Introduction

Prone positioning has been used for more than 30 years in patients with acute hypoxemic respiratory failure and in particular with acute respiratory distress syndrome (ARDS). The initial reason for prone positioning in ARDS patients was to alleviate severe hypoxemia and it turned out that prone positioning was an efficient means to improve oxygenation, sometimes dramatically, in a large number of patients. Therefore, prone positioning has been thought of, from the very outset, as a rescue treatment in case of life-threatening hypoxemia. However, it was later recognized that the prevention of ventilator-induced lung injury (VILI) [1] is as important a goal as maintaining safe gas exchange in mechanical ventilation of ARDS patients. It is now clear, and data are still accumulating, that prone positioning is able to prevent VILI. Therefore, prone positioning is a strategy that covers the two major goals of ventilator support in ARDS patients, maintaining safe oxygenation and preventing VILI. This review will cover the mechanisms by which prone positioning improves oxygenation, the effects of prone positioning on respiratory mechanics, the effects of prone positioning on hemodynamics, and the mechanisms by which prone positioning prevents VILI. All together, these mechanisms support a large body of physiologic benefits that should explain why prone positioning improves survival in severe ARDS patients as demonstrated in a recent trial [2] in line with a previous meta-analysis [3].

Effects of prone positioning on oxygenation

Improvement in oxygenation with prone positioning comes from a reduction in intrapulmonary shunt (Q s/Q t) and change in lung ventilation (V A) and lung perfusion (Q) distribution and better V A/Q matching. Three preliminary considerations underlie the effect of prone positioning in ARDS on oxygenation. First, for the Q s/Q t to go down either Q should be directed towards well-ventilated lung areas or V A should be redirected towards well-perfused lung areas. In both instances overall matching of V A/Q should improve and so should the oxygenation. Second, owing to the effect of gravity Q s/Q t increases along the ventral-to-dorsal gradient and is maximal in the dorsal lung regions in the supine position. Third, by “reversing” the effect of gravity, the prone position should decrease Q s/Q t in the dorsal lung regions, now non-dependent. Let us first examine whether prone positioning would reduce Q s/Q t from the redistribution of perfusion towards well-ventilated areas away from those edematous in ARDS. Figure 1 (bottom) displays the four zones of lung perfusion described by West in normal human lung in the sitting position. The figure was adapted for the supine position, even though the gravitational gradient should be less than for the sitting one, as the ventral-to-dorsal dimension is lower than its cranial-to-caudal counterpart. For the blood to flow through the alveoli the pulmonary artery pressure upstream must be higher than the pulmonary vein pressure downstream. According to the model in Fig. 1 (bottom), the amplitude of the upstream–downstream hydrostatic force into the pulmonary circulation, and hence the magnitude of pulmonary blood flow, depends on the alveolar pressure as the amount of blood flow upstream of the alveoli is assumed to be similar. In normal subjects, gravity imposes a progressive reduction of alveolar pressure (P A) from the top to the bottom and from the non-dependent to the dependent parts of the lungs. What is expected from the reversal of gravity is that lung perfusion should move towards ventral regions in the prone position. If this holds true this should furthermore stem from the reversal of P A along the same gradient. Wiener et al. [4] measured lung perfusion by using radiolabeled microspheres, the gold standard technique, in normal dogs before and after oleic acid injection in supine and prone positions. Pulmonary blood flow remained preferentially distributed to the dorsal non-dependent regions when they were prone, contrary to the above expectations. The magnitude of the gravitational gradient was, however, reduced in the prone position. Furthermore, in this study, the distribution of lung edema was not influenced by gravity. In pigs injured with oleic acid we found similar findings [5] (Fig. 1) by measuring pulmonary blood flow with positron emission tomography, a method we had previously validated using radiolabeled microspheres [6]: pulmonary blood flow was still prevalent to the dorsal regions and more homogeneously arranged through the lungs. Moreover, we found homogenization in the distribution of lung densities (not shown) across the lung in the prone position. These findings were confirmed in other animal species including humans.

Fig. 1
figure 1

Distribution of pulmonary blood flow (PBF) along the ventral-to-dorsal gradient in the supine (SP) (open triangles) then after 120 min (blue symbols) and 160 min (red symbols) in the prone (PP) position, after injection of oleic acid in five pigs. PBF was measured by using 15OH2 and positron emission tomography (PET). Superimposed on the PBF distribution are the four theoretical zones described by West. PA alveolar pressure, Ppa pulmonary artery blood pressure, Ppv pulmonary vein blood pressure. For further explanations, see text

Full size image

To sum up the above considerations, the reduction in Q s/Q t in the prone position cannot be explained by more perfusion to well-ventilated areas or less perfusion to poorly ventilated areas.

Let us now look at the second hypothesis for why the Q s/Q t goes down in the prone position, namely the increase in V A towards well-perfused areas. In ARDS the distribution of P A in Fig. 1 is further complicated by the loss of aeration due to lung edema, inflammation, and atelectasis, and also by the ventilator settings. More than 40 years ago, Milic-Emili et al. [7–9] described the gravitational distribution of lung ventilation in normal humans in various body positions. They estimated pleural pressure from esophageal pressure measurements [10] and assessed lung ventilation with xenon and lung planar scintigraphy [11]. They found that pleural pressure was more negative, and hence transpulmonary pressure was more positive, in non-dependent regions. Mutoh et al. [12] directly measured pleural pressure in normal dogs and after hypervolemia. They found that in the lung injury condition (Fig. 2) turning from the supine to the prone position makes the pleural pressure in the non-dependent regions less negative and less positive in the dependent regions. The net and important effect is a marked reduction in the pleural pressure gradient, which has also been shown in humans [13] resulting in homogenization of pleural pressure and hence of transpulmonary pressure across the ventral-to-dorsal direction.

Fig. 2
figure 2

Distribution of pleural pressure in non-dependent (ND) and dependent (D) lung regions in supine (SP) and prone (PP) positions in normal dogs and after lung injury induced by hypervolemia. For further explanations, see text

Full size image

As a result the V A/Q matching should improve in the dorsal regions in the prone position with no change in the ventral regions. This has been demonstrated in dogs injured with oleic acid by using SPECT [14] and simultaneous ventilation and perfusion measurements with radioactive compounds. Richter et al. [15] used intravenous injection of nitrogen-13 and PET to estimate the amount of regional Q s/Q t in surfactant-depleted sheep. They found that Q s/Q t was significantly reduced in the dorsal parts of the lung owing to the combined effect of increase in aeration and ventilation and maintenance of perfusion in those areas.

To sum up, in the prone position the Q s/Q t goes down following the increase in V A towards well-perfused areas.

In ARDS patients receiving mechanical ventilation, oxygenation also depends on ventilator settings, in particular PEEP, and on their effects on the determinants of oxygenation (V A/Q distribution and Q s/Q t). To further explore the effect of the interaction of body position and PEEP on these variables we measured lung ventilation, lung perfusion, aeration, and recruitment with PET at a PEEP of 0 and 10 cmH2O in pigs injured with oleic acid [16]. PEEP was associated with significant alveolar recruitment in each position, whereas recruitment induced by posture was not statistically significantly different from 0 at each PEEP. The prone position was associated with recruitment in dorsal regions with concomitant derecruitment in ventral regions, the magnitude of this being reduced by PEEP. The prone position redistributed ventilation toward dorsal regions at PEEP 0 and to ventral regions at PEEP 10, and perfusion toward ventral regions at PEEP 10. There were therefore two distinct scenarios for the improvement in oxygenation with prone positioning depending on PEEP level. At PEEP 0, the improvement in oxygenation in the prone position follows the common view of lower Q s/Q t in dorsal regions. At PEEP 10, the improvement in oxygenation resulted from a better V A/Q matching throughout the lungs owing to the homogenization of lung ventilation and perfusion distribution. As ARDS patients commonly receive PEEP of at least 5 cmH2O, the second scenario should be prevalent in most clinical situations. In this study, we also assessed the determinants of PaO2 due to PEEP in any given position and due to position at any single level of PEEP (Table 1, unpublished results). As an example, between PEEP 0 and 10 cmH2O in the supine position, the change in PaO2 negatively correlated with the lung perfusion to the ventral regions (the lower the lung perfusion to the ventral regions, the higher the PaO2 at PEEP 10 cmH2O) and positively correlated with both the lung perfusion to and lung recruitment in the dorsal lung regions (the higher the perfusion and the greater the recruitment to the dorsal regions, the higher the PaO2 at PEEP 10 cmH2O).

Table 1 Correlation between regional redistribution of PET measurements and PaO2 variation induced by any tested intervention
Full size table

In summary, in ARDS the increase in oxygenation in the prone position comes from the reduction of Q s/Q t that is mainly due to better ventilation in perfused lung areas.

Effects of prone positioning on respiratory mechanics and lung volume

The fitting of the lungs into the thorax is a main determinant of the effects of prone positioning on gas exchange and VILI in ARDS patients. The mechanics of the chest wall involved in the behavior of the lungs in the rib cage and abdomen include increase in chest wall elastance (Est,cw), abdominal wall elastance, diaphragm curvature, and heart and mediastinal mass. The respiratory system is commonly modeled as two elastic bodies, the lungs and the chest wall (rib cage and abdominal wall), arranged serially. Accordingly, their respective elastance, lung elastance (Est,L) and Est,cw, add up to the respiratory system elastance (Est,rs). It should be noted that the elastic nature of the abdominal wall has been questioned and the mechanical role of abdomen ascribed to the displacement of its mass rather to elastic properties [17]. Let us go briefly over the effect of prone positioning on each component of the respiratory mechanics in ARDS.

In adult humans with ARDS, Est,cw has consistently been found to be higher in the prone than in the supine position (Table 2). This can be the result of the increase in abdominal pressure and/or cranial diaphragm displacement in patients with baseline high intra-abdominal pressure. A stiffer (anterior) chest wall in prone position implies that the lungs are operating in-between two rigid bars, the spine and the sternum. This arrangement would make the distribution of tidal volume more homogeneous and gas exchange improved. Indeed, Pelosi et al. [18] found that the higher the chest wall compliance in supine the better the oxygenation in prone and that the magnitude of change in oxygenation negatively correlated with the increase in chest wall elastance with prone positioning.

Table 2 Values of static chest wall elastance and end-expiratory lung volume in patients with ARDS in the supine and in the prone position
Full size table

Assuming a systematic increase in Est,cw in the prone position (Table 2) and given that Est,rs = Est,L + Est,cw, the range of values of Est,rs found in the prone position can be interpreted as follows. No change in Est,rs implies a reduction in Est,L proportional to the increase in Est,cw. Higher Est,rs can be due to an increase in Est,cw with or without an increase in Est,L. Lower Est,rs suggests marked reduction in Est,L, and presumably significant recruitment, in the prone position. In the randomized controlled trials comparing supine and prone positions, the results on Est,rs are not consistent. Mancebo et al. [19] found higher compliance of the respiratory system (lower Est,rs) in the prone position, whereas Guérin et al. [2]. and Taccone et al. [20] did not. In these trials the prone position was applied for long periods and measurements were done at the end of each of them.

In seven patients without ARDS and breathing spontaneously, CT scan analysis found that prone positioning eliminates the compressive force of the heart weight as almost no lung is located under the heart [21]. The compressive force of the heart is then directed towards the sternum [21]. There is no such data available in ARDS patients. This finding may explain why Nakos et al. [22] found that oxygenation response to prone positioning was observed in 100 % of patients with cardiogenic pulmonary edema, in whom cardiac enlargement is likely. Since pulmonary artery occlusion pressure has been found to be higher than 18 mmHg in almost one-quarter of ARDS patients [23] and since, on this basis, it is no longer mandated to rule out left cardiac failure or hypervolemia to define ARDS [24], unless no risk factor for ARDS has been identified, this mechanism might be of paramount importance and warrants further studies.

The chest wall elastance contributes to set the end-expiratory lung volume (EELV), which is the functional residual capacity on PEEP 0. The values of EELV in supine and prone position in ARDS patients are limited to date and not fully consistent (Table 2). Pelosi et al. [25] found that sighs superimposed on the prone position further increased EELV and oxygenation with no effect when applied in the supine position. This would suggest that either the potential for recruitment was not maximal and the prone position extended it, or overdistension may have occurred. As discussed below CT studies are consistent in discarding overdistension occurrence with prone positioning.

Recruitment of non-aerated or poorly aerated lung is an important strategy to improve oxygenation and make the lung more homogeneous, and hence to prevent VILI. Lung CT scan studies provided consistent evidence that the prone position promotes lung recruitment as compared to the supine position. Galiatsou et al. [26] found that this process was more prevalent in lobar than in diffuse ARDS patients. Cornejo et al. [27] extended these findings in showing that prone position-induced lung recruitment was observed regardless of whether the patient was a high or a low recruiter in response to the change in airway pressure in the supine position or whether PEEP was low or high in the supine position [28]. There is no evidence to date that lung recruitment is beneficial to patient outcome according to the negative results of the three large randomized controlled trials on high versus low PEEP [29–31]. It should be noted that in these trials the potential for recruitment was not taken into account and low recruiters may have received high PEEP and developed overdistension with its harmful consequences. Another issue to have in mind when dealing with lung recruitment is that the recruited lung tissue, even re-aerated, may still have impaired lung mechanics, e.g., higher regional elastance, as compared to normal lung [32]. This result seems unlikely in the prone position, which works without an increase in airway opening pressure relative to the supine position.

Translating these findings into the strain and stress concept [33] shows that prone positioning homogenizes the strain imposed by mechanical ventilation and overall reduces the resulting stress.

The main effects of prone positioning on respiratory mechanics and lung volume are summarized in Supplementary material.

Effects of prone positioning on hemodynamics

As a whole, prone positioning is associated with a preservation of hemodynamics in ARDS patients. Beneficial hemodynamic effects of prone positioning observed in ARDS patients are reported Table 3. One of the most relevant effects may be the reduction of the transpulmonary gradient (the difference of mean pulmonary arterial pressure relative to pulmonary artery occlusion pressure), as vascular dysfunction is a major independent factor associated with ARDS mortality [34]. Increase in pulmonary arterial occlusion pressure induced by prone positioning is another striking feature, which may result in pulmonary vascular recruitment by increasing pulmonary venous pressure above alveolar pressure in some areas of the lungs. This would decrease the dead space, another factor independently related to ARDS mortality [35]. However, the substantial heterogeneity observed in clinical studies (Table 3) suggests that several confounding factors may interfere with the effect of prone positioning on hemodynamics (e.g., case mix, volume status before prone positioning, and PEEP level). In this regard, prone positioning-associated improvement in oxygenation may offset the deleterious effect of high PEEP on hemodynamics, provided PEEP adjustment is made following oxygenation improvement with prone positioning, as can be done by using the PEEP-FiO2 table [36]. This strategy has been used in the PROSEVA trial [2], and may have contributed to the significant increase in cardiovascular dysfunction-free days observed in the prone positioning group, along with the beneficial hemodynamic effects directly related to prone positioning. It is also noteworthy that long prone positioning sessions can reverse right heart failure in severe ARDS patients, a finding that could play a role in patient outcome [37].

Table 3 Effects of prone positioning on hemodynamics in ARDS patients
Full size table

However, a few adverse effects of PP on hemodynamics have been reported (Table 3). An increase in left ventricle afterload related to a slight increase in mean arterial pressure may adversely affect left heart function, although the magnitude of this increase (ca. 5 mmHg) may be of little clinical relevance. Splanchnic perfusion may also worsen in response to intra-abdominal pressure increase due to abdominal compression in prone positioning, although this effect is not consistently reported and is not associated with detectable renal dysfunction (Table 3).

Effects of prone positioning on VILI

Once an ARDS patient is under invasive mechanical ventilation the intensivist’s primary concern should be to protect the lungs. This can be first achieved by lowering tidal volume [36]. Lowering tidal volume prevents VILI and improves survival. Data indicate that prone positioning prevents VILI. Broccard et al. [38] ventilated normal dogs for 6 h with 77 ml/kg tidal volume to reach 35 cmH2O transpulmonary end-inspiratory plateau pressure in supine or prone position. The results were striking with a marked reduction in the lung injury both macroscopically and microscopically and a homogeneous redistribution of VILI. Valenza et al. [39] found that the time for getting Est,rs higher than 50 % from the baseline in normal rats submitted to high tidal volume was significantly increased in the prone as compared to the supine position. CT scan studies found a reduction in overall overdistension in the prone position [26, 27]. However, the reduction in atelectrauma (repeated opening and closing of terminal respiratory units) and tidal hyperinflation measured statically during the breathing cycle was only observed with prone positioning in high recruiters in the supine position, who were receiving higher PEEP (15 cmH2O) in the prone position. Papazian et al. [40] found lower concentrations of pro-inflammatory cytokines in the bronchoalveolar lavage fluid in ARDS patients after 12 h the in prone position as compared to the same settings applied in the supine position for the same duration. Finally, it has recently been shown that prone positioning can work at the molecular level of VILI in modulating the expression and activation of mitogen protein kinase in rodents submitted to injurious mechanical ventilation [41]. So, there are several levels of evidence supporting the preventive effect of the prone position on VILI, making of prone positioning a full component of lung protective mechanical ventilation.

Long-term effects of prone positioning

Most of the physiologic data about the prone position are taken from early measurements of arterial blood gas or respiratory mechanics, i.e., 1–2 h after prone installation. Therefore, there is a need for further longitudinal physiologic information over the prone positioning session. However, Albert et al. [42] investigated the effects of prone positioning on oxygenation and survival, and found no significant result whether early or late PaO2 was used during the prone positioning session. Patients who did not respond to prone positioning in terms of oxygenation should be maintained prone unless gas exchange worsened during the session. Indeed, should the prone position work by preventing VILI this has nothing to do with the effect on gas exchange. As an example, in the ARMA trial [36], patients in the lower tidal volume group had better survival but worsened oxygenation during the first week as compared to the control group. In line with this argument, the prone position should be maintained for longer periods such as a continuous 24 h or more. This approach was planned by Mancebo et al. in their trial [19] but was not feasible for practical reasons. It should be noted that derecruitment may occur over time in the ventral lung regions in the prone position but, as discussed previously, the data supporting this are lacking. To sum up this section, the prone position should be used for long sessions, even though there are no physiologic data on prone positioning over time, because clinical benefits were obtained with such long sessions and should VILI prevention be the main factor in those benefits the longer the prone positioning the less the VILI.

How does prone positioning reduce mortality in ARDS?

There is a strong pathophysiological rationale for the beneficial physiologic effects of prone positioning to translate into clinical benefits in ARDS. This has been demonstrated in a recent single randomized controlled trial [43], which has confirmed a previous meta-analysis [3]. Nevertheless, it is not so clear whether these clinical benefits stem from VILI reduction/prevention as VILI markers were not assessed in that trial. The reduction in mortality was not explained by improvement in oxygenation [42]. Another mechanism that may be involved in the increase in survival is the improvement/preservation of hemodynamics in the prone position, as discussed above. The right ventricle is unloaded in the prone position with the reversal of acute core pulmonale as evidenced by transesophageal echocardiography [44]. The PROSEVA trial [43] also found that there were fewer days free of cardiovascular dysfunction in the prone group than in the supine group.

It should also be stressed that the observed effect of prone positioning on mortality in the PROSEVA trial may result from some of the inclusion criteria used, in particular both the PaO2/FIO2 cutoff and the stabilization period. A cutoff PaO2/FiO2 of 150 mmHg has been found to predict outcome within the first 24 h of ARDS onset in the ACURASYS trial [45] in which the intervention tested, i.e., neuromuscular blocking agent, was beneficial. These enrollment criteria fit with the universal definition of ARDS that characterizes ARDS patients according to early PEEP response.

Conclusions

With prone positioning we have an effective weapon that works at every step of lung protection and cardiocirculatory function preservation from modulation of pathways involved in VILI to the level of integrated cardiorespiratory physiology in mechanically ventilated ARDS patients. The mechanisms by which prone positioning improves survival are likely related to these physiologic effects. Therefore, prone positioning should be applied systematically as a first-line therapy in patients with severe ARDS.