 ml/kg via the central venous catheter. If the oxygen saturation (Sa_O2) decreased below 85% during the injection, no further oleic acid was given. During injection, blood pressure was stabilized with titrated doses of adrenaline.

Ventilatory Parameters

VT and minute ventilation (E) were recorded by the flow sensors in the ventilator. Airway pressure (Paw) and flow were measured on the inspiratory side of the ventilator and recorded on a personal computer for on-line signal processing, taking gas compression within the ventilatory circuit into account (Labview version 3.1 software; C-O Sjöberg Engineering, Stockholm, Sweden). Static compliance (Crs) of the total respiratory system (lung and chest wall) was determined during an inspiratory hold maneuver (duration  4 s) followed by an expiratory hold maneuver (duration  4 s), and was corrected for PEEP_i. Crs was calculated as VT divided by the inspiratory plateau pressure (Paw after > 2 s of inspiratory hold, when the plateau pressure had become stable) minus the end-expiratory airway pressure (Crs = VT [Paw_plateau  Paw_{endexpiration}]). The mean value of Crs in two inspiratory hold maneuvers was used for statistical evaluation. PEEP_i was determined at the end of the expiratory hold maneuver as PEEP_i = Paw_{endexpiration}  PEEP. Since PCV resulted in a nearly instantaneous increase and decrease in Paw, mean airway pressure (Paw_mean) was estimated as (Paw_high · TI + Paw_{endexpiration} · TE) · (TI + TE)¹.

Hemodynamics

For pressure measurement and arterial blood sampling, an 18-gauge catheter was inserted in the left carotid artery, together with a thermistor-tipped fiberoptic catheter (Pulsiocath 4F FT PV 2024; Pulsion Medical Systems, Munich, Germany) that was advanced into the descending aorta for measurements of cardiac output (CO), intrathoracic blood volume (ITBV) and extravascular lung water (EVLW). A Swan- Ganz catheter and an 18-gauge catheter were introduced into the right external jugular vein. The exact position of the catheters was confirmed by pressure tracing as well as radiologically during CT scanning.

Systemic, pulmonary arterial, and central venous pressures were displayed on a bedside monitor, together with Sa_O2 (Series 7010, Tram; Marquette Electronics Inc., Milwaukee, WI), and were recorded with reference to atmospheric pressure at the midthoracic level at end-expiration. CO and EVLW were determined randomly within the respiratory cycle by injection of 8 to 10 ml of a double indicator bolus consisting of 1 mg/ml indocyanine green (ICG-Pulsion; Pulsion Medical Systems) mixed in sterile water (temperature ~ 5 to 7° C). The thermistor-tipped fiberoptic catheter in the descending aorta generated the dye-dilution and temperature curves, CO, ITBV, and EVLW are automatically calculated by a computer connected to the catheter (COLD Z-021; Pulsion Medical Systems). The mean of triplicate measurements was calculated and used for statistical evaluation. Oxygen delivery (O₂), oxygen consumption (O₂), systemic vascular resistance (Rsv), and pulmonary vascular resistance (Rpv) were calculated with standard equations.

Gas Exchange and A/ Relationship

Arterial and mixed venous blood gas samples were analyzed with ABL 300 and OSM 3 oximeters (Radiometer, Copenhagen, Denmark).

Determination of the A/ distribution was done with the MIGET technique (17) at baseline and at the end of the stabilization period after induction of lung injury, as well as during PEEP ₂₀, IRV₃₀, and IRV₁₅. This method is based on the constant infusion of six inert gases that have different solubilities in blood, and on their steady-state elimination from the lung. The inert gases are dissolved in isotonic saline and infused into a peripheral vein. Arterial and mixed venous blood samples are tonometered with gas and analyzed together with an expired gas sample through gas chromatography (Model 5890, Series II; Hewlett-Packard, Waltham, MA). These data enable the construction of a virtually continuous distribution of A/ ratios against blood flow or ventilation, with separation of shunt (A/ < 0.005) from regions with low A/ ratios (0.005 < A/ < 0.1; poorly ventilated lung units in relation to their perfusion), as well as separation of dead space (A/ > 100) from regions with high A/ ratios (10 < A/ < 100). The standard deviations of the logarithmic distributions of perfusion (LogSDQ) and ventilation (LogSDV) were calculated as measures of the dispersion (mismatch) of blood flow and ventilation distribution, respectively.

CT Scanning

During ventilation, a frontal topogram of the chest and four transverse scans (140 kV, 111 mA) 8 mm apart and directly cephalad of the diaphragm were obtained with a Somatom Plus 4 CT scanner (Siemens, Erlangen, Germany). The scanning time for the transverse images was adjusted to match one ventilatory cycle (3 s for PEEP₂₀, 2 s for IRV₃₀, and 4 s for IRV₁₅). Since CT scanning was performed with only one ventilatory pattern in each animal, five independent observations were obtained for each ventilatory pattern.

The CT images were analyzed with the Sienet-Magic View, version VA30A computer program (Siemens). All transverse CT scans were analyzed through two different approaches as follows:

1. 1The entire left and right lungs in a single scan were chosen as the region of interest (ROI) by drawing the external boundaries of the lungs along the inside of the ribs and the internal boundaries along the mediastinal organs. The total lung area and its mean density were determined by including pixels with density values between 1,000 and +100 HU. Selecting a wider range, of 1,000 to +1,000 HU, resulted in an increase of < 1% in the total area. Thereafter, for the same ROI, the area of pixels with HU values in the range of 100 to +100 HU, representing atelectasis or lung parenchyma with a maximum of 10% gas (16, 18), and poorly aerated lung tissue with a density between 100 and 500 HU (16), were measured.
2. For analysis of differences in regional aeration, four circular ROIs, each averaging 1.6 ± 0.5 cm² in area, were drawn in the uppermost, ventral, dorsal, and lower dorsal parts of the right lung (Figure 4, inset), respectively. The positions of the ROIs were kept constant for each animal, in order to ensure evaluation of approximately the same area in each scan. The density of each ROI was determined by including pixels within the range of 1,000 and +100 HU. An average density of the four transverse scans of each animal was then calculated for each ROI.

View larger version (25K): [in this window] [in a new window]   Figure 4.   Regional lung density. Four circular ROIs were drawn in the midline of the right lung, as shown in the inset. Their vertical distance from the ventral lung border is given on the x-axis, and the mean density is shown on the y-axis. Data for five animals for each ventilatory mode are shown as mean values, with error bars indicating SEM. CT was performed so that the scanning time matched one ventilatory cycle (PEEP20 = 3 s, IRV30 = 2 s, IRV15 = 4 s). Ventilation with PEEP20 caused less gravitational-directed density difference than did IRV30 or IRV15, but this was not statistically significant (p = 0.18 for differences between PEEP20 and IRV).

Perfusion Scintigraphy

The animals were put in the supine position with their front legs stretched cranially, and three measurements were made with single-photon computed tomography (SPECT) after intravenous injection of ^99mTc-labeled macroaggregated albumin (^99mTc-MAA) (Pulmocis; CISbiointemational, Gif sur Yvette, France).

The first SPECT scan was done 3 h after the induction of lung injury, on pigs that were ventilated in the control mode. Thereafter, two additional measurements were made in random order after 45 min of ventilation with PEEP₂₀ and IRV₃₀, respectively.

In order to obtain high pulmonary emission activity in relation to the contribution of activity from preceding measurements, the injected activity was increased from 50 MBq ^99mTc-MAA for the first SPECT scan to 200 MBq and 1,000 MBq for the second and third SPECT scans, respectively.

Images were acquired on a dual-head gamma camera (Maxxus; General Electric Systems, Milwaukee, WI) equipped with all-purpose, low-energy collimators. The SPECT acquisitions were made in sixty-four projections (32 per head) and stored in a 64 × 64 matrix.

The acquisition time was 15 s per projection during the first SPECT scan and 5 s per projection during the second and third scans.

The data were reconstructed on a Nuclear Diagnostics (Stockholm, Sweden) HERMES workstation. The acquired data were prefiltered with a two-dimensional Butterworth filter (cutoff frequency = 0.14 and filter order = 10). Filtered back-projection reconstruction was performed without applying a correction for attenuation. Through the addition of coronal slices, the lungs were divided into 10 equally thick portions in the ventral-to-dorsal direction, for assessment of the vertical perfusion distribution. Similarly, the lungs were divided into 20 equally thick transverse slices from the apex to the base for analysis of the perfusion distribution in that plane.

After corrections were made for different scanning times and background counts, the number of counts was measured in each volume element of the lungs.

Statistics

All data are presented as mean ± SD, unless stated otherwise. A value of p < 0.05 was chosen as the level of significance. Not all data were normally distributed as tested with the Shapiro-Wilk's W test. Therefore Friedman's analysis of variance (ANOVA) was used to analyze differences between the ventilatory modes, and was followed by Wilcoxon's signed ranks test if significant differences were detected. Data recorded during periods of ventilation in the control mode were analyzed with a parametric ANOVA for repeated measures in addition to nonparametric testing. CT scans and hemodynamic parameters among the groups of five animals each on which CT scanning was performed were statistically analyzed with a Kruskal-Wallis ANOVA. Perfusion scintigraphy was analyzed for an interactive effect of the ventilatory mode with perfusion of the different lung levels, using a two-way ANOVA with a repeated measures design for study groups and lung levels. Calculations were made with the Statistica software package (Statsoft, Inc., Tulsa, OK) on a personal computer.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lung Injury

Induction of lung injury resulted in significant effects on hemodynamics, oxygenation, and respiratory mechanics (Table 1), and 16 of 20 animals fulfilled the strict criteria for ARDS (19). No significant differences were seen (Friedman's ANOVA and parametric ANOVA for repeated measures) among the three periods of ventilation with the control mode (ZEEP), which preceded PEEP₂₀ and the two IRV modes. Therefore, data obtained during ventilation in the control mode were pooled in Table 2, although testing was done for significance of differences in data collected during the test mode and the preceding control mode.

Respiration

Normoventilation was achieved with all three ventilatory patterns. Paw_peak was significantly lower with the two IRV modes than with PEEP₂₀, but all three ventilatory patterns resulted in similar values for Paw_mean. PEEP_i was minor during ventilation with the control mode and PEEP₂₀, whereas IRV₃₀ and IRV₁₅ produced PEEP_i values of 15.3 ± 6.2 cm H₂O and 10.3 ± 6.1 cm H₂O, respectively. Data are summarized in Table 2.

Oxygenation

All three ventilatory modes improved Pa_O2 significantly as compared with the control mode (Table 2), although there was large variation within each group. However, only a PEEP of 20 cm H₂O restored nearly normal oxygenation (p < 0.05 for PEEP₂₀ versus the two IRV patterns) in most of the animals. With PEEP₂₀ (n = 20), Pa_O2 improved in 95% of the animals by  50 mm Hg, and in 70% of the animals by > 250 mm Hg. Contrastingly, during IRV₁₅ (n = 15) and IRV₃₀ (n = 20), Pa_O2 improved by more than 50 mm Hg in only 67% and 60% of the animals, respectively, and by  250 mm Hg in only 20% and 35%, respectively. However, all three ventilatory modes caused a significant decrease in CO, so that oxygen delivery had a tendency to move toward lower values during ventilation with PEEP₂₀ or IRV as compared with the control mode.

Multiple Inert Gas Data

The quality of the MIGET data (Table 3) was technically good, as assessed with the residual sum of squares (RSS) for the measured versus the calculated A/ distributions. RSS averaged 1.6 ± 1.5 (95% confidence interval: 1.2 to 1.9) and exceeded 6 (7.4) in only one measurement. This should be compared with the suggestion that 50% of the A/ measurements should have a fit with an RSS < 6 (20).

Oleic acid injection caused a significant increase in the shunt and perfusion of poorly ventilated lung areas, and increased ventilation in areas with high A/ ratios (Figure 1). Thus, A/ mismatch increased, as indicated by an increased LogSD.

View larger version (22K): [in this window] [in a new window]   Figure 1.   Baseline and lung injury (A/) distribution as determined with MIGET. The continuous distributions of ventilation and blood flow are plotted against the A/ ratio. Data are shown as mean + SEM. (A) Data obtained during baseline measurements. (B) Data obtained at the end of the stabilization period following the induction of lung injury. Note that lung injury causes a decrease of perfusion of lung areas with a A/ ratio close to 1, and increases perfusion of poorly ventilated lung regions. Shunt increased from 8.0% during baseline to 38.2% after induction of lung injury. LogSDQ = logarithmic standard deviation of perfusion distribution; VD = dead space ventilation.

All three ventilatory patterns decreased shunt and A/ mismatch as compared with the control mode (Figure 2). However, only ventilation with PEEP ₂₀ restored nearly baseline values for shunt and LogSD. The differences in shunt between PEEP₂₀ and IRV₃₀, as well as between PEEP₂₀ and IRV₁₅, were significant (p = 0.009 and p = 0.019, respectively). PEEP₂₀ and IRV₃₀ had no effect on dead space ventilation (VD) as compared with the control ventilatory mode, but IRV₁₅ decreased VD by almost 30% (p = 0.001 as compared with the control mode).

View larger version (23K): [in this window] [in a new window]   Figure 2.   A/ distribution during PEEP 20 and IRV as determined with MIGET. The continuous distributions of ventilation and blood flow are plotted against the A/ ratio. Data are shown as mean ± SEM. (A) Data for pressure controlled ventilation with PEEP20 and an expiratory time interval of 2 s. For IRV30 (B) and IRV15 (C ), the same I:E ratio of 3:1 was used and no PEEPe was applied. This resulted in an TE interval of 0.5 s for IRV30, which was expected to nearly prevent end-expiratory lung collapse, whereas the 1-s time interval with IRV15 was expected to allow some end-expiratory lung collapse. However, only PEEP20 reduced shunt and perfusion of poorly ventilated lung areas to nearly baseline values, whereas IRV30 and IRV15 led to very similar perfusion distributions. Log SD2 = logarithmic standard deviation of perfusion distribution; VD = dead space ventilation.

Shunt and Pa_O2 showed an excellent linear correlation after induction of lung injury regardless of the ventilatory pattern used (r² = 0.88, n = 15; Figure 3).

View larger version (20K): [in this window] [in a new window]   Figure 3.   PaO2 versus shunt. The x-axis shows the PaO2 (mm Hg) and the y-axis shows shunt measured with the MIGET technique. PaO2 and shunt showed an excellnt linear correlation over the entire range of observed data. Thus, true shunt was the major mechanism for the observed decrease in PaO2 after induction of lung injury.

CT

The mean lung density was similar for the three groups of five animals each that were ventilated with PEEP₂₀, IRV₁₅, or IRV₃₀ (Table 4).

The gravity-dependent density gradient tended to be smaller during ventilation with PEEP than with the two IRV modes (p = 0.18; Figure 4). It should be noted, however, that the statistical power of the analysis was limited by the small number of independent observations for each ventilatory mode (n = 5).

The atelectatic area located between the diaphragm and the base of the heart, according to standard definition in CT studies (16, 18), was smaller than normally seen either in ALI in clinical studies or in experimental lung damage. However, the exposure was made over an entire breath (see DISCUSSION), which should have resulted in a smaller area of atelectasis because of recruitment of tissue during the inspiration. The mean area of atelectasis was similar during PEEP₂₀ and IRV₃₀, and was significantly larger during IRV₁₅ than with the other two ventilatory modes. The amount of poorly aerated tissue, which should include lung that collapses during expiration and is recruited during inspiration, corresponded to approximately one third of the total lung, with no significant difference between the ventilatory modes (Table 4).

Lung density and the vertical density gradient (HU/cm gravitational direction) explained as much as 78% of the variation of shunt, according to the equation: shunt = 0.076 · density + 0.192 · density gradient 55.1 (r² = 0.78, p = 0.00054). Changes in density alone explained 68% of the variation of shunt (shunt = 0.08 · density +65 HU; r² = 0.68; p < 0.001; Figure 5A), and a similar correlation was obtained for shunt and the combination of atelectasis and poorly aerated lung tissue (500 to +100 HU; r² = 0.65). In contrast, no significant correlation was obtained for the amount of atelectasis and shunt (Figure 5B).

View larger version (15K): [in this window] [in a new window]   Figure 5.   Shunt versus lung density and shunt versus atelectasis. (A) Shunt versus mean lung density. (B) Shunt versus atelectasis. Shunt increased almost linearly with increasing lung density (A), but no such correlation was found for shunt and atelectasis.

Perfusion Scintigraphy

The spatial perfusion distribution (% of CO) during ventilation with the control mode, PEEP₂₀ and IRV₃₀ is shown in Figure 6. As compared with the control mode, PEEP₂₀ and IRV₃₀ caused a small, but nevertheless significant (p < 0.001) redistribuition of pulmonary blood flow from cranially located lung regions toward areas located close to the diaphragm (Figure 6, left panel ); analogously, PEEP₂₀ and IRV₃₀ caused significant (p = 0.001) and, on average, similar redistributions of blood flow from ventral to dorsal lung areas (Figure 6, right panel ).

View larger version (20K): [in this window] [in a new window]   Figure 6.   Regional lung perfusion. Pulmonary perfusion (% of CO) measured with SPECT is shown for separate lung regions. Data are given as mean ± SEM. (Left panel ) Lungs were divided into 20 equally thick slices (x-axis) from the apex of the lungs (cranial) to the diaphragm (caudal). Pulmonary perfusion (% of CO) is shown on the y-axis. Redistribution of blood flow resulting from the use of PEEP20 (solid squares) or IRV30 (solid diamonds) was small, but significant (p < 0.001) as compared with the control mode (solid triangles). (Right panel ) Lungs were divided into 10 equally thick slices in the ventral-dorsal direction (y-axis). As compared with control ventilation, PEEP20 and IRV30 caused a redistribution of pulmonary blood flow from ventral to dorsal lung regions (p = 0.001), but this effect was similar for IRV30 and PEEP20.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A/ matching and oxygenation in porcine oleic acid-induced lung injury improved more during pressure-controlled ventilation with the application of PEEP ₂₀ (I:E = 1:2) than with IRV (I:E = 3:1) with TE intervals of 1.0 s or 0.5 s, respectively. No significant differences among the three ventilatory patterns were noted for mean arterial blood pressure, CO, D.O ₂, or mean airway pressure. Thus, the ventilatory strategy of stabilizing the lung through a short TE interval did not offer any short-term advantages over the more conventional approach of stabilizing the lung with PEEP_e. However, there was a considerable overlap of PEEP_e and IRV in their effects on A/ and gas exchange. What was more consistent was that when aeration of the lung improved the most, irrespective of the ventilatory mode with which this occurred (although more often with PEEP _e than with IRV), then shunt was reduced the most and Pa_O2 was the highest. The perfusion distributions were similar with PEEP₂₀ and IRV₃₀, and they did not explain the better average, oxygenation with PEEP₂₀ than with IRV.

Relevance of PEEP₂₀ and IRV₃₀ to Clinical Practice

IRV₃₀ and PEEP₂₀ were matched in order to equally effectively prevent end-expiratory lung collapse and to yield the same Pa_CO2 of 40 mm Hg. Therefore, the ventilatory settings studied may seem somewhat unusual from a clinical viewpoint. However, PEEP levels above 20 cm H₂O and inspiratory pressure levels above 40 cm H₂O, as applied during PEEP₂₀, are not uncommon in clinical investigations (21, 22), and tidal volumes of 12 to 15 ml/kg are still considered standard in treating ARDS patients (23). During IRV₃₀, no PEEP_e was applied, although the use of PEEP is recommended for ARDS patients (19). Since PEEP_i averaged 15 cm H₂O during IRV₃₀ (Table 2), the application of moderately low levels of PEEP_e (5 to 10 cm H₂O), as recently used in a large clinical trial (23), would most likely have been without major effects on gas exchange. Thus, ventilatory settings similar to those for PEEP₂₀ and IRV₃₀ in the present study may be used for treating patients with acute respiratory failure.

Methodologic Aspects

Lung injury has been shown to be stable between 1 and 4 h after oleic acid infusion in pigs (24, 25). In the present study, in contrast, lung function tended to deteriorate during the course of the experiments. This is obvious when the results of Table 1, with lung injury data recorded 2 h after induction of lung injury, and Table 2, with mean values of ventilation with the sequentially applied control mode, are compared. In addition, all six pigs that died during the experiments did so when ventilation was switched from PEEP₂₀ or IRV to the control mode. Thus, the sudden withdrawal of PEEP_e or PEEP_i most likely causes additional lung dysfunction. This might be important from a clinical perspective, since it may suggest that in patients with severe respiratory failure, disconnection from mechanical ventilation for therapeutic or diagnostic interventions should be avoided whenever possible. For our study, however, a consistent volume history seemed necessary, in order to avoid carryover effects when ventilation was changed. Since we randomized the order of application of the three ventilatory modes, and found no significant differences for the ventilatory periods (control mode) before PEEP₂₀ or the two IRV patterns, a systematic influence in our results caused by changing degrees of lung injury is very unlikely.

CT has been used in numerous studies of patients with acute respiratory failure (26), and CT scanning in these studies was usually performed during breathholding procedures. However, lung collapse and recruitment occur within seconds (15), and CT scans obtained during breathholding procedures are therefore not ideally suited for the comparison of different ventilatory modes. Synchronization of CT scanning with the expiratory or inspiratory phase of mechanical ventilation was not possible in our study for technical reasons. Therefore, we performed CT scanning during mechanical ventilation and adjusted the scanning time to match one ventilatory cycle. This may explain why we did not find a significant correlation between shunt and atelectasis (defined here as a range of density from +100 to 100 HU), since lung tissue with a density above 100 HU, averaged over one ventilatory cycle, is most likely collapsed throughout the expiratiory and inspiratiory phases of the cycle. Lung tissue that is in contrast collapsed during only a part of the ventilatory cycle causes intermittent shunt, but may have an average density below 100 HU. In addition, pixels may contain both aerated and collapsed alveoli, resulting in attenuation values below 100 HU and thus causing an underestimation of atelectasis (partial volume effect). These methodologic limitations are circumvented by using the mean lung density, which we did, and with which we observed a significant correlation between lung density and the vertical gradient on one hand and shunt on the other hand. The circular ROIs for evaluation of regional aeration were positioned at least 0.5 cm away from the lung borders in order to avoid partial volume effects caused by tissue adjacent to lung parenchyma.

Oxygenation with PEEP and IRV

The finding that PEEP₂₀ but not IRV₃₀ restored almost normal oxygenation can mainly be attributed to a larger reduction of shunt with PEEP. The perfusion of poorly ventilated areas was minor with both PEEP₂₀ and IRV₃₀ (Table 3, Figure 2), but this deficit should in any case have contributed very little to the oxygenation impairment in view of the high FI_O2 used in the study. These results are in accord with the finding in a recent study of ARDS patients of a larger reduction of shunt with PEEP_e as compared with the same level of PEEP_i (11). PEEP₂₀ and IRV₃₀ should prevent expiratory lung collapse to a similar degree (15), and in our study were indeed associated with the same amount of atelectasis (within the limits set by the technique we used, as described) and with a similar mean lung density during a ventilatory cycle (Table 4). Thus, other potential mechanisms may explain the better average oxygenation with PEEP₂₀, as subsequently discussed.

Effect of PEEP_i. PEEP_i caused by IRV has been suggested as a "selective" form of PEEP that preferentially expands parts of the lung with prolonged time constants (4). In contrast to PEEP_e, PEEP_i has therefore been proposed to produce more homogenous aeration and less overextension of healthy lung parenchyma (4). Our data do not support this view, but show a better matching of the A/ distribution when PEEP _e was applied. The mean of the ventilation distribution was higher with IRV₃₀ (p = 0.21) and IRV₁₅ (p = 0.02) than with PEEP₂₀ (Table 3, Figure 2), although Paw_mean was similar for the three ventilatory modes. Zavala and coworkers (11) reported a similar change in the A/ distribution in ARDS patients when comparing volume-controlled mechanical ventilation with PEEP _e and pressure-controlled IRV with PEEP_i. In addition, the gravitational density gradient tended to be smaller for PEEP₂₀ than for IRV, indicating a more homogenous aeration with PEEP₂₀ than with IRV (Figure 4). Thus, morphologic data derived from CT scanning may offer an explanation for the results obtained by MIGET and blood gas analysis. Since different time constants in individual lung regions are crucial for a "selective effect" of PEEP_i, it should also be considered that PEEP_i may allow the "selective collapse" of regions with short time constants, which might result from the combination of a low compliance and a normal resistance. This would allow edematous regions to collapse when more normal lung regions remain open, an effect that may not be the desired one.

Effect of inspiratory Paw. Inspiratory Paw was highest (48 cm H₂O) with PEEP₂₀. Since a Paw of up to 40 cm H₂O does not completely recruit collapsed lung tissue in our animal model (15), more lung tissue was most likely opened up during inspiration with PEEP₂₀ than with IRV. Thus, Paw_peak should also be considered an important factor for oxygenation, as previously suggested by Jousela and coworkers (29).

Effect of perfusion distribution. During inspiration, transmission of Paw to the pulmonary vascular bed may force blood flow away from well aerated and toward more collapsed lung regions. Such a redistribution of pulmonary blood flow would increase with an increasing relative duration of the inspiratory phase of the breath cycle and could therefore explain the higher shunt fraction seen during IRV than during PEEP₂₀, as well as the similar shunt fractions obtained with IRV₃₀ and IRV₁₅. We tested this hypothesis with perfusion scintigraphy in five animals with oleic acid-induced lung injury. In comparison to control ventilation, PEEP₂₀ and IRV₃₀ redistributed perfusion away from anterior to posterior regions, resembling what can be expected from an increase in intrathoracic pressure (14). However, the redistribution of blood flow to lower, less aerated or collapsed lung regions was on average not larger with IRV₃₀ than with PEEP₂₀ (Figure 6). Differences in perfusion distribution as an explanation for the differences in oxygenation with PEEP₂₀ had therefore to be rejected in the present study.

Conclusion

A/ matching and oxygenation were improved more by the application of PEEP ₂₀ (I:E = 1:2) than by IRV (I:E = 3:1) with TE intervals of 1.0 s or 0.5 s, in accord with some previous findings in ARDS patients (9, 11). Moreover, our study showed that this result was due to a better and more even aeration with PEEP than with IRV. Distribution of pulmonary blood flow, on the other hand, was on average similar with PEEP and IRV, and could not explain the differences in gas exchange seen with PEEP versus IRV.