Changes in Pulmonary Blood Flow during Gaseous and Partial Liquid Ventilation in Experimental Acute Lung Injury

The experimental protocol was approved by the appropriate governmental institution (Bezirksregierung Koeln, Koeln, Germany, 23.203.2 AC 38, 21/97), and the study was performed according to the Helsinki convention for the use and care of animals.

After premedication with 5 mg/kg azaperone, 10 mg/kg ketamine, and 0.01 mg/kg atropine intramuscularly, anesthesia was induced in 25 healthy female pigs with a mean body weight of 29.4 ± 2.6 kg (PLV group, 29.9 ± 3.6 kg; GV group, 29.1 ± 2.3 kg; attenuation control group, 29.5 ± 1.6 kg) with a bolus of 2–3 mg/kg thiopental intravenously and maintained with the continuous infusion of 6–10 mg · kg⁻¹· h⁻¹thiopental and 0.1 mg · kg⁻¹· min⁻¹fentanyl. Muscle relaxation was achieved with 3 mg · kg⁻¹· min⁻¹pancuronium bromide. The animals were positioned supine and intubated with a 8.0- to 9.0-mm internal diameter endotracheal tube. Volume-controlled ventilation was instituted in all animals using a respirator (Servo 300; Siemens Elema, Lund, Sweden) with a fraction of inspired oxygen (Fio²) of 1.0, a respiratory rate of 20 breaths/min, a mean tidal volume of 8 ml/kg, a positive end-expiratory pressure of 5 cmH²O, and an inspiration–expiration time ratio of 1:2 without pause time. The ventilator setting remained unchanged throughout the entire study protocol.

Arterial access was achieved by introducing a 18-gauge catheter (Vygon, Ecouen, France) into a femoral artery. A pulmonary artery catheter (model 93A-431-7.5 F; Baxter Healthcare Corporation, Irvine, CA) was advanced into a pulmonary artery under transduced pressure guidance through a 8.5-French venous sheath (Arrow Deutschland GmbH, Erding, Germany) positioned in a femoral vein.

The blood temperature, determined by means of the pulmonary artery catheter, was maintained at 37.5 ± 1.0°C in all animals during the investigation by use of an infrared warming lamp and a warming pad. A balanced electrolyte solution was administered continuously at 4–5 ml · kg⁻¹· h⁻¹to maintain adequate hydration.

All hemodynamic measurements were taken with the animal in the supine position. Pressure transducers were calibrated to the center of the lateral chest, and central venous pressure, mean arterial pressure, mean pulmonary arterial pressure, and pulmonary artery occlusion pressure of all animals were transduced (PVP, Kirschseeon/Eglharting, Germany) and recorded (Model CS/3 Compact; Datex, Achim, Germany). Cardiac output was determined using standard thermodilution techniques (Baxter Deutschland GmbH, Unterschleiβheim, Germany) and expressed as the mean of three measurements at end-expiration of different respiratory cycles. Heart rate was obtained from the blood pressure curve.

Arterial and mixed venous blood samples were collected anaerobically and immediately analyzed for partial pressure of oxygen (Po²), partial pressure of carbon dioxide (Pco²), and pH using standard blood gas electrodes (ABL 520; Radiometer Copenhagen, Copenhagen, Denmark). Species-specific spectrophotometry was performed to obtain arterial and mixed venous oxygen saturation and total hemoglobin concentration (OSM 3 Hemoximeter; Radiometer Copenhagen, Copenhagen, Denmark).

The oxygen content (ml/dl) of arterial oxygen concentration (Cao²) and mixed venous oxygen concentration (Cvo²) samples was calculated using the following formula: Content of oxygen = (hemoglobin concentration × 1.34 ×%O²− saturation/100) + (Po²× 0.0031). To calculate the pulmonary capillary oxygen content (Cco²), the pulmonary capillary oxygen tension was assumed to be equivalent to the alveolar Po²which was estimated as follows: Barometric pressure − Paco²/respiratory quotient (assuming that the respiratory quotient = 0.8) 10− water vapor pressure (47 mmHg) − perfluorocarbon vapor pressure (76 mmHg at 37.5°C for the perfluorocarbon used in this study, only when PLV was performed). 11The venous admixture (Q^s/Q^t) was derived from the standard shunt equation: (Cco²− Cao²)/(Cco²− Cvo²). Some reservations have to be made regarding the calculation of shunt during PLV with this formula. As previously outlined by Mates et al. , 11calculated shunt during PLV may not only be attributed to the perfusion of unventilated lung areas but also to some diffusion limitation for oxygen through perfluorocarbon in the ventilated but liquid-filled alveoli. The quantitative effect of this diffusion limitation as well as the correlation between the increase of intrapulmonary perfluorocarbon volume and the increase of calculated shunt cannot be estimated as yet. The impact of ventilation–perfusion heterogeneity on arterial oxygenation seems minimal as long as a Fio²of 1.0 is used.

Macroaggregates of human serum albumin were labeled with ^99mTc according to the protocol of the manufacturer (^99mTc-Lyo-MAA, Mallinckrodt Medical B.V., Petten, The Netherlands) who guarantees a size between 10 and 100 μm for 95% of the particles. After the intravenous administration of the radioactive-labeled macroaggregated human serum albumin particles in the supine animals, regional pulmonary blood flow distribution expressed as percentage of total activity measured in both lungs was determined using single-photon emission computed tomography (SPECT) with a triple-head gamma camera (Multispect 3; Siemens, Erlangen, Germany) fitted with low-energy high-resolution collimators. SPECT previously has been validated for this purpose by other investigators. 10,12–14Acquisition parameters were 360-degree rotation with 120 steps of 3 degrees each in a 128 × 128 matrix, with acquisition times depending on the injected activity at each study point and an acquisition zoom of 1.23. Reconstruction was done by filtered back-projection using a Butterworth Filter of order 3.0 and a cutoff frequency of 0.5 cycles/pixel. This resulted in transaxial slices of 3.51 mm. Spatial resolution (full-width, half-maximum) measured with a 0.6-mm line source was 11.7 mm.

Using the ICON 7.1 software (Siemens, Erlangen, Germany), the lungs were divided into four transaxial slices numbered from one for the most apical to four for the most basal (fig. 1). Changes in blood flow distribution between dependent and nondependent lung regions were analyzed by an automatic algorithm; in each transaxial slice, a rectangular region of interest, subsequently divided into three geometrically equal segments along the vertical axis (ventral, middle, and dorsal), was drawn around the lung shape using four points representing 13% of the maximum activity. This isocontour-based calculation guarantees an exact reproducibility of region of interest definition. 15,16In this manner, a total of 12 lung segments was generated for each animal (fig. 1). For each study point, pulmonary blood flow to a particular lung segment was calculated as percentage of the total activity in both lungs using the following formula: Segmental pulmonary blood flow [%]= counts per segment/total counts in both lungs.

Lung injury was induced in all animals in supine position by surfactant depletion caused by repeated lung lavage as previously described and evaluated by Lachmann et al.  17,18Briefly, after disconnecting the animals from the ventilatory circuit, the lungs were filled with 30 ml/kg of prewarmed saline (38°C, 0.15 m) through the endotracheal tube. Removal of the lavage fluid was achieved by tilting the animals head down at an angle of about 45 degrees using the gravitational gradient. ALI was defined as a Pao²consistently lower than 100 mmHg for 1 h at a Fio²of 1.0 and a positive end-expiratory pressure of 5 cmH²O after the last lung lavage.

Partial liquid ventilation was initiated in the supine animal with the administration of two sequential doses of 15 ml/kg of perfluorocarbon (C⁸F¹⁸, “FC 3280”; 3M Chemical Products, Neuss, Germany) through the endotracheal tube during inspiration, using a swivel connector (Portex, Kent, United Kingdom). Each dose was administered into the airway during a 10-min time interval resulting in a volume of approximately 2–2.5 ml of perfluorocarbon per respiratory cycle. The total volume of 30 ml/kg perfluorocarbon is estimated to be equivalent to the normal functional residual capacity of pigs. The perfluorocarbon used in our investigation is a highly purified industrial perfluorocarbon with a density of 1.75 g/cm³, a surface tension of 12 mN/m, and a vapor pressure of 61 mmHg at 25°C (76 mmHg at 37.5°C). Up to 40 ml O²and 192 ml CO²can be dissolved in 100 ml of this liquid (data from 3M Chemical Products, Neuss, Germany).

Gas exchange, hemodynamic, and pulmonary blood flow data were assessed in supine position in all animals before (T^Baseline) and after the induction of acute lung injury (T^ALI). Subsequently, 20 animals were randomly assigned to one of two groups.

One group was submitted to PLV (n = 10, PLV group) by administering 15 ml/kg of perfluorocarbon through the endotracheal tube. As evaluated in previous studies, losses of perfluorocarbon as a result of evaporation were continuously replaced during the rest of the study at a volume of 4 ml · kg⁻¹· h⁻¹of perfluorocarbon. 19,20Measurements of hemodynamic, gas exchange, and pulmonary blood flow distribution data were performed 1 h after instillation of this first dose of perfluorocarbon (T^1h). Subsequently, a second dose of 15 ml/kg of perfluorocarbon was applied, and again, data of all parameters tested were determined 1 h after achieving the final intrapulmonary perfluorocarbon volume of 30 ml/kg (T^2h).

In the second group of animals (n = 10, GV group), controlled mechanical ventilation was continued with no changes in the respirator setting, as outlined previously. Measurements of hemodynamic, gas exchange, and pulmonary blood flow distribution data was performed 1 h (T^1h) and 2 h (T^2h) after the assessment of ALI, corresponding to the study points of the PLV group.

To allow relative quantification of pulmonary blood flow by SPECT imaging at the four serial study points, appropriate fourfold-increasing amounts of radioactivity were injected at these points of time: 20 MBq ^99mTc-Lyo-MAA at T^Baseline, 80 MBq at T^ALI, 320 MBq at T^1h, and 1280 MBq at T^2h. Adjusted to these increasing amounts of radioactivity, the acquisition times dropped from 80 s per step at T^Baselineto 40 s per step at T^ALI, 10 s per step at T^1h, and 5 s per step at T^2h. A similar procedure to perform repeated measurements has already been used by other groups and is in accordance with the clinical routine. 10,13The remaining acquisition parameters were kept constant. The overall amount of ^99mTc-Lyo-MAA injected was about 1.5 mg. To avoid a possible error as a result of residual activity from previous injections of the radiotracer, measurements at T^ALI, T^1h, and T^2hwere corrected by subtracting the counts obtained during the previous SPECT scan corrected for different acquisition time and decay. 10,13 

Attenuation correction for the effects of the chest wall, heart, and lung tissue was not performed in the animals submitted to PLV and GV because of the lack of a reliable transmission source. However, although the importance of attenuation in intact lungs is not universally accepted, 21–23it seems possible that measurements of regional blood flow distribution in the group randomized for PLV were influenced by attenuation of the photon energy as a result of the dense perfluorocarbon, especially in the dorsal lung segments. Therefore, a third group of animals (n = 5, attenuation control group) was investigated. After induction of anesthesia and ALI as described previously, a single dose of 400 MBq ^99mTc-Lyo-MAA was injected, and a first SPECT scan with an acquisition time of 10 s per step was performed (SPECT 1). Subsequently, perfluorocarbon was instilled in two sequential doses of 15 ml/kg each, and a SPECT scan with an acquisition time of 10 s per step was performed after each instillation (SPECT 2 and 3), with no further injection of ^99mTc-Lyo-MAA.

All data are expressed as mean ± standard deviation. Data analyses were conducted using the NCSS 2000 software package (NCSS, Kaysville, UT). All data regarding the PLV group and the GV group were analyzed by two-way repeated-measures analysis of variance (ANOVA) for comparisons between and within groups. The statistical design for the ANOVA procedure was based on two within-group factors: group having two levels (GV group, PLV group) and study point having four levels (T^Baseline, T^ALI, T^1h, T^2h). For significant ANOVA results, post hoc  correction for multiple comparisons was performed using Bonferroni test for all-pairwise comparison. Values of P < 0.05 were considered significant. Data analyses of the third group of animals designated to study the attenuation of photon energy by perfluorocarbon were performed using two-way repeated-measures ANOVA followed by Bonferroni test for all-pairwise comparison when ANOVA revealed significant results.

All pigs (n = 25) survived the entire study period. A total of 11 ± 3 lung lavages (PLV group 11 ± 3 lung lavages, GV group 12 ± 3 lung lavages) were necessary to achieve a persistent decrease in Pao²from 557 ± 38 mmHg to 58 ± 12 mmHg with an increase of Q^S/Q^Tfrom 13 ± 1% to 54 ± 11% in the animals randomized to GV or PLV. Concomitantly, Paco²increased from 32 ± 4 mmHg to 57 ± 13 mmHg after the induction of ALI in these animals. Analyses of hemodynamic, gas exchange, and pulmonary blood flow data at T^Baselineand T^ALIrevealed no differences between the PLV and the GV groups for all parameters tested. Total hemoglobin concentration and calculated capillary oxygen content remained unchanged in both groups throughout the study.

Gas exchange and metabolic data for the PLV and GV groups are summarized in table 1. PLV resulted in a significant improvement of Pao²and Q^S/Q^Twhen compared with GV only when the higher of the two doses of perfluorocarbon was applied (P < 0.05). No alterations in arterial oxygenation or venous admixture were observed in the GV group during the entire study period. Paco²and pH remained constant in both groups of animals after the initial changes following the induction of ALI.

Hemodynamic data for the PLV and GV groups are presented in table 2. Statistical analyses revealed no changes in any of the hemodynamic parameters tested within or between the two groups during the investigation.

To investigate regional changes in pulmonary blood flow distribution, the lungs were divided in four transaxial slices which consisted of three segments each (ventral, middle, dorsal). Analyses were performed separately for each of the 12 segments for each animal, as well as for each of the four transaxial slices and the combined ventral, middle, and dorsal segments, respectively (fig. 1). When all animals randomized for either GV or PLV were analyzed together (n = 20), induction of ALI was found to cause an increase in pulmonary blood flow to the ventral segment of slice 2 and the middle segments of slices 1 and 2 (P < 0.05). Additionally, an increase in blood flow was observed in the combined middle segments of all slices, whereas it decreased in the dorsal slices (P < 0.05). In contrast, no effect of ALI on the pulmonary blood flow distribution was observed compared with T^Baselinewhen the GV group (n = 10) and the PLV group (n = 10) were analyzed separately.

Neither PLV nor GV resulted in a redistribution of pulmonary blood flow regarding any of the 12 segments when values at T^1hand T^2hwere compared with data obtained at T^ALI(fig. 2). The trend toward a decrease in the combined dorsal segments and a concomitant increase in the ventral and middle segments did not reach statistical significance in either group (fig. 3). Representative SPECT images of slice 3 at all study points are presented in figure 4. Analyses in each of the four transaxial slices revealed no influence of PLV or time on regional blood flow distribution within the PLV or GV groups (fig. 5). Comparison of corresponding time points between the two groups of animals did not show any statistically significant difference.

In the attenuation control group, 11 ± 2 lung lavages with saline were required to induce ALI with a Pao²of 52 ± 14 mmHg. Analyses of the segmental counts obtained during ventilation without perfluorocarbon and PLV with 15 and 30 ml/kg revealed no differences among the three study points (fig. 6).

The purpose of this study was to investigate changes in pulmonary blood flow distribution during PLV in an experimental model of ALI. We found that PLV did not cause changes in pulmonary blood flow distribution compared with values obtained after induction of ALI in the PLV group or with animals submitted to GV. Arterial oxygenation improved significantly after the instillation of 30 ml/kg of perfluorocarbon in the PLV group but not in the GV group.

Compression of the pulmonary vasculature in the posterior lung and subsequent redistribution of pulmonary blood flow to nondependent, better-ventilated lung areas is one mechanism that has been proposed to account for the beneficial effect of PLV on arterial oxygenation in clinical and experimental ALI. 6To investigate this hypothesis, we employed a surfactant washout animal model of ALI which demonstrates histologic findings similar to those in patients with acute respiratory distress syndrome 17and has been extensively used by other investigators. As in acute respiratory distress syndrome, pulmonary shunt and oxygen-refractory hypoxemia caused by atelectasis and increased radiologic density in the dorsal pulmonary regions are typical features of the lung lavage model. 24,25 

Pulmonary blood flow distribution was determined by SPECT, a technique which is routinely used in clinical practice for this purpose and has been validated. Tow et al.  14investigated whether the distribution of I¹²⁵-labeled macroaggregates of human serum albumin (MAA) in the lung parallels that of Cr⁵¹-labeled red blood cells. They found a high linear correlation between the regional distribution of the particles and the cells (r = 0.97), indicating that the number of radiolabeled MAA particles trapped in a particular lung region is actually proportional to the pulmonary arterial blood flow to that region. Further evidence for the validity of the technique found in their study were the optimal size of the MAA particles (85%≥ 9.2 μm, majority 50 μm) regarding the average internal diameter of the pulmonary capillaries (8 μm), 26the high extraction efficiency of 80% of the administered dose of particles in the first passage through the lung, and the uniform mixing of the particles prior to their arrival in the pulmonary arterial vessels. Using the injection of radiolabeled MAA to examine changes in regional pulmonary blood flow caused by changes in posture in six subjects, they were able to demonstrate the well-known effects of the erect and supine position on the pattern of pulmonary arterial blood flow. Lamm et al.  10compared relative regional pulmonary blood flow (Qr) measured by SPECT-scanning with ^99mTc-MAA and by radiolabeled microspheres in three animals and found no statistical difference for the median Qr (SPECT 0.90 ± 0.1, microspheres 1.00 ± 0.15, P = 0.07) and the coefficient of variation (SPECT 48.7 ± 13.3, microspheres 32.1 ± 1.1, P > 0.1) comparing the two methods. Using the SPECT technique, they measured the frequency distribution of relative regional ventilation–perfusion ratios (V/Q) as well as the regional redistribution of relative V/Q depending on posture. A good correlation between SPECT in an intact canine thorax and scintillation camera counts of frozen slices of the thorax was reported by Osborne et al. , 12indicating that a reasonable accuracy is possible when imaging an extended source of ^99mTc-labeled microspheres in the lung and therefore allowing the regional quantification of a physiologic function such as blood flow. Based on these results, SPECT with ^99mTc-labeled MAA has been recently used to study the ventilation–perfusion distribution and correlation to atelectasis during anesthesia 23and to make detailed quantitative estimates of regional pulmonary blood flow following changes in posture and breathing with or without a continuous positive airway pressure. 13 

Attenuation correction is considered to be an important factor when performing SPECT with respect to gamma photons being attenuated during their way through the body. Because the gamma camera system used in our study was not equipped with an attenuation correction device, we had to rule out the possibility that changes in pulmonary blood flow during PLV were influenced or even simulated by attenuation artifacts caused by perfluorocarbon. We therefore performed SPECT acquisitions in five pigs with ALI using a single injection of 400 MBq ^99mTc-Lyo-MAA and acquisition of three sequential SPECT scans during GV and PLV with 15 or 30 ml/kg of perfluorocarbon. The rationale of this procedure was based on the finding that approximately 80% of the radiolabeled macroaggregates of albumin are extracted and fixated in the pulmonary capillaries during the first passage through the lung. 14Therefore, only attenuation of the radioactive signal by perfluorocarbon, and not the redistribution of the ^99mTc-Lyo-MAA–labeled particles following the onset of PLV, were likely to cause changes in the regional activity measured with SPECT. Using the surfactant washout animal model for this validation seemed appropriate to avoid differences in the experimental setting that could have influenced the results. When comparing the percentage of total activity measured in each pulmonary segment, we could not detect any significant differences indicating increased attenuation following the administration of 15 or 30 ml/kg perfluorocarbon. Therefore, we conclude that the administration of perfluorocarbon does not interfere with the determination of pulmonary blood flow distribution performed with SPECT itself. In addition, measured attenuation correction with the use of a transmission scan has been shown to reduce performance of SPECT by introducing noise into the corrected emission scan. 27Especially in sequentially performed SPECT scans including the comparison of low-dose and high-dose studies as performed in our study, downscatter may result in serious errors of transmitted attenuation correction. 22 

Residual radioactivity in the mediastinal structure also could have influenced the measurements of regional pulmonary activity. However, because of the high first-pass extraction of ^99mTc-Lyo-MAA, the size and number of the circulating aggregates of albumin after the first passage are small. They are metabolized in the liver and spleen and do not introduce an appreciable error in the measurements. 14 

Changes of pulmonary blood flow distribution along the vertical axis of the lung following the intrapulmonary instillation of perfluorocarbons have been shown in healthy animals. 8However, ALI can interfere with the regional distribution of pulmonary blood flow, and the effect of PLV in such settings may be different from results obtained in subjects without respiratory failure. The severe ventilation–perfusion mismatch in the posterior lung segments caused by ALI is likely to cause hypoxic pulmonary vasoconstriction in these regions, which itself can result in a redirection of pulmonary blood flow to more anterior lung areas. This has been shown in animals with oleic acid–induced acute respiratory failure that maintained almost normal ventilation–perfusion ratios and arterial oxygenation as a result of a redistribution of regional pulmonary blood flow from dorsal to middle and ventral zones of the lungs after initiation of lung injury. 28Elimination of hypoxic pulmonary vasoconstriction caused by the application of low doses of endotoxin prior to the induction of lung injury with oleic acid reversed this effect, and the animals demonstrated a severe deterioration of arterial oxygenation that was not observed when oleic acid was administered alone. A decrease in pulmonary blood flow directed to dependent lung regions in animals with oleic acid–induced ALI submitted to conventional mechanical ventilation has also been reported by Enrione et al.  29They explained this observation by vascular closure following lung edema, which is most pronounced in the dependent lung regions when oleic acid is injected in the supine position. 30Whereas the injection of oleic acid causes preferentially a severe endothelial damage with a sometimes excessive increase of mean pulmonary arterial pressure and subsequent deterioration of gas exchange, 31lung lavage with saline results in atelectasis, alveolar epithelial lesions, and inflammatory reactions, thereby impairing arterial oxygenation. 18Although suggested by our data from the GV group, it remains speculative whether these differences between the lung injury models, i.e.,  differences in the homogeneity of the lung injury, exert different effects on pulmonary blood flow distribution. A high experimental variability in the degree of lung injury will increase the difficulties in finding a significant effect of PLV on the pulmonary blood flow distribution when the total changes in regional activity are small.

In contrast to the GV animals in their study and previous investigations by other groups, Enrione et al.  29found a preservation of pulmonary blood flow distribution in animals treated with PLV compared with data prior to and after the induction of ALI. They suggested that the preserved pattern of regional pulmonary blood flow was the result of the inhibition of hypoxic pulmonary vasoconstriction and a combined effect of liquid positive end-expiratory pressure and improved alveolar ventilation. Similar effects are likely to be the reason for the improvement of arterial oxygenation during PLV in our investigation. However, whereas injection of oleic acid caused a clear decrease of pulmonary blood flow to the most dependent lung segments due to its preference for the dorsal regions in the study by Enrione et al. , 29saline lung lavage failed to produce such a clear effect on pulmonary blood flow distribution. Therefore, the changes following the instillation of perfluorocarbon in our investigation may have been less pronounced as well.

There are no studies investigating changes in pulmonary blood flow between apical and basal slices of the lung during PLV in animals with ALI. Doctor et al.  8studied the effect of PLV on regional blood flow in healthy lambs and found a flow pattern favoring dependent over nondependent lung segments and diaphragmatic over apical transverse planes prior to the onset of PLV. We did not observe such a change in blood flow to apical or basal lung segments in animals with ALI submitted to PLV. This finding may have been due to the same reasons as the differences in pulmonary blood flow distribution between dorsal and ventral lung segments during PLV in animals with and without lung injury as discussed earlier. 29 

We found no redistribution of pulmonary blood flow during PLV, although arterial oxygenation improved significantly following the intrapulmonary instillation of perfluorocarbon. Improvement of Pao²in ALI is usually the result of an improvement in ventilation–perfusion ratios (V^A/Q). Possible mechanisms by which PLV may affect V^A/Q include the improvement of alveolar ventilation as a result of the recruitment of previously atelectatic lung regions as well as an increase in perfusion of ventilated regions, probably as a result of the inhibition of hypoxic vasoconstriction in liquid-filled and thereby stabilized alveoli. Each of these mechanisms may be differentially important and expressed depending on the type of lung injury. In this regard, the improvement of V^A/Q, and not only the redistribution of pulmonary blood flow, may be important for the increase of arterial oxygenation during PLV.