Introduction

Since its first description in 1967 [1], it has been accepted that acute respiratory distress syndrome (ARDS) includes a number of lung injuries of various origins whose consequences are decreased lung capacity available for ventilation, leading to the concept of "baby lung" [2]. Considerable progress has been made over the past decade in the ventilatory management of patients with ARDS. In particular, a strict limitation of tidal volume (VT) and plateau pressure (Pplat) below 30 cmH2O reduces mortality [3]. The application of positive end-expiratory pressure (PEEP) is recognized to recruit the lung and to restore functional residual capacity [4], but its optimum level is still widely debated [5].

The prone position (PP) may also be part of the ventilatory strategy. This method was proposed more than 30 years ago, initially in pathophysiological studies [6, 7]. Recently, Sud et al. [8] suggested, on the basis of pooled data from randomized, controlled trials, that PP may improve survival in the subgroup of patients with the most severe ARDS, that is, those with a ratio of partial pressure of arterial O2 to fraction of inspired O2 (PaO2/FiO2) < 100 mmHg. Many questions remain unresolved. In particular, response to PP is usually defined according to changes in PaO2, with responders being those in whom the PaO2/FiO2 ratio increases > 20 mmHg after one to six hours in the PP [9–11]. However, we have previously reported that PP allows recruitment of a slow compartment previously excluded from ventilation [12]. This was associated with a decrease in partial pressure of arterial CO2 (PaCO2), an indirect reflection of the reduction of the alveolar dead space (VDalv) [12]. Gattinoni et al. [10] also reported that the prognosis is improved in patients in whom PaCO2 declines after an initial PP session. Finally, VDalv appears to be an independent risk factor for mortality in patients with ARDS [13]. In a recent study, Siddiki et al. [14] proposed evaluating the physiological dead space fraction (VDphysiol/VT) by using a rearranged alveolar gas equation for PaCO2 without any expired CO2 measurement.

In this context, we conducted a prospective physiological study to evaluate the impact of PP on ventilatory mechanics, gas exchange and VDalv. Our main objective was to validate our hypothesis that changes in PaCO2 and VDalv might be more relevant than changes in PaO2 in defining the respiratory response to PP. Our second objective was to validate the method of evaluation of the VDphysiol/VT proposed by Siddiki et al. [14].

Materials and methods

In our unit, patients with a PaO2/FiO2 ratio < 100 mmHg after 24 to 48 hours of mechanical ventilation are systematically turned to PP when hemodynamically stable [15]. Our study was approved by the Ethics Committee of the "Société de Réanimation de Langue Française" (SRLF-CE 07-213). After obtaining informed consent from the patients' relatives, 15 patients were included in the study between January 2008 and March 2010. Inclusion criteria were (1) the presence of ARDS according to the definition of the Acute Respiratory Distress Syndrome Network [3]; (2) persistence of severe hypoxemia after 48 hours of mechanical ventilation, defined as a PaO2/FiO2 ratio < 100 mmHg; and (3) hemodynamic stability, defined as systolic blood pressure > 90 mmHg with norepinephrine infusion at a rate < 0.5 μg/kg/minute. Patients with chronic obstructive pulmonary disease were excluded.

All patients were ventilated in volume-controlled mode (Servo-i; Maquet SA, Ardon, France), sedated and paralyzed by infusion of atracurium. The heat and moisture exchanger was routinely removed and replaced by a heated humidifier to reduce instrumental dead space as previously reported [16]. The ventilator settings included a "moderately restricted" VT of 6 to 8 mL/kg measured body weight, a respiratory rate allowing us to limit hypercapnia without generating intrinsic PEEP and an inspiration/expiration ratio of 1:2 with an end inspiratory pause of 0.5 seconds. Pplat was strictly limited < 30 cmH2O, and the PEEP selected was that which corrected the intrinsic PEEP, if any [17]. Ventilator settings were kept constant throughout the study. A recruitment maneuver was never used, and suction was not systematically performed. All patients were continuously monitored in terms of blood pressure with an arterial catheter, heart rate and O2 saturation by pulse oximetry.

The study was conducted during the first session of PP. Our sessions routinely last 15 to 18 hours per day. Blood gas analysis, Pplat, total PEEP, end-tidal CO2 (PetCO2) and mixed expired CO2 (PECO2) were recorded with the patient in the supine position, just before turning the patient to the PP, and every 3 hours in the PP until 15 hours had elapsed. Expired CO2 was measured by a sensor positioned between the proximal end of the endotracheal tube and the Y piece of the ventilator circuit (COSMO; Novametrix, Wallingford, CT, USA). The ratio of VD/VT was calculated using the simplified Bohr equation [18] as follows: (1) VDalv/VT = 1 - PetCO2/PaCO2 and (2) VDphysiol/VT = 1 - PECO2/PaCO2.

The estimated VDphysiol/VT ratio was calculated as 1 - [(0.86 × VCO2est)/(VE × PaCO2)], where VCO2est is the estimated CO2 production calculated using the Harris-Benedict equation [19] and VE is the expired minute ventilation.

Intrinsic PEEP was measured during a four-second end-expiratory occlusion period. Pplat was measured during a 0.5-second end-inspiratory pause. Respiratory system compliance (Crs) was calculated as Crs = VT/(Pplat - PEEPtotal). Responders to PP were defined in two different ways: (1) an increase in PaO2/FiO2 ratio > 20 mmHg after 15 hours of PP or (2) a decrease in PaCO2 > 2 mmHg after 15 hours of PP.

Statistical analysis

Statistical analysis was performed using StatView 5 software (SAS Institute Inc., Cary, NC, USA). The continuous variables were expressed as medians (1st to 3rd interquartile range). Analysis of variance for repeated measurements was used for each parameter, and P < 0.05 was considered statistically significant. Measured VDphysiol/VT and estimated VDphysiol/VT were compared according to Bland-Altman analysis, together with the concordance correlation coefficient in 78 paired data. The same method was used to compare variations of measured and estimated VDphysiol/VT every three hours while the patient was in PP.

Results

Two patients were excluded from the study because of a history of severe chronic obstructive pulmonary disease, which left a study population of 13 patients. The patients' median age was 53 years (1st to 3rd interquartile range, 48 to 59 years), their median Simplified Acute Physiology Score II score was 62 (1st to 3rd interquartile range, 35 to 71) and their median Sequential Organ Failure Assessment score was 11 (1st to 3rd interquartile range, 8-13). All patients except one had ARDS of pulmonary origin. Eight patients had pneumonia, with six cases related to streptococcus pneumonia and two due to influenza (H1N1 virus). Two patients had aspiration, one had toxic shock syndrome and two had ARDS due to miscellaneous causes. No patient had abdominal hypertension or traumatic lung injury. Eleven patients required norepinephrine infusion. Respiratory parameters and blood gas analysis at the time of inclusion are reported in Table 1.

Table 1 Respiratory parameters and blood gas analysis at inclusiona
Full size table

A significant increase in PaO2/FiO2 ratio occurred after 15 hours of PP, from 70 mmHg (51 to 77) in the supine position to 99 mmHg in the prone (83 to 139) (P < 0.0001) (Table 2). A significant decrease in PaCO2 was also observed, from 58 mmHg (52 to 60) to 52 mmHg (47 to 56) (P = 0.04) (Table 2), with the lowest value occurring after nine hours of PP. As noted in Table 2, Pplat was significantly reduced (P = 0.0004) and Crs improved (from 16 mL/cmH2O (13 to 30) to 18 mL/cmH2O (15 to 30); P = 0.02). Finally, the VDalv/VT ratio was significantly reduced from 0.42 (0.35 to 0.47) to 0.40 (0.26 to 0.45), with the lowest value occurring after three hours in PP (hour 3) (0.31) (Table 2).

Table 2 Changes in respiratory mechanics, blood gas analysis and VDalv in PP
Full size table

Seven patients were classified as "PaO2 responders" and six were classified as "PaO2 nonresponders" according to PaO2/FiO2 ratio changes. No differences in VDalv/VT ratios or PaCO2 or Pplat alterations during PP were observed between groups (Table 3 and Figure 1), whereas Crs increased more in the responders (Table 3). Seven patients were also classified as "PaCO2 responders" and six as "PaCO2 nonresponders" according to the PaCO2 changes. However, when compared with the PaO2/FiO2 classification, four patients were classified differently. As shown in Table 4 and Figure 2, VDalv/VT, PaO2/FiO2, PaCO2, Pplat and Crs were significantly more altered in responders than in nonresponders. As shown in Figure 3, we found no correlation between changes in VDalv/VT and changes in PaO2/FiO2 (P = 0.95), whereas we found a negative correlation between changes in VDalv/VT and changes in Crs (r = 0.29, P = 0.03).

Table 3 Changes in respiratory mechanics, blood gas analysis and VDalv in PaO2 responders (n = 7) and PaO2 nonresponders (n = 6)a
Full size table
Figure 1
figure 1

Alterations during PP in PaO 2 /FiO 2 , PaCO 2 , plateau pressure (Pplat) and alveolar dead space (VD alv /V T ) in responders (solid lines) and nonresponders (dotted lines) according to PaO 2 /FiO 2 changes. "PaO2 responders" were defined by an increase in PaO2/FiO2 > 20 mmHg after 15 hours of PP (PP H15). Shown are box and whisker plots. Median = horizontal line inside the box; upper and lower quartiles = whisker plot. Boxes and triangles represent values higher or lower than the upper or lower quartiles. *P < 0.05 for comparison of changes in responders versus nonresponders. PP: prone position.

Full size image
Table 4 Changes in respiratory mechanics, blood gas analysis and VDalv in PaCO2 responders (n = 7) and PaCO2 nonresponders (n = 6)a
Full size table
Figure 2
figure 2

Alterations during PP in PaO 2 /FiO 2 , PaCO 2 , plateau pressure (Pplat) and alveolar dead space (VD alv /V T ) in responders (solid lines) and nonresponders (dotted lines) according to PaCO 2 changes. "PaCO2 responders" were defined by a decrease in PaCO2 > 2 mmHg after 15 hours of PP (PP H15). Shown are box and whisker plots. Median = horizontal line inside the box; upper and lower quartiles = whisker plot. Boxes and triangles represent values higher or lower than the upper or lower quartiles. *P < 0.05 for comparison of changes in responders versus nonresponders. PP: prone position.

Full size image
Figure 3
figure 3

Correlation between changes in alveolar dead space (ΔVD alv /V T ) and changes in compliance of the respiratory system (ΔCrs, left) or in PaO 2 /FiO 2 (ΔPaO 2 /FiO 2 , right) at each time of the study when compared with the supine position.

Full size image

As shown in Figure 4, estimated VDphysiol/VT systematically underestimated measured VDphysiol/VT, with a poor concordance correlation coefficient of 0.19 (95% confidence interval (95% CI) 0.091 to 0.28), a bias of 0.16 and an agreement between -0.05 and 0.37. Concerning changes in VDphysiol/VT during PP, estimated VDphysiol/VT had a concordance correlation coefficient of 0.51 (95% CI 0.32 to 0.66) (Figure 4).

Figure 4
figure 4

Comparison between measured VD physiol /V T and estimated VD physiol /V T [14]using a Bland and Altman representation (left) and a linear correlation (right). (A) and (B) Comparison for each paired data set (n = 78) in the supine position and after 3, 6, 9, 12 and 15 hours in the prone position. (C) and (D) Comparison of changes in VDphysiol/VT assessed according to the two methods between each time of measurement and the previous one. VDphysiol/VT: ratio of physiological dead space to tidal volume.

Full size image

Discussion

One of the objectives of our study was to describe alterations in VDalv induced by PP. ARDS is characterized by a heterogeneous lung with the existence of a slow compartment [18, 20], defined as areas available for, but partially or totally excluded from, ventilation due in part to a bronchiolar collapse [12, 21]. In a previous study, we reported that PP may induce recruitment of this slow compartment, as suggested by its ability to counteract intrinsic PEEP and to decrease the expiratory time constant [12]. In the same study, we also reported that PP leads to a decrease in PaCO2, suggesting diminution of VDalv (alveolar dead space) [12]. Our present study demonstrates that PP may induce a decrease in VDalv. It occurred from the third hour and was maintained throughout the PP session. VDalv may be the consequence of nonperfused or poorly perfused lung areas in ventilated anterior areas, but also of a slow compartment partially excluded from ventilation. Our results suggest that PP induces functional lung recruitment, especially since decreases in VDalv related to PP were associated with a decrease in Pplat and strongly correlated with improvement in compliance. Interestingly, in a previous study of 16 ARDS patients, Pelosi et al. [22] did not find a decrease in VDphysiol after 120 minutes in PP. One of the explanations for this discrepancy could be the different levels of PEEP in the two studies: 12.3 cmH2O in Pelosi et al.'s study and only 6 cmH2O in our study. However, Protti et al. [23], in a study of patients ventilated with a PEEP of 13 cmH2O, demonstrated a strong relation between lung recruitability and decreased PaCO2 related to PP. Pelosi et al. also did not report a decrease in Pplat in PP, as we found, but after returning patients to the supine position [22]. This could be explained by the fact that they used roll under the upper part of the chest wall, leading to a significant impairment in chest wall compliance [22], whereas we did not.

The most beneficial reported effect of PP is oxygenation improvement [24, 25]. However, this better oxygenation can be due to (1) lung recruitment related to restoration of functional residual capacity [7] and improvement of the diaphragmatic movement in the posterior part [26–28] or (2) simply to an improvement in the ventilation/perfusion ratio due to a decreased hydrostatic gradient between the anterior and posterior parts of the lung [26, 29]. Whereas the first mechanism is crucial, one can say that the second mechanism is less important. This is why the second objective of our study was to test whether the response to PP in terms of PaCO2 was physiologically more relevant than in terms of PaO2/FiO2 ratio. Gattinoni et al. [10] reported that an increase in PaO2/FiO2 ratio > 20 mmHg after six hours of PP is not predictive of the patient's prognosis, whereas a decline in PaCO2 ≥1 mmHg is. In our present study, 7 of 13 patients were PaO2 responders (increased PaO2/FiO2 ratio > 20 mmHg after 15 hours of PP). However, changes in Pplat, PaCO2 and VDalv did not differ between PaO2 responders and PaO2 nonresponders. On the other hand, 7 of 13 patients were PaCO2 responders (decreased PaCO2 > 2 mmHg after 15 hours of PP). PaCO2 responders had a significant decrease in Pplat and VDalv, as well as a significant increase in oxygenation and compliance, compared with nonresponders. Our results are in accordance with a recent study of 32 ARDS patients [23], in which the investigators reported that PaCO2 variation induced by PP, and not PaO2/FiO2 variation, is associated with lung recruitability. Interestingly, in our study, changes in VDalv were not correlated with changes in oxygenation but were strongly correlated with changes in compliance of the respiratory system.

An unexpected result of our work concerns the change over time of respiratory mechanics, blood gas analysis and VDalv. For many years, our PP protocol has been to turn patients to PP for up to 15 to 18 hours per day for 3 days [15]. In the study by Mancebo et al. [30], which concluded that PP may reduce mortality in patients with severe ARDS, PP sessions lasted 20 hours/day. In a recent study, we demonstrated that PP sessions that lasted 18 hours/day were independently associated with survival [31]. In the present study, the maximum effect of PP for VDalv, PaCO2 and Pplat occurred six to nine hours after turning patients to PP. Later the effect seemed to be a decline. How this affects the effect of PP on patient prognosis remains to be elucidated.

The second objective of our study was to validate a recently proposed method to evaluate the VDphysiol/VT ratio [14]. The method is based on CO2 production calculated from the Harris-Benedict equation [19] and on the expired minute ventilation. Siddiki et al. [14] reported that it was associated with mortality in acute lung injury patients in a dose-response manner and proposed its routine use to estimate VDphysiol/VT. However, they did not report any comparison with measured VDphysiol/VT. In the present study, we have demonstrated that this method significantly underestimates VDphysiol/VT, rendering it not accurate enough to assess the degree of lung injury. Interestingly, changes in estimated VDphysiol/VT during PP appeared better correlated with changes in measured VDphysiol/VT and could be proposed in the future in this field. Siddiki et al. [14] proposed the method in the context of a much larger series than ours and in patients with less severe ARDS, rendering it difficult to draw any definitive conclusions.

Our work is limited by the small number of patients included. This is a consequence of our routine protocol, which strictly restricts PP to patients with the most severe ARDS, that is, those with a PaO2/FiO2 ratio < 100 mmHg after 48 hours of ventilation. This also explains why it is not possible to link our results to outcomes. However, despite this limitation, we consider our results relevant from a physiological point of view.

Conclusions

In conclusion, our study demonstrates that PP induces a decrease in PaCO2 and VDalv. This is related to an improvement in respiratory mechanics, with a decrease in Pplat and an increase in compliance. Testing the response to PP appeared to be physiologically more relevant using PaCO2 changes than PaO2/FiO2 changes. How this may affect management at the bedside remains to be studied. Estimated VDphysiol/VT ratios systematically underestimated measured VDphysiol/VT ratios.

Key messages

  • PP induced a decrease in VDalv/VT, which was correlated with an improvement in respiratory mechanics.

  • Defining the respiratory response to PP appeared more relevant when using PaCO2 changes rather than PaO2/FiO2 changes.

  • Estimated VDphysiol/VT using the Harris-Benedict equation systematically underestimated measured VDphysiol/VT.