J. Appl. Physiol. 88: 10761083, 2000.

Regional ventilation-perfusion distribution is more uniform in the prone position

MARGARETA MURE,1 KAREN B. DOMINO,2 STEN G. E. LINDAHL,1 MICHAEL P. HLASTALA,3 WILLIAM A. ALTEMEIER,3 AND ROBB W. GLENNY3 1Department of Anesthesiology and Intensive Care, Karolinska Hospital and Institute, SE-171 76 Stockholm, Sweden; and Departments of 2Anesthesiology and 3Medicine and Physiology and Biophysics, University of Washington School of Medicine, Seattle, Washington 98195
Mure, Margareta, Karen B. Domino, Sten G. E. Lindahl, Michael P. Hlastala, William A. Altemeier, and Robb W. Glenny. Regional ventilation-perfusion distribution is more uniform in the prone position. J. Appl. Physiol. 88: 10761083, 2000.The arterial blood PO2 is increased in the prone position in animals and humans because of an A) and perfusion (Q ) matching. improvement in ventilation (V A/Q is unknown. This However, the mechanism of improved V A/Q heterogeneity and the experiment measured regional V A and Q in supine and prone positions correlation between V in pigs. Eight ketamine-diazepam-anesthetized, mechanically ventilated pigs were studied in supine and prone A and Q were measured positions in random order. Regional V using uorescent-labeled aerosols and radioactive-labeled microspheres, respectively. The lungs were dried at total lung capacity and cubed into 603967 small (1.7-cm3) pieces. In the prone position the homogeneity of the ventilation distribu A and tion increased (P 0.030) and the correlation between V increased (correlation coefficient 0.72 0.08 and 0.82 Q 0.06 in supine and prone positions, respectively, P 0.03). A/Q distribution increased in the The homogeneity of the V prone position (P 0.028). We conclude that the improve A/Q matching in the prone position is secondary ment in V A distribution and increased to increased homogeneity of the V A and Q . correlation of regional V aerosol; uorescent microspheres; pulmonary blood owventilation heterogeneity

, and the correlation between them. Using this relaQ tionship, Wilson and Beck (33) postulated that the A/Q distribution is more uniform in the prone than in V the supine posture primarily because of more uniform A and Q in the prone posture. They distributions in V A and Q was assumed that the correlation between V less in the prone position but that this had little impact A/Q distribution because of the uniformity in on the V VA and Q. The present experiment is the rst to A/Q distributions and correlation of measure regional V A and Q in the supine and prone position, regional V thus directly testing the model of Wilson and Beck.
METHODS

ARTERIAL BLOOD OXYGENATION is often improved in the prone position in animals and humans with normal and injured lungs (1, 8, 10, 14, 16, 1820, 31). The prone position improves oxygenation by improving ventilation A/Q ) matching, as measured by the mulperfusion (V tiple inert gas elimination technique (MIGET) (1, 8, 18). Using single-photon emission-computed tomography, Lamm et al. (14) found that the prone posture A/Q heterogeneity in dogs with normal and decreased V oleic acid-injured lungs. However, the mechanism of A/Q heterogeneity in the prone position the decreased V A and Q distributions can be characteris unclear. If V ized as normal distributions in the logarithmic domain, A/Q can be described by V A, the expected variance in V

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1076

This study represents a further analysis of data collected from seven of eight animals by Mure et al. (18), which described the inuence of abdominal distension and position on pulmonary gas exchange by use of the MIGET. One additional animal was added to the present study. Only data from the control conditions are analyzed in this study, and A/Q distribution data were not analyzed as part of regional V the original study. Animal preparation and experimental protocol. The study was approved by the University of Washington Animal Care Committee. The animal preparation and experimental protocol were described in detail in the earlier publication (18). Briey, the investigation was performed in eight 30- to 45-day-old pigs [15.4 2.0 (SD) kg body wt (range 1320 kg)]. The pigs were healthy and free from signicant diseases. The pigs were allowed to eat and drink ad libitum until premedication, which consisted of an intramuscular injection of xylazine (2 mg/kg) and ketamine (20 mg/kg) given 10 min before the start of the investigation. Anesthesia was induced with ketamine (20 mg/kg iv) and diazepam (0.5 mg/kg iv) and continued with a mixture of diazepam (1.7 mg/ml) and ketamine (67 mg/ml) at 4 ml/h. Anesthetic agents for maintenance were given in sufficient doses to prevent spontaneous ventilatory effort and to maintain a surgical plane of anesthesia. No muscle relaxants were used. After tracheotomy and endotracheal tube insertion, all pigs were mechanically ventilated with a fractional inspiratory O2 of 0.4 and a tidal volume of 15 ml/kg at a respiratory rate to achieve normocapnia. Body temperature was adjusted to normal with heating pads. One arterial catheter was inserted into the carotid artery to monitor mean systemic blood pressure and heart rate and another into the femoral artery for blood-gas sampling (model ABL 4, Radiometer, Copenhagen, Denmark). A 5-F pulmonary artery catheter was inserted via the internal jugular vein to measure body temperature and cardiac output in triplicate (Edwards COM 2, Baxter, Irvine, CA). Pulmonary
http://www.jap.org

8750-7587/00 $5.00 Copyright 2000 the American Physiological Society

 A/Q HETEROGENEITY PRONE POSITION REDUCES V

arterial pressure and pulmonary capillary wedge pressure were recorded. Both femoral veins were catheterized. One vein was used for infusion of six inert gases, of which acetone was analyzed in a gas chromatograph (Varian 300, Walnut Creek, CA) to determine anatomic dead space (9, 30). Microspheres and maintenance uids were administered via the second femoral venous catheter. All pigs were studied in prone and supine positions in random order. In the prone position the pigs rested on their abdomen. After a period of 30 min to achieve steady-state conditions, the rst series of measurements were obtained. The animals were allowed to stabilize in the other position for 30 min before the next sets of measurements. A and Q distributions. Regional Measurements of regional V A was measured using inhaled aerosolized uorescent (blueV green, yellow-green, orange, and red) microspheres (26) with a particle size of 1.0 m (FluoSpheres, Molecular Probes, Eugene, OR). Simultaneously with the inhaled microspheres, was measured by injection of radioactive (113Sn, regional Q 103Ru, 95Nb, and 46Sc) microspheres (10) with a particle size of 15 m (Dupont NEN Research Products, Boston, MA). After the last measurement, heparin (10,000 U iv) and papaverine (60 mg iv) were administered, and the animals were exsanguinated while saline was freely infused intravenously. The lungs were harvested and perfused with a dextran solution. The lungs were visually inated to total lung capacity and air-dried for 3 days at transpulmonary pressure of 25 cmH2O. The lobes were glued in their anatomic position with cyanoacrylate glue (Duro Superglue, Locite, Cleveland, OH). The dried lungs were coated with a cold setting foam and then encased in rapidly setting isocyanate foam (2 lb Polyol Isocyanate, International Sales, Seattle, WA). A miter box was used to cut the lungs into 1.7-cm3 cubes. Any foam adhering to the lung piece was removed, and the pieces were weighed. Pieces weighing 8.0 mg were discarded. Each lung piece was assigned a unique three-dimensional (x, y, z) coordinate, where x represents distance in the right-to-left plane, y represents distance in the dorsal-to-ventral plane, and z represents distance in the caudal-to-cranial plane. Piece radioactivity was read in a gamma counter (Minaxi gamma counter system, model 5550, Packard, Downers Grove, IL). Each piece was soaked for 48 h in 1.5 ml of 2-ethoxyethyl acetate (Cellosolve, Aldrich Chemical, Milwaukee, WI) to extract the uorescent markers. The extract was transferred into a cuvette, and the uorescent intensities of each color were measured in a uorescent spectrophotometer (model LS 50B, Perkin-Elmer, Norwalk, CT). Calculations. The data were treated in four different fashions depending on the analysis. Linear gradients and A and Q distributions were detercoefficient of variation of V mined in milliliters per minute. Natural logarithm (ln) trans A and Q data were used to calculate variances formations of V and the correlation coefficient as postulated by Wilson and A/Q ratios Beck (33). A logarithmic (log) transformation of V A/Q distribution. was used to assess linear gradients in the V A/Q distributions were calcuFlow- and ventilation-weighted V A/Q distributions and lated from the microsphere-derived V were compared with those derived by MIGET in the traditional log domain. Ventilation to each lung piece (ml/min) was calculated as follows

1077

Perfusion to each lung piece (ml/min) was calculated as follows

cardiac output (piece counts/total counts) Q

(2)

A to each piece was then divided by Q to the same piece to V A/Q ratio to each piece. obtain the V A and Q was assessed by the The heterogeneity of V coefficient of variation (SD/mean) and the variance (2) of A and lnQ . The coefficient of correlation () between V A lnV was calculated using the Pearson coefficient of correlaand Q A/Q distribution was calcution. The heterogeneity of the V 2 A/Q (V lated directly by the variance of the lnV A/Q obs) and indirectly using an equation derived by Wilson and Beck (33) 2 (V A/Q cal), where VA/Q, VA, and Q are measured in the ln domain according to the following equation
2 2 2 V AQ A/Q cal V A Q 2V

(3)

A, Q , and log V A/Q distributions were characterSlopes in V ized as a linear function of distance in centimeters in the dorsal-to-ventral (y) and caudal-to-cranial (z) spatial vectors by using least-squares regression analysis. Although somewhat of an oversimplication, a linear slope is an easily understood method to describe a general trend in the data. To assess the gravitational (y) gradient within a transverse A, Q , and log V A/Q in the dorsal-tosection, new slopes of V ventral direction were recalculated after correction for trends in the caudal-to-cranial (z) direction (32). Because MIGET relies on whole lung inert gas exchange to A/Q distribution, it cannot directly measure determine the V A/Q compartments. Instead, MIGET calcudiscrete regional V A/Q lates the amount of blood ow and ventilation to 50 V compartments evenly distributed along a logarithmic axis between 0.0005 and 1,000. The MIGET software then calculates a perfusion- or ventilation-weighted mean and standard A/Q distribution (log SDQ deviation of this V and log SDV A, respectively). In contrast, the microsphere method measures A, Q , and therefore V A/Q to many discrete compartments. To V compare microsphere-measured data with MIGET results, A/Q data measured with microspheres must be weighted the V in a manner analogous to that used by MIGET. The mean of A/Q distribution (Q) is calculated as the perfusion-weighted V follows

Qe

3 4
iln Q i1 Q i Q
n n

12
i V
i

i1

(4)

i and V i are the blood ow and ventilation, respecwhere Q tively, to piece i of n pieces and ln is the natural logarithm of A/Q ratio. The standard deviation of the perfusionthe V A/Q distribution (log SDQ ) is calculated by weighted V

log SDQ

A (total ventilation dead space ventilation) V

(piece uorescence / total uorescence)

(1)

A/Q distribution (VA) The mean of the ventilation-weighted V A/Q and the standard deviation of the ventilation-weighted V A) are calculated in a similar manner. distribution (log SDV

i1

53

i ln Q

i V i Q

ln 1Q 2 Q
n

i V

i1

Q
i1 i

46
2

Q
i1

(5)

1078

A/Q HETEROGENEITY PRONE POSITION REDUCES V

A and Q distribuFrom these data, standard deviations of V A/Q (log SDV A and log SDQ , tions with reference to log V A reects heterogerespectively) were calculated (3). Log SDV A distribution with reference to V A/Q . Log SDQ neity in the V distribution with reference to reects heterogeneity in the Q A/Q . These values are therefore similar in concept to log V A and log SDQ derived from MIGET (9, 30). SDV Statistics. Slopes of linear gradients from all animals were compared with zero with a single-sample two-tailed t-test. Differences in all heterogeneity data and ow gradients from all animals were compared in supine and prone positions by two-tailed paired t-tests. Differences in the correlation coefficient were compared using Fishers z transformation. P 0.05 was considered statistically signicant. Values are means SD.
RESULTS

Hemodynamics and respiratory variables are presented in Table 1. Arterial PO2 (PaO2) increased and alveolar-arterial PO2 difference decreased in the prone position (P 0.03). Otherwise, there were no differences in these variables between the supine and prone positions. The number of lung pieces analyzed per animal ranged between 603 and 967 pieces, with 1214 rightto-left planes, 1013 dorsal-to-ventral planes, and 16 21 caudal-to-cranial planes. There was considerable A and Q distribuisogravitational heterogeneity of the V A was tions (Fig. 1). The coefficient of variation of V decreased in the prone compared with the supine position (P 0.012; Table 2). Although the coefficient of decreased, it was not statistically signivariation of Q A was increased in ventral cant (P 0.11; Table 2). V regions in both positions (Fig. 1), as reected by dorsalto-ventral (vertical) gradients greater than zero A was increased in the cranial compared (Table 2). V Table 1. Hemodynamic and respiratory variables
Supine Prone

A; A) and perfusion (Q ; B) as a function of Fig. 1. Ventilation (V dorsal-to-ventral distance in supine and prone position in a representative pig. Independent and dependent axes have been interchanged A and Q are plotted for each lung piece at each for presentation. V plane in dorsal-to-ventral (y) directions. Solid line, linear regression equation. Drawings of lungs serve as schematics to signify position of pig and are not accurate representations of lung shape. Linear A vs. dorsal-to-ventral distance: V A 0.43 regression equations for V A 0.27 (cm of (cm of lung) 2.18 (r 0.31) in supine position and V A in ventral lung regions lung) 2.29 (r 0.25) in prone position. V was increased in both positions, although considerable isogravita A heterogeneity was present. Linear regression equations for tional V vs. dorsal-to-ventral distance: Q 0.2 (cm of lung) 4.44 (r Q 0.1 (cm of lung) 2.22 (r 0.13) in 0.21) in supine position and Q tended to be higher dorsally, prone position. In supine position, Q although there was considerable isogravitational heterogeneity. with the caudal regions in the supine position (P A in the supine 0.01; Table 2). The vertical gradient in V position remained after correction for trends in the caudal-to-cranial dimension (Table 2). The magnitude A decreased (P of the caudal-to-cranial gradient in V tended to be 0.03) in the prone position (Table 2). Q increased in dorsal and cranial regions in the supine position (Fig. 1), although the dorsal-to-ventral (vertical) and caudal-to-cranial gradients were not signicantly different from zero (P 0.18 and P 0.15, berespectively). However, the vertical gradient in Q came signicantly different from zero in the supine position, after correction for trends in the caudal-tocranial dimension (P 0.004; Table 2). In contrast, there were no vertical or caudal-to-cranial gradients in distribution in the prone position. V A and Q the Q tended to decrease in the peripheral lung regions (Fig. 1), but when normalized by piece weight and mean ventilation or ow, respectively, this trend was not observed.

Ventilation Plateau pressure, cmH2O VT, ml Compliance, ml/cmH2O RR, breaths/min HR, beats/min Psa, mmHg Ppa, cmH2O Pw, cmH2O , l/min Q PAO2 PaO2, Torr PaO2 , Torr PaCO2 , Torr PvO2 , Torr pH 13 3 197 42 15.2 3 27 2 109 16 94 21 25 6 10 6 2.6 1.0 Gas exchange 52 26* 193 45* 39 4 42 7 7.38 0.07 23 13* 223 29* 40 3 44 7 7.38 0.09 13 3 196 30 15.5 2.8 26 3 105 19 103 19 26 6 95 2.2 0.7

Hemodynamics

Values are means SD; n 8. Fraction of inspiratory O2 0.40. VT, tidal volume; RR, respiratory rate; HR, heart rate; Psa, mean systemic arterial pressure; Ppa, mean pulmonary arterial pressure; , blood ow; PAO2 , alveolar Pw, pulmonary capillary wedge pressure; Q PO2 ; PaO2 , arterial PO2 ; PaCO2 , arterial PCO2 ; PvO2 , mixed venous PO2 . * Signicantly different from supine, P 0.05.

 A/Q HETEROGENEITY PRONE POSITION REDUCES V Table 2. Coefficient of variation and gradients as linear function of spatial vectors
Gradients Coeff of Variation, % Dorsal-to-ventral (vertical) Dorsal-to-ventral (corrected) Caudal-to-cranial

1079

Supine

A V Q A/Q Log V A V Q A/Q Log V

81.6 13.1 73.2 6.4

0.228 0.351* 0.083 0.156 0.098 0.058* Prone 0.122 0.135* 0.011 0.056 0.005 0.021

0.183 0.230* 0.124 0.084* 0.077 0.039* 0.072 0.099 0.019 0.060 0.005 0.015

0.215 0.180* 0.070 0.122 0.043 0.028* 0.079 0.104 0.024 0.088 0.003 0.013

66.8 7.1 67.9 8.8

A) and perfusion (Q ) are ml min1 cm1; gradients for log V A/Q are cm1. * P Values are means SD; n 8. Gradients for ventilation (V 0.05, gradient vs. zero; P 0.05, supine vs. prone; P 0.01, supine vs. prone; P 0.001, supine vs. prone.

A/Q distribution also had signicant isogravitaThe V A/Q ratios became more tional heterogeneity; however, V uniform when the animals were prone (Fig. 2, Table 2). In the supine position there were signicant dorsal-toventral (P 0.002) and caudal-to-cranial gradients A/Q (Fig. 2, Table 2), such that V A (P 0.004) in log V in ventral was relatively increased compared with Q and cranial lung regions (Fig. 2, Table 2). The vertical gradient remained after correction of trends in the caudal-to-cranial direction (P 0.001). In contrast, there was no dorsal-to-ventral (vertical) gradient of log A/Q in the prone position (Fig. 2, Table 2). The vertical V A/Q gradient and caudal-to-cranial gradients of log V decreased in the prone compared with the supine position (P 0.002 and P 0.005, respectively; Table 2). A distribution, measured The heterogeneity of the V 2 by V A, decreased in the prone position ( P 0.03; distriTable 3). In contrast, the heterogeneity of the Q 2 bution, measured by Q , did not change signicantly A (P 0.18; Table 3). Correlation between regional V increased in the prone position ( 0.82 0.06 and Q and 0.72 0.08 in prone and supine positions, respec A/Q heterogeneity, as tively, P 0.03; Fig. 3, Table 3). V

2 measured by V A/Q , decreased in the prone position (P 2 0.028; Table 3). The variable of V A/Q calculated indirectly 2 by Wilson and Beck (33), V A/Q cal, was identical to the ob2 A/Q (V served variance of V A/Q obs ; Table 3). A and Q distributions When the heterogeneity of the V was compared with reference to log VA/Q ratios (Table 4), calculated according to Altemeier et al. (3), the prone A (P position was associated with a lower log SDV (P 0.015). The heterogeneity of 0.010) and log SDQ A/Q distribution, measured by log SDV A/Q , was also V lower in the prone position (P 0.016; Table 4).

DISCUSSION

A/Q as a function of dorsal-to-ventral (y) distance in Fig. 2. Log V supine (A) and prone (B) position in same pig used in Fig. 1. Independent and dependent axes have been interchanged for presen A/Q 0.14 (cm of lung) tation. Linear regression equations: log V A/Q 0.01 (cm of lung) 0.31 0.6 (r 0.64) in supine position and V A/Q was lower in (r 0.07) in prone position. In supine position, log V A/Q was dorsal and higher in ventral lung regions. Distribution of V more uniform in prone position, although considerable isogravitational heterogeneity remained.

A/Q distribuThe major nding of this study is that V tion became more uniform in the prone position be A distribucause of an increase in homogeneity of the V A tion and an increase in correlation between regional V . and Q Methodological issues. Before discussing the signicance of these ndings, we have to consider the limitations of the methods used. The lungs were dried ex vivo at total lung capacity, giving all alveoli uniform size. To estimate regional blood ow reliably, the radioactive microspheres need to be totally trapped by the pulmonary microcirculation. Microspheres with a 15-m diameter are almost completely entrapped by the pulmonary circulation (25) and adequately reect the distribution of blood ow (6, 16). Studies comparing the distribution of N,N,N-trimethyl-N[2-hydroxy-3-methyl-5-iodobenzyl]1,3-propanediamine, a diamine with a near-complete rst-pass extraction by the lungs, have shown that the principle used in the present study reects regional pulmonary blood ow (16). With use of similar reasoning, instead of radioactive microspheres for calculation of lung perfusion, aerosolized uorescent microspheres were used to measure ventilation. The uorescent signals were recently shown by Robertson et al. (26) and Melsom et al. (17) to A. We therefore employed the method derepresent V scribed by Altemeier et al. (3) to simultaneously mea A and Q in 1.7-cm3 cubes of lung with sure regional V microsphere techniques.

1080

A/Q HETEROGENEITY PRONE POSITION REDUCES V

A and Q distributions Table 3. Heterogeneity of V

Animal No.
2 V A 2 Q

V A/Q Prone Supine Prone Supine

2 V A/Q cal

2 V A/Q obs

Supine

Prone

Supine

Prone

Supine

Prone

1 2 3 4 5 6 7 8 Mean SD
2

1.27 5.15 2.05 1.35 2.99 2.32 1.64 2.39 2.39 1.25*

1.24 1.20 0.71 0.82 1.13 0.76 1.57 1.85 1.16 0.40*
2

1.43 2.12 1.86 1.37 1.61 1.26 2.83 1.67 1.77 0.51

1.04 1.07 1.96 0.99 1.18 0.76 2.67 2.35 1.50 0.72

0.78 0.66 0.81 0.82 0.70 0.67 0.58 0.73 0.72 0.08*

0.85 0.90 0.80 0.84 0.78 0.74 0.79 0.89 0.82 0.06*

0.59 2.89 0.76 0.50 1.55 1.28 1.96 1.14 1.33 0.80*

0.35 0.23 0.77 0.30 0.50 0.40 1.02 0.47 0.51 0.26*
2

0.59 2.90 0.75 0.48 1.56 1.28 1.96 1.14 1.33 0.81*

0.35 0.23 0.78 0.30 0.48 0.40 1.03 0.47 0.51 0.27*

A distribution; Q distribution; V A and Q distributions; V A/Q , correlation between V V A , Variance of V , variance of Q A/Q cal , calculated variance 2 A/Q distribution; V A/Q distribution. * P 0.05, supine vs. prone. of V A/Q obs , observed variance of V Because the lungs in this study were dissected along an orthogonal grid, peripheral lung pieces are not full cubes. Hence, all the lung pieces used in this study are not uniform in volume. We used weight normalization in the past to correct for this artifact. We have chosen not to weight normalize ows in this analysis, because we believe that respiratory and inert gas exchange is determined by the relationship between local ventilation and perfusion and their ow rates in milliliters per minute. The additional variability in lung piece size A and Q . The adds to the observed heterogeneity of V values of VA and Q heterogeneity in Table 2 are therefore signicantly larger than previously reported. Because this increased variability occurs in supine and prone postures, the relative differences between the postures remain similar to those presented without use of weight-normalized ows, and the conclusions of the study are unchanged. The correlation between local ventilation and perfusion is also slightly increased: small pieces tend to have less ventilation and perfusion, whereas larger pieces have greater ventilation and perfusion. Although directional gradients in ventilation and perfusion are presented in milliliters per minute per centimeter, we also explored the spatial A and Q after weight normalization. distributions of V There were no signicant differences in the directional gradients between weight-normalized ows and ows in milliliters per minute. A/Q . During mechanical Spatial distributions of V A was increased to ventral ventilation, we found that V lung regions in the supine and prone positions, as demonstrated by signicant dorsal-to-ventral (vertical) A to dependent gradients (Fig. 1, Table 2). Increased V ventral lung in the prone position has been demonstrated previously in humans (5, 12, 23) and animals (14). Our results are different from those in supine unanesthetized, spontaneously breathing humans, in A was increased to dorsal, dependent lung (5, whom V 12, 22, 24). The variations may reect differences between the mechanical and spontaneous ventilation, species differences, and methodological factors, such as use of aerosols vs. radioactive-labeled gases, spatial resolution, and the lung volume at which ventilation was normalized. The role of anesthesia and mechanical ventilation is likely to be quite signicant, inasmuch as A in supine it reduced the gravitational gradient of V humans by increasing ventilation of nondependent ventral lung (24). to dorsal lung regions in the supine The increase in Q position (Fig. 1, Table 2) is consistent with prior work in humans (4, 12, 15, 21) and animals (7, 10, 17, 31). Our in the prone ndings of a lack of a vertical gradient of Q position (Fig. 1, Table 2) are similar to prior results in animals with use of similar methodology (10, 31, 32). Beck and Rehder (7) demonstrated a higher conductance for blood ow in dorsal lung regions in the dog, to dorsal which may result in a relative increase in Q lung regions in quadruped animals independent of position (7, 10, 31).

A as a function of Q in each individual lung Fig. 3. V piece plotted according to method of Altemeier et al. (3) in same pig used in Figs. 12 (A is supine, B is A and Q are highly correlated in prone). Regional V A and both positions, although correlation between V is higher in prone position. Q

 A/Q HETEROGENEITY PRONE POSITION REDUCES V

1081

A, Q , and V A/Q Table 4. Heterogeneity of V A/Q ratio distributions with reference to log V
Supine Prone P

Log Log A/Q Log SDV

A SDV SDQ

0.798 0.252 0.603 0.205 1.110 0.339

0.514 0.143 0.398 0.065 0.691 0.177

0.010 0.015 0.016

A A , standard deviation of V Values are means SD; n 8. Log SDV A/Q ratio; log SDQ , standard distribution with reference to log V distribution with reference to log V A/Q ratio; log deviation of Q A/Q distribution. A/Q , standard deviation of log V SDV Our study found vertical and caudal-to-cranial gradi A/Q distributions in the supine position, ents in the V A/Q ratios were lower in dorsal and caudal such that V lung regions than in ventral and cranial regions (Fig. 2, Table 2). In the prone position the distribution of A/Q was more uniform, reected by a lack of vertical V and caudal-to-cranial gradients. Although our results are consistent with those in anesthetized, mechanically ventilated animals (14), studies in unanesthetized, spontaneously ventilating humans have demonstrated A/Q , such that V A/Q is a gravitational dependence of V increased in dependent lung in supine and prone positions (12, 22). However, the presence of anesthesia and mechanical ventilation may reverse this relation A/Q to nondependent lung, as ship and increase V shown by Landmark et al. (15). This nding is in agreement with that in the present series when the animals were in the supine position. The more uniform A/Q in the prone position contributes to distribution of V A/Q in that posture and constitutes the well-matched V the primary mechanism for increases in PaO2 (1, 8, 10, 18, 19) and improvements in pulmonary gas exchange reported in the prone position (8, 14, 18). A/Q . The prone position increased Heterogeneity of V A/Q distribution as a result of homogeneity of the V A distribution and inincreased homogeneity of the V A and Q . Improvecreased correlation between regional V A distribution is consisment in the uniformity of the V tent with studies that suggest a more even distribution A in the prone position (1, 5, 14). This speaks of V against a marked overventilation in nondependent A may regions, as we observed in the supine position. V be more uniform in the prone position, because the pleural pressure gradient is more uniform, inasmuch as there is less change in pleural pressure per centime heterogeneity did ter of distance (20, 34). Although Q not change signicantly with the prone position in our study with pigs, increases in the homogeneity of the Q distribution have been demonstrated in the prone position in dogs (8, 10), sheep (31), and humans (21). The lack of change in the present study probably reects inadequate power due to interanimal variability, although methodological differences and species differences [intensity of the hypoxic pulmonary vasoconstrictor (HPV) response] may be important. The more intense HPV response in pigs (13) may attenuate heterogeneity. The presposition-related differences in Q ent study uniquely demonstrates an improvement in the A and Q in the prone position. correlation of regional V

Wilson and Beck (33) speculated that the prone A/Q heterogeneity by improved position decreases V A and Q distributions. They postuhomogeneity of the V lated that the VA distribution was more uniform in the prone position, because there is no gravitationally related pleural pressure gradient. On the basis of studies in the dog, they estimated that two-thirds of the A/Q is a result of nonuniform Q and variance in V A. Wilson and one-third is the result of nonuniform V A and Q must be weakly Beck reasoned that regional V A and Q do correlated in the prone position, because V not share a gravitational inuence. Although the scale of measurement (1.7 cm3) used in the present study is considerably larger than some of the data used by Wilson and Beck, their model is not scale dependent. A/Q The variance we observed in the distribution of V 2 (V ) exactly equaled the variance in V A /Q predicted A/Q obs 2 by the their model (V A/Q cal ) in both positions (Table 3). However, the magnitude of the variance was considerably larger (Table 3) than was estimated by Wilson and A was Beck. In addition, the variance of regional V , in contrast to the larger than the variance of regional Q prediction for dogs (33). This difference in results may be due to species differences, scale of measurement, or comparability of techniques used to measure regional A and Q . Dogs, which have extensive collateral ventiV lation (13), may have a more homogeneous distribution of ventilation than pigs. In addition, ventilation in 1.7-cm3 lung pieces is primarily dependent on convective gas movement, whereas in smaller units of measurement, gas diffusion dominates. Wilson and Beck A also relied on different methods to measure regional V (e.g., parenchymal density, external detectors) and Q (e.g., microspheres). A and Q , as was A high correlation between regional V demonstrated in the present study, has also been observed using identical methodology in the prone pig (26). Although counter to the prediction of Wilson and A/Q is Beck (33), an excellent correlation of regional V not surprising because of the importance of anatomic (10), and probably structure in determining regional Q A, and physiological mechanisms, such as HPV and V A/Q matchcollateral ventilation, which act to improve V A ing on the local level. It is possible, however, that V correlation may be lower when a smaller scale of and Q measurement is used. Relationship with MIGET-derived indexes of heterogeneity. The microsphere method has an advantage over traditional measurements of gas exchange, in that it A, Q , and provides spatial measurements of regional V A/Q . However, microsphere-measured data may be V compared with data from more traditional methods, such as MIGET with appropriate transformation. In and log SDV A calculated these experiments, log SDQ from the microsphere data were less than the previously reported results using MIGET (18). This repre A/Q heterogeneity by sents an underestimation of V microspheres or an overestimation by MIGET. The microsphere method may potentially underestimate A/Q heterogeneity because of its resolution limit true V

1082

A/Q HETEROGENEITY PRONE POSITION REDUCES V the ability to determine the mechanism for the change A/Q matching, i.e., changes in regional V A distriin V distribution, and/or correlation of bution, regional Q A and Q . regional V A/Q distribution was more uniform In summary, the V in anesthetized, mechanically ventilated pigs in the A distribution prone position. The homogeneity of the V A and Q was imwas increased, and correlation of V proved.
The authors gratefully acknowledge the excellent secretarial assistance of L. Hubbard-Hamacher and the expert technical help of D. An, E. Anderson, and Dr. S. Bernard in completion of the studies. This study was supported by National Heart, Lung, and Blood Institute Grants HL-12174 and HL-24163, The Swedish Heart and Lung Association, and The Swedish Society of Medicine (Carin Tryggers Minnesfond). Address for reprint requests and other correspondence: K. B. Domino, Dept. of Anesthesiology, University of Washington, Box 356540, Seattle, WA 98195-6540 (E-mail: kdomino@u.washington. edu). Received 8 April 1999; accepted in nal form 25 October 1999. REFERENCES 1. Albert, R. K., D. Leasa, M. Sanderson, H. T. Robertson, and M. P. Hlastala. The prone position improves arterial oxygenation and reduces shunt in oleic acid-induced acute lung injury. Am. Rev. Respir. Dis. 135: 628633, 1987. 2. Altemeier, W. A., M. Mure, S. E. McKinney, H. T. Robertson, and R. W. Glenny. Regional ventilation has fractal properties that are posture dependent (Abstract). Am. J. Respir. Crit. Care Med. 155: A117, 1998. 3. Altemeier, W. A., H. T. Robertson, and R. W. Glenny. Pulmonary gas-exchange analysis by using simultaneous deposition of aerosolized and injected microspheres. J. Appl. Physiol. 85: 23442351, 1998. 4. Amis, T. C., H. A. Jones, and J. M. B. Hughes. Effect of posture A/Q on inter-regional distribution of pulmonary perfusion and V ratios in man. Respir. Physiol. 56: 169182, 1984. 5. Amis, T. C., H. A. Jones, and J. M. B. Hughes. Effect of posture on inter-regional distribution of pulmonary ventilation in man. Respir. Physiol. 56: 145167, 1984. 6. Beck, K. C. Regional trapping of microspheres in the lung compares well with regional blood ow. J. Appl. Physiol. 63: 883889, 1987. 7. Beck, K. C., and K. Rehder. Differences in regional vascular conductances in isolated dog lungs. J. Appl. Physiol. 61: 530 538, 1986. 8. Beck, K. C., J. Vettermann, and K. Rehder. Gas exchange in dogs in the prone and supine positions. J. Appl. Physiol. 72: 22922297, 1992. A/Q distributions 9. Evans, J. W., and P. D. Wagner. Limits on V from analysis of experimental inert gas elimination. J. Appl. Physiol. 42: 889898, 1977. 10. Glenny, R. W., W. J. E. Lamm, R. K. Albert, and H. T. Robertson. Gravity is a minor determinant of pulmonary blood ow distribution. J. Appl. Physiol. 71: 620629, 1991. 11. Glenny, R. W., and H. T. Robertson. Fractal properties of pulmonary blood ow: characterization of spatial heterogeneity. J. Appl. Physiol. 69: 532545, 1990. 12. Kaneko, K., J. Milic-Emili, M. B. Dolovich, A. Dawson, and D. V. Bates. Regional distribution of ventilation and perfusion as a function of body position. J. Appl. Physiol. 21: 767777, 1966. 13. Kuriyama, T., L. P. Latham, L. D. Horwitz, J. T. Reeves, and W. W. Wagner, Jr. Role of collateral ventilation in ventilationperfusion balance. J. Appl. Physiol. 56: 15001506, 1984. 14. Lamm, W. J. E., M. M. Graham, and R. K. Albert. Mechanism by which the prone position improves oxygenation in acute lung injury. Am. J. Respir. Crit. Care Med. 150: 184193, 1994. 15. Landmark, S. J., T. J. Knopp, K. Rehder, and A. D. Sessler. /Q in awake and anestheRegional pulmonary perfusion and V tized-paralyzed man. J. Appl. Physiol. 43: 9931000, 1977.

of 1.7 cm3. Observed heterogeneity of regional perfusion increases in a predictable fashion as resolution increases (11). Similarly, the observed heterogeneity of ventilation increases as resolution improves at least to and likely beyond the resolution obtained in this study (2, 27). Given the relationship between the variances of A/Q , Q , and V A distributions dened by Eq. 4, the V improved resolution will increase the observed hetero A/Q distribution, unless the regional geneity of the V A and Q increases. Alternatively, correlation between V A/Q heterogeneity MIGET may overestimate the true V because of airway excretion of highly soluble gases A/Q (28, 29) or because of enforced smoothing of the V and distribution. This effectively limits how different Q A data points can be assigned to compartments with V A/Q ratios. similar V Altemeier et al. (3) found that measurement of A/Q with microspheres more accurately preregional V dicted PaO2 and arterial PCO2 than MIGET in normal lungs, although microspheres underestimated areas A/Q ratios after administration of glass emboli with low V (3). The correlation between measured (by MIGET) and predicted (by microspheres) inert gas retention was high (r 0.99) in normal lungs (3). These results A/Q suggest that, in the normal lung, analysis of regional V with aerosolized and injected microspheres is a valid method to study pulmonary gas exchange and has the advantage of providing high spatial resolution (3). Although PaO2 increased in the prone position in the , log present study, MIGET indexes, including log SDQ A, and the arterial-alveolar difference area, were SDV not signicantly different with control conditions (18). In contrast, we observed signicant decreases in log A, log SDQ , and log SDV A/Q derived simultaneously SDV using microspheres. The lack of sensitivity of MIGET to detect small, but physiologically signicant, changes in A/Q heterogeneity in the normal lung may be the V result of errors induced by MIGET algorithms and smoothing procedures and/or airway excretion of highly soluble gases (29). Comparison of gas exchange data derived from microspheres in this study to the gasexchange indexes measured by MIGET (18) suggests that the microsphere technique may possess greater A/Q in the normal sensitivity to detect changes in V lung. The present study nicely illustrates that changes in , as obtained using MIGET, do not necessarily log SDQ changes. Inasmuch as log SDQ mean that regional Q distribution with referreects the variance of the Q A/Q ratio, a change in the V A distribution ence to the V , and/or correlation of VA and Q will change log SDQ distribution is unchanged. A similar even if the Q A distribution for log SDV A. reasoning applies to the V Therefore, inferences about changes in the regional A or Q distributions cannot be accurately made using V MIGET-derived variables. In addition, a more homog A or Q distribution does not necessarily mean enous V A/Q improved arterial blood oxygenation or decreased V heterogeneity. A unique advantage of microsphere A/Q distributions over MIGET is, therefore, derived V

 A/Q HETEROGENEITY PRONE POSITION REDUCES V

16. Melsom, M. N., T. Flatebo, J. Kramer-Johansen, A. Aulie, O. V. Sjaastad, P. O. Iversen, and G. Nicolaysen. Both gravity- and non-gravity-dependent factors determine regional blood ow within the goat lung. Acta Physiol. Scand. 153: 343353, 1995. 17. Melsom, M. N., J. Kramer-Johansen, T. Flatebo, C. Muller, and G. Nicolaysen. Distribution of pulmonary ventilation and perfusion measured simultaneously in awake goats. Acta Physiol. Scand. 159: 199208, 1997. 18. Mure, M., R. W. Glenny, K. B. Domino, and M. P. Hlastala. Pulmonary gas exchange improves in the prone position with abdominal distension. Am. J. Respir. Crit. Care Med. 157: 17851790, 1998. 19. Mure, M., C.-R. Martling, and S. G. E. Lindahl. Dramatic effect on oxygenation in patients with severe acute lung insufficiency treated in the prone position. Crit. Care Med. 25: 1539 1544, 1997. 20. Mutoh, T., R. J. Guest, W. J. E. Lamm, and R. K. Albert. Prone position alters the effect of volume overload on regional pleural pressures and improves hypoxemia in pigs in vivo. Am. Rev. Respir. Dis. 146: 300306, 1992. 21. Nyren, S., M. Mure, H. Jacobsson, S. A. Larsson, and S. G. E. Lindahl. Pulmonary perfusion is more uniform in the prone than in the supine position: scintigraphy in healthy humans. J. Appl. Physiol. 86: 11351141, 1999. 22. Orphanidou, D., J. M. B. Hughes, M. J. Myers, A.-R. AlSuhali, and B. Henderson. Tomography of regional ventilation and perfusion using krypton 81m in normal subjects and asthmatic patients. Thorax 41: 542551, 1986. 23. Rehder, K., T. J. Knopp, and A. D. Sessler. Regional intrapulmonary gas distribution in awake and anesthetized-paralyzed prone man. J. Appl. Physiol. 45: 528535, 1978. 24. Rehder, K., A. D. Sessler, and J. R. Rodarte. Regional intrapulmonary gas distribution in awake and anesthetizedparalyzed man. J. Appl. Physiol. 42: 391402, 1977.

1083

25. Ring, G. C., A. S. Blum, T. Kurbatov, W. G. Moss, and W. Smith. Size of microspheres passing through pulmonary circuit in the dog. Am. J. Physiol. 200: 11911196, 1961. 26. Robertson, H. T., R. W. Glenny, D. Stanford, L. M. McInnes, D. L. Luchtel, and D. Covert. High-resolution maps of regional ventilation utilizing inhaled uorescent microspheres. J. Appl. Physiol. 82: 943953, 1997. 27. Rodarte, J. R., M. Chaniotakis, and T. A. Wilson. Variability of parenchymal expansion measured by computed tomography. J. Appl. Physiol. 67: 226231, 1989. 28. Souders, J. E., S. C. George, N. L. Polissar, E. R. Swenson, and M. P. Hlastala. Tracheal gas exchange: perfusion-related differences in inert gas elimination. J. Appl. Physiol. 79: 918 928, 1995. 29. Swenson, E. R., H. T. Robertson, N. L. Polissar, M. E. Middaugh, and M. P. Hlastala. Conducting airway gas exchange: diffusion-related differences in inert gas elimination. J. Appl. Physiol. 72: 15811588, 1992. 30. Wagner, P. D., H. A. Saltzman, and J. B. West. Measurement of continuous distributions of ventilation-perfusion ratios: theory. J. Appl. Physiol. 36: 588599, 1974. 31. Walther, S. M., K. B. Domino, R. W. Glenny, and M. P. Hlastala. Pulmonary blood ow distribution in sheep: effects of anesthesia, mechanical ventilation, and change in posture. Anesthesiology 87: 335342, 1997. 32. Walther, S. M., K. B. Domino, R. W. Glenny, N. L. Polissar, and M. P. Hlastala. Pulmonary blood ow distribution has a hilar-to-peripheral gradient in awake, prone sheep. J. Appl. Physiol. 82: 678685, 1997. 33. Wilson, T. A., and K. C. Beck. Contributions of ventilation and A/Q distribution. J. Appl. perfusion inhomogeneities to the V Physiol. 72: 22982304, 1992. 34. Yang, Q.-H., M. R. Kaplowitz, and S. J. Lai-Fook. Regional variations in lung expansion in rabbits: prone vs. supine positions. J. Appl. Physiol. 67: 13711376, 1989.