Marked differences between prone and supine sheep in effect of PEEP on perfusion distribution in zone II lung

Sten M. Walther, Mats J. Johansson, Torun Flatebø, Anne Nicolaysen, Gunnar Nicolaysen


The classic four-zone model of lung blood flow distribution has been questioned. We asked whether the effect of positive end-expiratory pressure (PEEP) is different between the prone and supine position for lung tissue in the same zonal condition. Anesthetized and mechanically ventilated prone (n = 6) and supine (n = 5) sheep were studied at 0, 10, and 20 cmH2O PEEP. Perfusion was measured with intravenous infusion of radiolabeled 15-μm microspheres. The right lung was dried at total lung capacity and diced into pieces (≈1.5 cm3), keeping track of the spatial location of each piece. Radioactivity per unit weight was determined and normalized to the mean value for each condition and animal. In the supine posture, perfusion to nondependent lung regions decreased with little relative perfusion in nondependent horizontal lung planes at 10 and 20 cmH2O PEEP. In the prone position, the effect of PEEP was markedly different with substantial perfusion remaining in nondependent lung regions and even increasing in these regions with 20 cmH2O PEEP. Vertical blood flow gradients in zone II lung were large in supine, but surprisingly absent in prone, animals. Isogravitational perfusion heterogeneity was smaller in prone than in supine animals at all PEEP levels. Redistribution of pulmonary perfusion by PEEP ventilation in supine was largely as predicted by the zonal model in marked contrast to the findings in prone. The differences between postures in blood flow distribution within zone II strongly indicate that factors in addition to pulmonary arterial, venous, and alveolar pressure play important roles in determining perfusion distribution in the in situ lung. We suggest that regional variation in lung volume through the effect on vascular resistance is one such factor and that chest wall conformation and thoracic contents determine regional lung volume.

  • regional blood flow
  • perfusion heterogeneity
  • microspheres
  • pulmonary circulation
  • gravity
  • positive end-expiratory pressure

distribution of pulmonary blood flow is thought to be determined by the interplay between alveolar, arterial, and venous pressures resulting in a systematic distribution down the lung (12). This familiar concept, the zone I, II, and III model, evolved from studies demonstrating that pulmonary exchange of carbon dioxide, labeled with a short-lived isotope, differed between upper and lower lung regions (30). A fourth zone was added when Hughes and coworkers (12) using intravenously injected radioactive xenon found a reduction in blood flow in the most dependent lung regions. Compression of extra-alveolar vessels by increased interstitial pressure was postulated as one factor leading to this increased vascular resistance. These authors also pointed out a key influence of lung volume on the distribution of blood flow down the lung by showing that the area of the fourth zone was much less at higher lung volumes. A more complex picture was perceived when radiolabeled microspheres were introduced for measurement of blood flow. Reed and Wood (23) used microspheres with a diameter of 35 μm to study pulmonary blood flow distribution in anesthetized dogs in prone, supine, head-up, head-down, right decubitus, and left decubitus positions. They also observed that pulmonary blood flow increased down the hydrostatic pressure gradient in the lung. However, they also found a more than fivefold increase in blood flow from lung base to apex at similar hydrostatic vascular pressures in the supine position. They concluded that factors other than those addressed by the zonal model had to be considered to reach an adequate description of pulmonary perfusion distribution. Since then, a series of studies have corroborated findings of perfusion distribution not easily interpreted within the framework of the four-zone model. For instance, high-resolution imaging of pulmonary perfusion have demonstrated no, or only minute, gravitational flow gradients in the prone position (2, 22, 28), gravity-independent inequality in pulmonary blood flow (4, 5, 17, 20), fairly fixed perfusion distribution with reduced and increased gravitational force (1, 3), and centripetal perfusion gradients (6, 22, 29), all at least in partial conflict with the zonal paradigm.

There is obviously also an abundance of work supporting the classic paradigm. Of principal relevance to the present study is a set of carefully designed studies in which lung volume was altered by increasing airway pressure. Hedenstierna and coworkers (8, 9) analyzed the mechanisms by which positive end-expiratory pressure (PEEP) improves gas exchange. With use of 15-μm microspheres, they demonstrated redistribution of lung blood flow to dependent regions with PEEP (20 cmH2O) ventilation as predicted by the current perfusion pressure-lung volume paradigm. These studies were done in anesthetized dogs in the supine position. More recently, quite similar results were obtained in anesthetized supine dogs and sheep with redistribution of perfusion to dependent lung regions with 5 cmH2O PEEP (14, 27). In prone sheep, the effect of 5 cmH2O PEEP on distribution of blood flow was different with no redistribution to dependent lung (27). The low level of PEEP used in these latter studies of pentobarbital sodium-anesthetized and mechanically ventilated animals increased lung volumes to awake levels only. The purpose of the present work was therefore to challenge the classical paradigm by examining redistribution of pulmonary blood flow with a greater amount of PEEP (10 and 20 cmH2O) in the prone posture. We sought to characterize zonal conditions within the lungs to be able to relate our findings to the classic model.


The experimental protocol was approved by the local Animal Experimentation Committee. Eleven healthy sheep (30–40 kg) were studied. A foreleg vein was cannulated, and anesthesia was induced with pentobarbital sodium and maintained by continuous infusion of 4–7 ml/h of pentobarbital sodium (40 mg/ml) and meperidine (8 mg/ml). The animals were tracheotomized, and catheters were introduced into the carotid artery, the external jugular vein, and the pulmonary artery via surgical cutdowns on the neck. Left atrial pressure was measured with a Brockenbrough needle introduced via a femoral vein transseptally into the left atrium in two animals at the end of the experiment. Heparin (5,000 units) was given to avoid clotting of catheters. Vascular pressures were referenced in both postures to a level one-third of the sagittal thoracic diameter from the sternum, the presumed position of the right atrium. Cardiac output was measured with thermodilution technique. Mechanical ventilation was provided with a Servo 900C ventilator (Siemens Elema, Solna, Sweden) in pressure-controlled mode with tidal volumes appropriate to maintain normocapnia (6–8 ml/kg). Blood gases were measured with an automated blood-gas analyzer (model ABL 520, Radiometer, Copenhagen, Denmark).

Labeling of perfusion.

Radiolabeled microspheres were infused into the right atrium as previously described (13). Briefly, ∼106 microspheres labeled with three different isotopes (diameter 15.5 μm; NEN, Boston, MA) suspended in saline were infused during 8 min at 1 ml/min with glass syringes while mixed continuously with a magnetic bar within the syringe.

Processing of the lungs.

The animals were exsanguinated at the end of the experiment while still deeply anesthetized. The lungs and trachea were excised en bloc and trimmed from adhering nonpulmonary tissue (e.g., lymphatic glands, cardiac muscle). They were then hung by the trachea and expanded with pressurized air (20–25 cmH2O). Numerous holes were made through the lung pleura with a 22-gauge needle to facilitate drying. After drying, the right lung was cut in 1-cm-thick slices, horizontal with respect to gravity when the microspheres were given. The slices were diced into pieces (weight 40–50 mg) while the location of each piece on the map was recorded. The pieces were weighed and placed in labeled plastic tubes, after which radioactivity was counted in a gamma counter (Cobra Autogamma 5002, Packard, Downers Grove, IL).

All animals were used in additional experiments that did not interfere with the present experiment, but they also involved microsphere injections. We used, in all, six different radiolabels (Tc, Ru, Sc, Nb, Sn, Co) and three different fluorescent-labeled microspheres. The radiolabels were corrected for decay and separated by the matrix inversion method (Compusphere, Packard).

Experimental protocol.

We studied six animals in the prone and five animals in the supine posture. Each animal was studied with 0, 10, and 20 cmH2O PEEP in random order. The animals were allowed to stabilize for 30–50 min before the experiments were started, and 20–30 min passed between each experimental permutation. Mean left atrial pressure was measured in two animals in both postures and all three PEEP levels (6 permutations per animals) at the end of the experiment.

Calculations and statistics.

Relative blood flow per piece was determined from radioactivity per milligram dry weight in each piece normalized to radioactivity per milligram dry weight for the whole lung. Height-corrected (isogravitational) relative flow was calculated as radioactivity per milligram in each piece normalized to radioactivity per milligram in the corresponding slice. To assess blood flow distribution in the gravitational field, we determined radioactivity per milligram in each slice normalized to radioactivity per milligram for the whole lung.

The junction between zones I and II was calculated as the lung plane where mean pulmonary vascular driving pressure was equal to mean airway pressure, taking the pressure drop from the zero reference point into account (set equal to the vertical distance). The junction between zones II and III was calculated as the lung plane where mean pulmonary venous pressure (equal to left atrial pressure) was equal to mean airway pressure, accounting for the pressure difference compared with the zero point. Left atrial pressure was measured in all six permutations in two animals (cf. Table 1), and the mean value was used when we calculated the level of the transition between zone II and III in the remaining animals.

View this table:
Table 1.

Lung function and hemodynamic parameters

Total perfusion heterogeneity was assessed per animal and permutation as the coefficient of variation (CVtotal) of relative blood flow in each piece. The isogravitational component (CVisograv) was calculated from the height-corrected relative blood flow values. The component due to different vertical height in the lung was calculated from the formula (CVtotal)2 = (CVisograv)2 + (CVgrav)2, where CVgrav is the gravitational component (18).

We applied linear regression to describe vertical perfusion gradients per animal. The slopes of linear relationships were compared with zero with Student’s t-test. Changes between postures were analyzed with the Kruskal-Wallis rank test, and changes within postures were examined with Friedman’s analysis of variance by ranks. When these tests were significant, pairwise comparisons with a two-sided Mann-Whitney U-test followed. P < 0.05 was deemed significant.


The lungs were sliced into median (range) 19 (16–23) slices and then cut into mean (SD) 1,070 (144) pieces per animal. The mean (SD) piece weight per animal was 48.5 (2.4) mg. Table 1 shows values for blood gases, airway pressure, and systemic circulatory parameters. Blood gases remained within physiological limits during the experiments. Mean systemic arterial pressure, mean pulmonary arterial pressure, and cardiac output were larger in the prone posture with 0 cmH2O PEEP compared with the supine posture with same PEEP level. Cardiac output was not affected in prone by increasing PEEP ventilation, but it was significantly lower with 20 cmH2O PEEP than with 0 and 10 cmH2O PEEP in the supine posture. Left atrial pressure appeared to increase with PEEP in the supine posture, whereas changes were smaller and inconsistent in prone.

The planes of transition between zones I and II, and zones II and III are shown in Table 2. In prone, most of the lung was in zone II, except for a few dependent and nondependent planes at 0 and 10 cmH2O PEEP. In supine, zone II conditions were still predominant but less so with 20 cmH2O PEEP. In both postures, zone III conditions were only present with 0 cmH2O PEEP.

View this table:
Table 2.

Transition levels between zone I, II, and III and perfusion gradients in zone II lung

Mean regional flow down the lung is shown in Fig. 1. In supine, we saw little regional perfusion in nondependent planes compared with the rest of the lung at all PEEP levels (Fig. 1, bottom). Blood flow increased gradually with distance down the lung, except for the most dependent planes. Regional flow actually decreased slightly in the most dependent planes with 0 and 10 cmH2O PEEP. Mean vertical perfusion gradients within zone II lung were uniformly negative, indicating that perfusion increased down the lung (Table 2). The pattern was different in prone in several aspects (Fig. 1, top). First, there was substantial flow in nondependent lung planes. Perfusion in this part changed little with 10 cmH2O PEEP; however, if anything, it increased with 20 cmH2O PEEP. Second, there was no decrease in flow to the most dependent planes. Third, although there appeared to be a small increase in blood flow with distance down the lung at 0 and 10 cmH2O PEEP, the mean vertical perfusion gradients within zone II lung were not significantly different from zero in the prone posture (Table 2).

Fig. 1.

Normalized regional perfusion per unit weight for each horizontal slice (≈1 cm thick) in supine and prone sheep. Each animal (5 supine and 6 prone) was studied at all positive end-expiratory pressure (PEEP) levels. The maximal number of planes was 21 and 23 in supine and prone, respectively. Values are means ± SE. Dotted lines identify limits of zones I–II and II–III (cf. Table 2). Arrows indicate when limits were outside the lungs. Note the remarkable difference in distribution of blood flow between the positions.

Total perfusion heterogeneity increased significantly with PEEP ventilation in the supine position as suggested by the perfusion profiles in Fig. 1 (bottom) (Table 3). When perfusion heterogeneity was partitioned into an isogravitational and a gravitational component, we found that the isogravitational component was similar irrespective of PEEP level within each posture, but it was significantly smaller in prone at all PEEP levels compared with supine. The residual heterogeneity accounting for variation due to different vertical height in the lung (gravitational component) was significantly lower in the prone compared with the supine posture at all PEEP levels. This component increased with PEEP in supine animals only.

View this table:
Table 3.

Perfusion heterogeneity


The novel findings in this study were that redistribution of pulmonary perfusion with 10 and 20 cmH2O PEEP differed remarkably between the prone and the supine postures. Furthermore, isogravitational perfusion heterogeneity was less in prone at all PEEP levels, and blood flow gradients in zone II were absent in prone whereas they were fairly large in supine. These findings are in apparent conflict with the classic zonal model; however, before discussing them in detail, we have to consider some methodological issues.

Microspheres with a diameter of 15 μm lodge in the pulmonary microcirculation in proportion to regional blood flow. Comparison of deposition of microspheres and a molecular tracer (HIPDM) provided evidence that regional microsphere deposition mirrors local blood flow within the lung (17). We analyzed perfusion distribution in the right lung only, because prior work demonstrated that distribution of blood flow was very similar in the right and left lungs (17). The lungs were inflated postmortem and dried at total lung capacity to expand alveoli uniformly throughout the lung. Weight per plane was similar between postures, indicating that prone and supine lungs were identically expanded ex situ (data not shown). An important question is to what extent the distribution of blood flow in these ex situ inflated and dried lungs represents blood flow in situ. Our findings in the prone animals should closely represent regional perfusion in situ because lung expansion is uniform in the prone posture (10). In the supine animals, it is likely that expansion ex situ changed linear gradients that were present in situ, because Hoffman (10) showed that dependent lung tissue was less inflated in situ in supine dogs [regional air content increased with 3.29%/cm up the lung at functional residual capacity (FRC)]. Flow in the most dependent plane in situ thus represents flow to more than one plane ex situ. If distribution of blood flow were to be given per unit volume, the ex situ distributions could differ markedly from that in situ. If, however, the distribution of blood flow is given per unit weight or per alveolus (alveoli probably have about equal weight), the ex situ distribution should mirror the in situ distribution reasonably close, except for some prolongation of any gradient in the most dependent supine lung. Hoffman also showed that expansion became more uniform with increasing lung volumes. Thus prolongation would be less when measurements were averaged over the entire respiratory cycle as in the present experiments. Given the dependence on lung volume of this effect, it is reasonable to believe that the distortion was even less significant in supine at 10 and 20 cmH2O PEEP. On the basis of these arguments, we conclude that the distributions we have recorded along the vertical axis most probably represents the in vivo situation and that prolongation of perfusion profiles and linear gradients in zone II lung was small and only present in supine lungs at 0 cmH2O PEEP.

Absolute regional blood flow was not measured because the focus was to examine gradients in pulmonary blood flow and relative flow distribution within the lung. Flows in each piece were thus normalized to the mean flow of all pieces per animal and condition. Tissue pieces contained, apart from gas-exchanging parenchyma, airways, which are heavier than alveolar tissue. Prior work in sheep showed that inclusion of pieces with more than 25% airways (assessed visually) had no influence on perfusion gradients (28). Hence, all lung pieces were included into the present analysis. By calculating vascular driving pressures per vertical plane, it was possible to relate regional perfusion patterns to zone I, II, and III conditions. Whereas the transition between zones I and II was calculated with confidence, the plane of transition from zone II to III was somewhat uncertain because we measured left atrial pressure in two animals only. However, given the high mean airway pressures at 10 and 20 cmH2O PEEP, we can still be confident that the larger part of each lung was in zone II, unless left atrial pressures rose excessively which our measurements, albeit few, did not indicate.

The cyclic changes in vascular and alveolar pressures with heartbeat and ventilation will make the levels of transition between zones less distinct. Such variations, which are basically similar in magnitude between postures, cannot explain the differences in distribution of perfusion between postures. Lung volumes are reduced by anesthesia and mechanical ventilation in the supine as well as in the prone posture (7, 24). Studies in supine dogs showed a 25–30% increase in FRC with 5 cmH2O PEEP (14), and it is reasonable to assume that lung volumes increase further with larger amounts of PEEP. Studies on the influence on lung volumes of PEEP in the prone posture are singular. Martynowicz et al. (16) found in oleic acid injured dogs that lobar lung volumes increased with 7.5 and 15 cmH2O PEEP (16). We believe that it is reasonable to assume that lung volumes also increase with PEEP ventilation in noninjured lungs in the prone posture.

Redistribution of lung blood flow with PEEP ventilation in the supine position was mainly consistent with the zonal concept and prior studies. Hedenstierna et al. (8, 9) found with 20 cmH2O PEEP ventilation that blood flow was redistributed from nondependent to dependent regions, the vertical gradient was augmented, and perfusion heterogeneity (calculated using inert gas data) increased, all in agreement with our findings. They also noted that some blood flow remained at a constant low value up to the top of the lung (zone I) and attributed this to persistent flow through corner vessels in the junctions between alveolar septa, a possibility suggested from work by Rosenzweig et al. (26). Flow through these corner vessels may account for up to 15% of the normal resting cardiac output in lungs that are completely in zone I (15). There appeared in the present study to be less flow in most dependent supine lung except with 20 cmH2O PEEP ventilation, much in agreement with earlier studies (9, 12). However, this must not be interpreted as a sign of zone IV conditions because our data indicate that these dependent regions were still in zone II or in an initial zone III (cf. Fig. 1). The similar vertical gradient with 0 and 10 cmH2O PEEP ventilation (cf. Table 2) is in accord with zone II conditions. The augmented gradient with 20 cmH2O PEEP must then be interpreted as an effect due to some other factor than an increase in hydrostatic pressure. Centrifugal displacement of perfusion with 20 cmH2O PEEP due to less vascular resistance in the periphery of the lung than in the core, as suggested by Hedenstierna et al. (9), may be one such factor. The effect of PEEP on distribution of blood flow in the prone posture was very different from that in supine. There appeared to be no abrupt change in perfusion that could correspond to a zonal transition. This may be explained by the fact that most of the lung was within zone II. However, given the presence of zone II conditions, one would expect to find a significant gravitational gradient due to increased hydrostatic pressure down the lung. We saw no such gradient.

The absence of a gravitational gradient in prone zone II lung is yet another observation in a series since the study by Reed and Wood (23) that questions the zonal paradigm. Such observations are not unique to animals but have also been observed in humans (5, 21). Of particular relevance is a study of perfusion distribution in spontaneous breathing humans. Blood flow distribution was measured in three transverse lung slices by single-photon-emission computerized tomography after intravenous injection of macroaggregated radiolabeled albumin, and the measurements were clustered into three groups per slice (dependent, midlung, and nondependent) (21). Changes with posture and 10 cmH2O continuous positive airway pressure (CPAP) were examined. In the supine position, perfusion became less uniform with CPAP, in agreement with prior (25, 27) and present work. However, also in the prone posture, perfusion became less uniform with CPAP, in contrast to the present study and prior work (25, 27). Given that mechanisms that control perfusion distribution are similar across species, and that anesthesia and mechanical ventilation have but minor influence on perfusion distribution in the prone posture (28), the difference may be due to limited sampling of lung in the human study.

It seems difficult to understand the pattern of blood flow distribution with PEEP within the framework of the zonal concept. The striking absence in prone of a gravitational gradient in zone II lung indicates that other factors than the interplay between alveolar, arterial, and venous pressures have to be considered when we try to understand the mechanisms that influence the distribution of blood flow in the lung. Regional vascular resistance is strongly influenced by regional lung volume (11). The marked discrepancies between postures in the response to 10 and 20 cmH2O PEEP suggest that regional lung volume is more influenced by chest wall conformation together with extrapulmonary factors, i.e., heart and position of the diaphragm, than currently believed. Increase in lung volume may, through radial traction on blood vessels (11), be accompanied by an increase in vascular volume and decrease in resistance to flow. We speculate that this effect may differ between postures. Another possibility is that, through pulmonary interdependence, a variable and posture-dependent siphon effect may exist on the venous side that contributes to less perfusion heterogeneity in the prone posture. Both these effects could make estimations of driving pressure in the pulmonary vascular bed imprecise. The heterogeneity of blood flow within planes (isogravitational heterogeneity) was systematically lower in the prone posture at all levels of airway pressure. Because blood flow is given per unit weight, this difference cannot be explained as being the result of different degree of distension. Interestingly, Musch et al. (19) found in humans indication of lower isogravitational heterogeneity in the prone posture. We have no easy explanation for this difference between postures.

We conclude that the present study forces us to reconsider the underlying mechanisms of pulmonary blood flow distribution. It appears difficult to explain, within the zonal paradigm, the perfusion pattern that was seen in zone II lung. The large differences between postures in the effect of PEEP on the distribution of blood flow and differences in isogravitational heterogeneity are also difficult to explain. Future studies have to clarify other factors that may influence perfusion distribution in the healthy lung. We speculate that distortion of the lung within the chest is one such factor.


This work was supported by The AGA Gas AB, The Swedish Heart-Lung Foundation, and The Anders Jahres Foundation for the Promotion of Sciences.


The skilled help of P. Björnstad was instrumental for the transseptal measurements of left atrial pressures.


  • 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.


View Abstract