Recent studies have questioned the classical gravitational model of pulmonary perfusion. Because the lateral position is commonly used during surgery, the authors studied the redistribution of pulmonary blood flow in the left lateral decubitus position using a high spatial resolution technique.
Distributions of pulmonary blood flow were measured using intravenously injected 15-microm diameter radioactive-labeled microspheres in eight halothane-anesthetized dogs, which were studied in the supine and left lateral decubitus positions in random order. Lungs flushed free of blood were air-dried at total lung capacity and sectioned into 1,498-2,396 (1.7 cm3) pieces per animal. Radioactivity was measured by a gamma counter, and signals were corrected for piece weight and normalized to mean flow.
Blood flow to the dependent left lung did not increase, and blood flow to the nondependent right lung did not decrease in the lateral position. The left lung received 39.3 +/-7.0% and 39.2+/-8.8% (mean +/- SD) of perfusion in the supine and left lateral positions, respectively. Detailed assessment of the spatial distributions of pulmonary blood flow revealed the lack of a gravitational gradient of blood flow in the lateral position. The distributions of blood flow did not differ in the supine and left lateral decubitus positions.
Perfusion to each lung did not change with movement from the supine to the left lateral position. These findings contradict the prediction of increased dependent lung and decreased nondependent lung blood flow based on the gravitational model. It was concluded that the distribution of blood flow in the lateral position in dogs is dominated by pulmonary vascular structure.
THE lateral decubitus position is commonly used during anesthesia, especially for lung surgery. Conventional wisdom based on the gravitational model is that blood flow is decreased to the nondependent lung and is increased to the dependent lung. Older studies using low spatial resolution suggested that blood flow was decreased in the nondependent lung in the lateral position. [3–7]
Recent studies using new high-resolution techniques have questioned the overwhelming importance of the gravity in determining the distribution of perfusion in the lung. [8–15]There is no consistent increase in blood flow to the more dependent lung regions, and there is a considerable degree of perfusion heterogeneity within isogravitational regions in prone animals. [8–12,14]In addition, gravity-independent central-to-peripheral and dorsal-to-ventral flow gradients have been observed. [8–13]These findings suggest that pulmonary vascular structure, rather than gravity, is the major determinant of the distribution of pulmonary blood flow in animals.
The current study applied these new techniques to determine the redistribution of pulmonary blood flow in the lateral decubitus position compared with supine position in dogs. According to the gravitational model, blood flow should increase to the dependent lung and decrease to the nondependent lung when rotating the animals from the supine to the lateral position.
Materials and Methods
Animal Preparation and Experimental Protocol
The animal preparation and experimental protocol were described in detail in the earlier publication. The study was approved by the University of Washington Animal Care Committee. Eight mongrel dogs (mean +/- SD weight, 26.9 +/- 1.5 kg) of both sexes were studied. Anesthesia was induced with thiamylal sodium (18 mg/kg given intravenously) and maintained with 1% inspired halothane. The dogs were orally intubated and mechanically ventilated using a piston ventilator (Harvard, South Natick, MA) with room air (fractional inspired oxygen tension, 0.21; tidal volume, 15 ml/kg; and a rate sufficient to maintain arterial partial pressure of carbon dioxide between 36 and 40 mmHg). Positive end-expiratory pressure and muscle relaxants were not used to ensure apnea during injections of microspheres. Ventilatory parameters were not changed during the study. Catheters were placed in the femoral artery and vein, and a pulmonary artery catheter was inserted via a jugular vein. Animals were studied in the supine, prone, and left lateral decubitus positions in random order. Data from the supine and prone positions were presented in the article by Glenny et al. A similar shape of the chest wall in both positions was sought. To accomplish this, bags filled with styrofoam beads (Vacu-Positioners; Shor-Line, Kansas City, MO) were placed around their chests when they were supine, and a cardboard box was then placed around the bags. When the air from the bags was evacuated, a semirigid mold of the dog's chest was formed. When the dogs were moved to the left lateral position, their chest wall shape remained relatively constant. The mold restricted chest wall movement somewhat but did not hinder abdominal motion and increased peak inspiratory pressures by <5 cm H2O. A lateral chest radiograph with a horizontal radiopaque marker was obtained for each animal.
After 20 min in each position, blood gases, hemodynamics, and pulmonary blood flow distribution were measured. Mean systemic arterial, pulmonary arterial pressures, and peak airway pressures were collected. Partial pressures of oxygen and carbon dioxide and pH were analyzed (ABL 4; Radiometer, Copenhagen, Denmark). Cardiac outputs by thermodilution were obtained in all animals (Edward's COM-2; Baxter, Irvine, CA). Measurements of pulmonary arterial pressure were obtained in five of the eight animals. The difference between alveolar and arterial oxygen tension was calculated assuming a respiratory quotient of 0.8.
Determination of Pulmonary Blood Flow
Radioactive 15-[micro sign]m diameter microspheres (Dupont, NEN Research Products, Boston, MA) were used to measure regional blood flow to the lungs in the different postures. At least 16 x 106microspheres of one of five different labels (141) Ce,113Sn,103Ru,95Nb, or46Sc) were injected during each study phase to ensure that adequate numbers of microspheres would be present in each piece of the lung. The microspheres were sonicated and vortexed, the lungs were inflated to an airway pressure of 25 - 30 cm H2O and then allowed to deflate passively to functional residual capacity (FRC), and each radiolabeled microsphere was then injected over 2 s via the femoral venous catheter during apnea to avoid regions with zone 1 conditions. At the end of the study, the animals were deeply anesthetized, received a bolus dose of heparin (20,000 U given intravenously), and were exsanguinated through the arterial cannulas. The animals were killed by injection of pentobarbital sodium (150 mg/kg given intravenously), a median sternotomy was performed, and large-bore catheters were placed in the pulmonary artery and left atrium. The lungs were perfused with normal saline until clear of blood. Then they were excised, reinflated to total lung capacity (TLC) with an airway pressure of 25 cm H2O, and suspended to dry. The lungs were glued together in anatomic position using cyanoacrylate glue (Duro Superglue; Loctite, Cleveland, OH). After drying for 6 - 8 days, the lungs were suspended vertically in a plastic-lined square box and embedded in rapidly setting urethane foam (2 lb Polyol and Isocyanate; International Sales, Seattle, WA) to provide a rigid form to which a three-dimensional coordinate system could be applied. The lateral chest radiograph and level marker were used to align the lungs in the box so that slicing produced true isogravitational planes.
A miter box was used to cut the foam block into uniformly sized cubes that were [almost equal to] 1.7 cm3in volume. Any foam adhering to the lung pieces was removed. Each lung piece was assigned a lobe designation and a unique three-dimensional (x, y, z) coordinate, where x represents distance in left-to-right planes, y represents distance in dorsal-to-ventral planes, and z represents distance in caudal-to-cranial planes. The number of airways ways present in the pieces was estimated visually (<25% of piece volume, 25 - 50%, 50 - 75%, >75%, trachea or part of trachea) and coded.
The radioactivity in each lung piece was determined in a 3 x 3.25-inch sodium well crystal gamma counter (Minaxi gamma counter system, model 5550; Packard, Downers Grove, IL). The effluent from the lung perfusion of two animals was collected for measurements of radioactivity.
Each sample was corrected for decay time and spill-over by a matrix inversion method. Pure isotope samples were used to select one energy window per isotope such that >90% of the principal peak was included and none of the energy windows overlapped. The spillover matrix was constructed from these samples by use of previously defined energy windows. Each lung piece was counted long enough to ensure a counting error of <1%.
Tissue samples with an airway content of >or= to 25% were not included in the final analysis. The weight-normalized relative flow per piece was calculated by first dividing the radioactive signal of each piece by the piece weight and then by dividing the result by the mean value of all of the final samples from the same animal and condition. The pulmonary hila were defined with spatial coordinates as the points of entry of the left and right pulmonary arteries into the lungs. The hilar-to-peripheral distance in centimeters from the ipsilateral hilum (dh) for a piece with spatial coordinates (x, y, z) were calculated as
dh= 1.2 [middle dot][(xh- x)2+(yh- y)(2)+(zh- z)2]0.5,
where the subscript h indicates coordinates for hilum and 1.2 is the conversion factor to change the units to centimeters.
Heterogeneity of pulmonary blood flow was assessed by the coefficient of variation (SD/mean) of regional pulmonary blood flow. Percentages of blood flow to each lung and lobe were calculated.
The regional pulmonary blood flow as a linear function of left-to-right, dorsal-to-ventral, caudal-to-cranial, and hilar-to-peripheral spatial vectors was characterized for each animal by least-squares regression analysis. Because the slopes were not different in the left and right lungs, the data for the entire lung are presented in results. Although the dimension of the slope is normalized flow units per centimeter, it can be expressed in more familiar terms (percent per centimeter) because the mean normalized flow for the entire animal was 100%. For example, a slope of -6.0%/cm means that flow decreases 0.060 normalized flow units per centimeter. Slopes of linear relationships (e.g., flow vs. x dimension) from all animals were compared with zero with a single-sample two-tailed t test. Pearson's correlation coefficient (r) was used to qualify the strength of the relationship. The square of linear correlation (r2) was used to qualify the proportion of pulmonary blood flow variability that was explained by variability in distance.
Blood flow to each piece was also calculated in ml [middle dot] min-1[middle dot] g-1by multiplying the weight-normalized relative flow per slice by the average cardiac output per piece (e.g., cardiac output/number of pieces in each lung). Pieces with >25% airways were not excluded from this analysis. Blood flow to each lung piece in the supine position was compared with the same piece in the lateral decubitus position using linear regression analysis. The difference in blood flow to each lung piece was calculated and characterized as a function of left-to-right distance by linear regression analysis, and the slope of this relationship from all animals was compared with zero. Blood flow in each left-to-right plane was calculated for each animal by the sum of blood flow to each individual lung piece within that plane. Values for all the animals were compared in the two positions using a two-factor repeated-measures analysis of variance (position, distance in left-to-right dimension), with significant differences compared by the paired t test with Bonferroni correction. The average difference in blood flow per piece (Qsupine- Qleftlateral decubitus) was compared in the dorsal and ventral regions of each left-to-right plane by a two-factor analysis of variance (dorsal or ventral, distance in left-to-right dimension). In addition, the slopes of the average Qsupine- Qleftlateral decubitus versus left-to-right distance were compared in dorsal and ventral areas by a two-tailed paired t test.
Differences in percent blood flow to each lung and lobe and pulmonary blood flow heterogeneity, and flow gradients were compared in the supine and left lateral decubitus positions using two-tailed comparison t tests. The data are presented as the mean +/- SD. A probability value <0.05 was considered significant.
Hemodynamics and Gas Exchange
All hemodynamic parameters were stable during the study (Table 1). No significant differences were observed between mean systemic arterial and pulmonary pressures, peak airway pressures, cardiac output, pH, partial pressures of oxygen and carbon dioxide, and the difference in alveolar-arterial partial pressures of oxygen in the supine and left lateral postures.
The number of pieces analyzed and their weight and number of planes per animal are summarized in Table 2. For the analysis of blood flow gradients (slopes), coefficient of variation flow, and percent flow to each lung and lobe, [almost equal to] 9%(154 +/- 66) of lung pieces per animal were excluded because of the presence of airways, resulting in a range of 1,250–2,194 pieces analyzed per animal. These pieces were included for the analyses in ml [middle dot] min-1[middle dot] g-1, for a total of 1,392 - 2,396 pieces analyzed per animal.
The left lung received 39.3 +/- 7.0% and the right lung received 60.7 +/- 7.0% of the perfusion in the supine position (Table 3, P < 0.01 between lungs). The left lung represented 43 +/- 3% and the right lung represented 57 +/- 3% of the total weight of the lung. The percent blood flow to each lung remained the same in the left lateral decubitus position, although the left lung was dependent and the right lung was nondependent. The percent blood flow to each lobe also did not change in the left lateral decubitus position, with the exception of a significant (P < 0.01) increase in perfusion to the right accessory lobe (Table 3). There was an excellent correlation (0.85 +/- 0.04, P < 0.001) between the blood flow to each lung piece in the supine and left lateral positions (Figure 1).
The distribution of pulmonary blood flow as a function of the left-to-right distance in a representative animal is illustrated in the supine (Figure 2) and left lateral decubitus (Figure 3) positions. The distribution of pulmonary blood flow per plane was highly variable (Figure 2and Figure 3and Table 4). None of the linear gradients accounted for >10% of the variability in flow in either position (r2= 0.02–0.09;Table 5). The coefficient of variation of blood flow was not different in the supine and lateral decubitus positions (Table 5).
The distribution of pulmonary blood flow did not differ between the supine (Figure 2) and the left lateral decubitus (Figure 3) positions. In the supine position, there was a small dorsal-to-ventral gradient (P < 0.05) and a hilar-to-peripheral gradient (P < 0.001;Table 5). The hilar-to-peripheral gradient decreased slightly in the left lateral position (P < 0.05 compared with supine;Table 5).
The difference in blood flow to each lung piece in the supine and left lateral decubitus positions demonstrated no relationship with the left-to-right spatial plane (Figure 4). A significant gravitational effect would show a positive slope. The mean slope for the eight animals was -0.73 +/- 3.57, which was not different from zero (P = 0.58). There was a lot of variability of blood flow to each piece in the two positions (Figure 4).
When analyzed according to each left-to-right plane, the blood flow for all eight animals showed a similar pattern of perfusion in the supine and left lateral decubitus positions (Figure 5). The sum of blood flow was reduced (P < 0.001) in isogravitational slices in the peripheral lung regions in both postures, represented by the extreme left or right planes. Blood flow in ventral versus dorsal regions within each left-to-right plane did not change with rotation to the lateral decubitus position. Gravitational gradients also were not observed within each lung in the left lateral position (Figure 5).
The primary finding of this study is that pulmonary blood flow to each lung in anesthetized, mechanically ventilated dogs did not redistribute with rotation from the supine to the left lateral position. These findings contradict the prediction of increased blood flow to the dependent lung and decreased blood flow to the nondependent lung based on the gravitational model.
Before discussing the significance of these findings, we have to consider the limitations of the methods used in this study. Flow signals in each lung piece normalized to the mean flow of all pieces per animal. The flow signals were corrected for piece weight because many tissues from peripheral parts of the lungs had a volume <1.7 cm3. The radioactive microspheres were injected at FRC, but the lungs were dried ex vivo at TLC. At TLC, all alveoli were uniform in size, and the density was the same; hence, weight is a surrogate of volume. To estimate regional blood flow reliably, the radioactive microspheres have to be totally extracted by the pulmonary microcirculation. Spheres with 15-[micro sign]m diameter are almost completely entrapped by the pulmonary circulation and adequately reflect the distribution of pulmonary blood flow. Deposition of microspheres reflects the true distribution of pulmonary blood flow and is not more susceptible to anatomic variables than are dissolved substances. Regional trapping of 15-[micro sign]m diameter microspheres correlated highly (r = 0.99) with endothelial uptake of simultaneously injected diamine. 
Interpretation of the data regarding distribution of pulmonary blood flow depends on preservation of the in vivo lung size and spatial orientation of the lung. We maintained lung size and shape by carefully opposing the left and right sides in anatomic position and drying the lungs at TLC. Some distortion of pulmonary parenchyma is possible when the lungs were kept inflated with a pressure of 25 cm H2O. Lung volume is greater than in the intact lung, which increases the linear dimensions. The orientation of gravity is slightly different when the lung is suspended by the trachea. The lung may be slightly distorted by this and the lack of physical constraints of the chest wall and diaphragm. In addition, the weight of the heart in vivo may result in some compression of the lung below it. Expansion of the lung from FRC to TLC may move adjacent parenchymal units 1 - 2 cm but has little effect on the overall orientation of the lung. Although these factors are difficult to quantify, they would not be expected to alter isogravitational planes and to affect blood flow differentially in either position. The lungs were oriented in the rigid box when foamed to assure sectioning of the lungs in true isogravitational planes. The postures were studied in random order, ruling out variations over time. The hymodynamics did not change between postures.
Lung pieces with >or= to 25% airways (9% of all samples) were excluded from the analysis of linear gradients (slopes), coefficient of variation of flow, and percent blood flow to each lobe and lung. These pieces are relatively heavy compared with lung parenchyma. As they contribute substantially to the weight of the sample, inclusion of airways would result in erroneously low weight-corrected pulmonary blood flow. To quantify any potential bias, we also evaluated the effect of exclusion of pieces with airways on the results. The absence of redistribution of pulmonary blood flow with rotation from the supine to the left lateral decubitus position was observed with inclusion and exclusion of pieces with airways.
According to the gravitational model, perfusion to the dependent left lung should increase and perfusion to the nondependent right lung should decrease in the lateral position. In addition, a gradient of decreasing blood flow in the left-to-right spatial dimension should be present in the lateral position. In contrast to these predictions, no gravitational gradient in flow was observed in the lateral position in dogs, between the dependent and nondependent lungs and within each lung separately. In addition, spatial mapping of blood flow using this-resolution technique demonstrated essentially identical distributions of perfusion in the supine and lateral positions.
Our results contradict those of older studies in the lateral position in animals [4,5,20]and humans. [3,6,7,21]Methodologic and species differences may play a significant role in the differences in results; however, the exact explanation for the contradictory findings requires further investigation.
The most important methodologic difference involves the lung volume at which flow was normalized. Our study injected microspheres at FRC but analyzed flow in a lung expanded ex vivo to TLC. At TLC, all alveoli are uniform size, and the density of the lung parenchyma is the same. In contrast, imaging studies evaluate flow at FRC, [3,6,7,21]during which parenchymal density and alveolar volume are different in dependent and nondependent regions of the lung. Although perfusion is standardized by the density of lung tissue that would represent the number of alveoli grossly, perfusion in these studies may be poorly adjusted in dependent lung because of the presence of collapsed and smaller alveoli. This may cause a relatively greater perfusion/alveolus in dependent lung regions. In contrast, in our study, expansion of the lung ex vivo would “dilute” the radioactive signal and reduce dependent lung flow compared with the detected by an imaging technique in the intact animal or human. Although this difference may alter the gradients of flow, however, it is unlikely to alter the results relating to the overall percent blood flow to each lung and lobe in each position.
The study most comparable to ours (Reed and Wood ) also measured pulmonary blood flow in dogs using microspheres with flow normalized at TLC. They observed a trend for a gravitational gradient in flow in the left lateral decubitus position, especially in the nondependent right lung, in the four animals studied. The explanation for the difference in our results is not entirely clear. Although the spatial resolution of our study was greater, this is unlikely to explain differences in the results measured in isogravitational planes. Larger (35 vs. 15 [micro sign]m in our study) microspheres were used by Reed and Wood, although differences in sizes of microspheres did not alter the results in other studies. Our study differed from that of Reed and Wood by use of halothane rather than pentobarbital anesthesia. Halothane-induced anesthesia may reduce FRC and result in atelectasis in the dependent lung. [22,23]Atelectasis may decrease blood flow in the dependent lung, which would reduce the gravitational gradient compared with pentobarbital anesthesia. Although differences in the distribution of pulmonary blood flow during halothane-and pentobarbital-induced anesthesia have not yet been studied fully, our pilot data suggest that halothane reduces the gravitational gradient of flow observed in the supine position. It is likely, however, that the amount of atelectasis in this study was small, as partial pressure of oxygen was 89 +/- 5 mmHg with room air ventilation. Our animals also were well hydrated, and the lungs were under zone 3 conditions in both positions. Hemodynamic and blood gas measurements were not reported in the Reed and Wood study. It is possible that their animals were hypocapnic, as they were mechanically ventilated with a greater minute ventilation than in our study. They also may have been less hydrated. The net result would be a decrease in blood flow in the nondependent lung because of the development of zone 1 regions.
Species differences are likely to be important contributors to differences in the results of dogs, which were used in our study, compared with humans [3,6,7,19]and rabbits. Rabbits may have marked variability in blood flow, with entire regions having low perfusion,[section] which may contribute to the gravitational gradient observed in rabbits in the lateral position. Gravity may play a greater role in the distribution of blood flow in upright humans compared with quadruped animals. Using single-photon emission computerized tomography, Ross et al. found that gravity primarily determines the nondependent to dependent distribution in supine and prone humans. In contrast, there is a dorsal preponderance of flow in most quadruped species, including dogs, [5,8,9]sheep, [10,11]and horses. 
The shape of the chest wall is also different in dogs and humans. The greatest dimension of the chest wall in dogs is in the ventral-to-dorsal plane. In contrast, the lateral dimension is larger in humans. This results in a greater hydrostatic pressure gradient in humans compared with dogs in the lateral position. The uppermost parts of the nondependent lung in humans may be under zone 1 conditions, in which alveolar pressure is higher than arterial pressure. Kaneko et al. found that perfusion per alveolus was relatively constant over the lower two thirds of the lungs up to a distance of [almost equal to] 18 cm from the most dependent part in eight healthy men in the lateral position (our dog lungs were [almost equal to] 18 cm in this direction). In the upper third, they found that perfusion per alveolus decreased progressively, and this is also a more peripheral part of the lung, which also would be under zone 1 conditions. In contrast, zone 1 conditions were not present in our dogs. The lack of a gravitational gradient in the lateral position in dogs in the current study therefore may be partially explained by the smaller side-to-side dimensions and lower hydrostatic gradient in the dog lung. In addition, species differences in mediastinal structures and compression of the dependent left lung by the heart also may contribute to reduction in gravitational gradient.
These results cannot be applied to the clinical situation of the mechanically ventilated human in the lateral position. Important species differences in the shape of the chest wall and the physiologic body position may account for the lack of gravitational gradient in the lateral position observed in this study in dogs. In contrast, studies in which perfusion was measured directly in humans demonstrated at least some gravity-dependent component of the perfusion distribution. [3,6,7,21]The one human study that did not demonstrate a decrease in nondependent lung blood in the lateral position assessed blood flow indirectly by elimination of carbon dioxide, a method that is sensitive to changes such as atelectasis. Studies of the distribution of pulmonary blood flow using high spatial resolution techniques, therefore, need to be performed in humans to address the relevance of our results to anesthetized human patients.
The spatial distribution of blood flow to each lung did not change when dogs were turned from the supine to the left lateral position. Methodologic factors and species differences may account for the contradictory findings compared with older studies. These results suggest that factors in addition to gravity are important determinants of pulmonary blood flow in the dog.
[section] Godbey PS, Graham JA, Hanger CC, Presson RG Jr, Doerschuk CM, Wagner WW Jr: Microcirculation perfusion patterns in the surface of dog and rabbit lungs (abstract). FASEB J 1992; 6:2044.