Effect of PEEP and Tidal Volume on Ventilation Distribution and End-Expiratory Lung Volume: A Prospective Experimental Animal and Pilot Clinical Study

Introduction Lung-protective ventilation aims at using low tidal volumes (VT) at optimum positive end-expiratory pressures (PEEP). Optimum PEEP should recruit atelectatic lung regions and avoid tidal recruitment and end-inspiratory overinflation. We examined the effect of VT and PEEP on ventilation distribution, regional respiratory system compliance (CRS), and end-expiratory lung volume (EELV) in an animal model of acute lung injury (ALI) and patients with ARDS by using electrical impedance tomography (EIT) with the aim to assess tidal recruitment and overinflation. Methods EIT examinations were performed in 10 anaesthetized pigs with normal lungs ventilated at 5 and 10 ml/kg body weight VT and 5 cmH2O PEEP. After ALI induction, 10 ml/kg VT and 10 cmH2O PEEP were applied. Afterwards, PEEP was set according to the pressure-volume curve. Animals were randomized to either low or high VT ventilation changed after 30 minutes in a crossover design. Ventilation distribution, regional CRS and changes in EELV were analyzed. The same measures were determined in five ARDS patients examined during low and high VT ventilation (6 and 10 (8) ml/kg) at three PEEP levels. Results In healthy animals, high compared to low VT increased CRS and ventilation in dependent lung regions implying tidal recruitment. ALI reduced CRS and EELV in all regions without changing ventilation distribution. Pressure-volume curve-derived PEEP of 21±4 cmH2O (mean±SD) resulted in comparable increase in CRS in dependent and decrease in non-dependent regions at both VT. This implied that tidal recruitment was avoided but end-inspiratory overinflation was present irrespective of VT. In patients, regional CRS differences between low and high VT revealed high degree of tidal recruitment and low overinflation at 3±1 cmH2O PEEP. Tidal recruitment decreased at 10±1 cmH2O and was further reduced at 15±2 cmH2O PEEP. Conclusions Tidal recruitment and end-inspiratory overinflation can be assessed by EIT-based analysis of regional CRS.


Introduction
It is well known that mechanical ventilation leads to lung injury. Dreyfuss and coworkers comprehensively showed the relationship between mechanical ventilation and pathologic lung tissue changes [1]. Liable factors inducing lung injury are high plateau pressures (P plat ), high tidal volumes (V T ) and cyclic opening and closing of alveoli (tidal recruitment) [2].
To minimize ventilator-induced lung injury the concept of lungprotective ventilation was introduced comprising the limitation of P plat , reduction of V T and optimization of positive end-expiratory pressure (PEEP) [3][4][5]. The current recommendations advocate V T of 6 ml/kg predicted body weight and P plat lower than 30 cm H 2 O in patients with acute lung injury (ALI) and acute respiratory distress syndrome (ARDS). Whether this recommended V T is optimal or whether the risk of ventilator-induced lung injury can be reduced by further reduction of V T is still under debate. However, V T lower than 6 ml/kg predicted body weight poses the risk of impaired alveolar ventilation and oxygenation [6,7].
Proposed measures to avoid derecruitment are increased PEEP and repeated recruitment maneuvers [8]. The effectiveness of these strategies to improve oxygenation was demonstrated in multiple studies [9,10,11]. However, other studies revealed that these measures resulted in regional lung overinflation with subsequent lung tissue injury [8,12,13]. This may be one of the reasons why studies comparing different PEEP strategies were inconclusive and mostly failed to demonstrate the benefit of PEEP [5,8,14,15]. Individualized identification of optimal PEEP and P plat according to mechanical properties of the lung and the thorax is discussed as a potential solution to the dilemma [4,16]. The goal should be to recruit the lung, to avoid tidal recruitment and endinspiratory overinflation.
Electrical impedance tomography (EIT) may offer new diagnostic possibilities in the assessment of recruitment (increase in end-expiratory lung volume (EELV)), tidal recruitment (change in respiratory system compliance (C RS ) with different V T ) and endinspiratory overdistension (decrease in C RS ) in mechanically ventilated patients with ARDS. Since EIT can determine regional distribution of ventilation, changes in EELV [17][18][19][20], and C RS [21,22] it might be applied at the bedside in addition to the measurement of global respiratory mechanics and gas exchange. Previous studies employed EIT to identify recruitment, derecruitment and overinflation. In these studies the following strategies were mainly used: stepwise variation of PEEP [23] and the low flow inflation [24,25] or the stepwise airway pressure increase [26].
To our knowledge, it is not predictable how distribution of ventilation, regional EELV, and regional C RS are affected by concomitant changes in V T and PEEP in patients with severe ARDS. Thus, the goal of this experimental study was to perform EIT measurements during ventilation with low and high V T at a preset PEEP to detect tidal recruitment. We hypothesized that low PEEP would lead to tidal recruitment whereas high PEEP, especially in combination with high V T , would lead to overinflation of at least some parts of the lung. The model and the hypothesis on which the model is based are described in more detail in Text S1 and Figure S1. We additionally checked whether this EIT-based analysis revealed tidal recruitment and endinspiratory overinflation in pilot examinations in critically ill patients with mild and moderate ARDS.

Experimental study
The experimental study was performed on ten anesthetized supine pigs of both sexes (Deutsches Landschwein, Institute of Animal Breeding and Husbandry, Christian Albrechts University, Kiel, Germany) with a body weight (BW) of 5065 kg (mean6SD). It was carried out in strict accordance with the guidelines on animal experimentation. The protocol was approved by the Committee for Animal Care of the Christian-Albrechts University, Kiel, Germany (Permit Number: V 742-72241.121-39 (80-10/ 03)). All surgery was performed under anesthesia with propofol and sufentanile and all efforts were made to minimize suffering.

Animal preparation
After sedation with azaperon (8 mg/kg BW) and atropine (0.1 mg/kg) general anesthesia was started with ketamine (5 mg/ kg BW), sufentanile (0.2 mg/kg BW) and propofol (1 mg/kg BW). Anesthesia was continued with intravenous infusion of propofol (6 to 12 mg/kg BW per hour) and sufentanile (10 mg/kg BW per hour). Vecuronium bromide (0.1 mg/kg BW) was administered for muscle paralysis. An infusion of lactated Ringers solution was given at a rate of 20 ml/kg BW per hour. If hypovolemia was suspected additional hydroxyethyl starch and Ringers solution were given. Norepinephrine was infused to maintain the mean arterial pressure above 70 mm Hg.

Mechanical ventilation and induction of ALI
Throughout the experiment, the animals were ventilated in a volume-controlled mode (Avea, CareFusion, Höchberg, Germany) at 20 breaths/min with the ratio of inspiration and expiration times (I:E) of 1:1.5. ALI was induced by repeated bronchoalveolar lavage with 1.5 L of warm saline solution until arterial partial pressure of oxygen (P a o 2 ) remained stable below 100 torr at an inspired oxygen fraction (F I o 2 ) of 1.0 and PEEP of 10 cm H 2 O for 30 min. The animals were ventilated with V T of 5 and 10 ml/kg BW both before and after ALI induction. These are referred to as low and high V T throughout the text.
A constant low-flow (2.5 l/min) inflation (pressure-volume (PV)) maneuver with an inspiratory volume limit of 1.5 l was performed starting at zero end-expiratory pressure. Afterwards, positive endexpiratory pressure (PEEP) was set 2 cm H 2 O above the lower inflection point (LIP) identified in the PV curve. An additional image file shows the PV curve obtained from a representative animal ( Figure S2).

Interventional lung assist device
A pumpless extracorporeal interventional lung assist device (ILA, Novalung, Hechingen, Germany) was applied in all animals allowing control of arterial partial pressure of carbon dioxide (P a co 2 ) independent of the ventilator pattern [27,28]. A 13 Fr cannula was inserted into the iliac artery with ultrasound guidance and a 15 Fr cannula into the iliac vein using Seldingers technique. The ILA device was prefilled with saline solution and connected to both cannulae. 5000 units of heparin were given after the instrumentation was completed. ILA was only used after induction of ALI during low V T ventilation. Oxygen flow was set to 10 l/ min to achieve a P a co 2 of 40 mmHg.

Electrical Impedance Tomography
EIT measurements were performed with the Goe-MF II system (CareFusion, Höchberg, Germany) using a set of 16 electrodes (Blue Sensor BR-50-K, Ambu, Ølstykke, Denmark) placed on the thoracic circumference at the fifth-sixth intercostal space. EIT data were acquired with a scan rate of 25 Hz. EIT images were generated using the filtered back-projection algorithm [29]. The data were filtered using a digital low-pass filter with a cut-off frequency of 1 Hz to eliminate small impedance changes synchronous with the heart beat.

Experimental protocol
A flowchart of the experimental protocol is provided in Figure 1. EIT scanning was performed during baseline conditions and at six subsequent measurement time points as described below:   Animals were then allocated to either ventilation with low V T and active ILA or to ventilation with high V T and inactive ILA in randomized order (5 animals in each randomization arm). After 30 minutes, the applied V T /ILA pattern was changed in a crossover design where each animal served as its own control.  At baseline and at each measurement time point, heart rate, mean arterial pressure, inspiratory peak pressure (P insp ), plateau pressure (P plat ) and end-expiratory partial pressure of CO 2 (Pco 2 ) were determined and EIT data acquired during 60 seconds. (To account for the crossover design, the data obtained at identical V T /ILA settings were combined. This resulted in the merged time points 5/3 and 6/4 for low V T and 3/5 and 4/6 for high V T ventilation).
Blood gases were measured at time point 1, after ALI induction (i.e. time point 2), immediately after the pressure-volume maneuver was performed (during ventilation with high V T and PEEP set 2 cmH 2 O above LIP) and at the end of the experiment to check for the stability of the ALI model.

EIT data analysis
Functional EIT scans were generated from each measurement using an established approach [19]. They showed the distribution of regional V T in the chest cross-section by calculating the tidal amplitudes of relative impedance change in 912 image pixels ( Figure 2).
Ventrodorsal profiles of fractional V T in 32 layers were generated from the functional ventilation scans and the geomet-rical centers of ventilation were calculated in relation to the ventrodorsal chest diameter as previously described [30][31][32].
To compare and display changes in ventilation distribution between the baseline and individual measurement time points, EIT ventilation difference images [33] and regional ventilation difference profiles were generated (Figures 2 and 3). Changes in EELV between individual time points and baseline were analyzed by calculating the differences between the minimum (endexpiratory) values of ventilation-related relative impedance change.
Finally, global C RS was calculated as V T /(P plat 2PEEP) and regional C RS determined in each of the 32 layers as ((fraction of V T )?V T )/(P plat 2PEEP). Corresponding to the ventilation profiles described above, regional C RS and C RS difference profiles were generated ( Figure 3).

Pilot patient study
The study was approved by the Ethics committee of the Christian-Albrechts University, Kiel, Germany (''Bestimmung der globalen und regionalen Atemmechanik bei unterstützter Spontanatmung'' Permit Number: A 125/12). Written informed consent was obtained from each patient or their legal representative, respectively.
We included five adult patients (age 7466 years, height 17465 cm, weight 77615 kg) with mild and moderate ARDS  according to the Berlin definition [34] treated in our surgical intensive care unit. ARDS resulted from severe sepsis (n = 4) and pneumonia (n = 1). Anaesthesia was performed with sufentanile and propofol and if required for an intervention or intubation muscle paralysis was induced with rocuronium. Patients were ventilated with Evita XL (Drä ger Medical AG & Co., Lübeck, Germany) in the pressure-controlled mode. During the study period volume-controlled mode was used. A low-flow (4 L/min) PV maneuver was started at PEEP of 0 cm H 2 O with an inspiratory volume limit of 2 L and a pressure limit of 35 cm H 2 O. Afterwards, PEEP was set at three different levels according to the LIP. We started with a PEEP of LIP+2 cm H 2 O followed by LIP25 cm H 2 O and LIP+7 cm H 2 O. At each PEEP level, patients were first ventilated with low V T (6 ml/kg BW) followed by high V T (10 ml/kg BW) for 5 minutes. The lowest PEEP was second in order to avoid unnecessary long derecruitment and followed by the highest PEEP. We a priori decided not to exceed a pressure limit of 40 cm H 2 O at this PEEP level and, therefore, we reduced V T to 8 ml/kg BW.
EIT examinations and data analyses were performed exactly as described above in sections on the experimental study. (The only difference was that we used the L-00-S electrodes (Ambu, Ølstykke, Denmark) in patients.) We calculated C RS at each PEEP level and each V T and the C RS differences between high and low V T at each PEEP level.

Statistical analysis
Since normal distribution assumption was not violated (Shapiro-Wilk-test), data are presented as means6SD and analyzed parametrically with paired t test or repeated-measures ANOVA as appropriate. The pilot patient data were analyzed using descriptive statistics as means6SD.
The statistical analysis was conducted using SPSS version 17.0 (SPSS Inc., Chicago, IL, USA). All statistical tests were two-sided and the level of significance was set at 5%.

Experimental study
Hemodynamics and global respiratory system mechanics. All animals studied were included in the final analyses. Hemodynamic and respiratory data are summarized in Table 1. Pco 2 increased to about 60 torr during ventilation with high V T and inactive ILA whereas ventilation with low V T and active ILA resulted in normocapnia. Increasing V T from 5 to 10 ml/kg BW did not significantly change global C RS in the healthy lungs. ALI reduced C RS markedly to about 50% of baseline values despite PEEP increase to 10 cm H 2 O. Application of PEEP according to LIP restored C RS only partially.
Regional distribution of ventilation and respiratory system compliance. Increasing V T in the healthy lung from 5 to 10 ml/kg BW led to a small but significant redistribution of ventilation in favor of the dependent lung regions. The geometrical center of ventilation moved slightly but significantly downwards ( Figure 4). This went along with an increase in regional C RS in the dependent parts of the lung (Figure 3). In ALI, a homogenous reduction in C RS throughout the lung was observed during ventilation with high V T despite increased PEEP of 10 cmH 2 O (Figure 4). Distribution of ventilation was not different compared to the healthy lung ( Figure 4). Increasing PEEP to 2 cm H 2 O above the LIP of the pressure-volume curve restored regional C RS to initial values in the dependent lung regions and reduced it in the non-dependent ones. Accordingly, distribution of ventilation was directed more toward the dependent lung regions as reflected by the downward shift in geometrical centers of ventilation. There was no difference in ventilation distribution and Examples of functional EIT scans of regional lung ventilation in animal 2 during all measurement time points (see Figure 1 for explanation). The orientation of the scans is indicated (ant., anterior; post., posterior). Panel A: Ventilated areas within the chest cross-section exhibit higher values of relative impedance change and are shown in red tones. In normal lungs (NL), symmetrical ventilation distribution between the right and left lung regions was found. With induction of acute lung injury (ALI), higher ventilation in the right lung region with pronounced ventilation in its ventral part and reduced left lung ventilation especially in its dorsal part were found. After PEEP was set 2 cm H 2 O above the lower inflection point (LIP) according to the pressure-volume curve, a shift in ventilation toward the dependent (dorsal) lung regions was observed. No obvious difference between ventilation with 10 ml/kg V T and inactive interventional lung assist (ILA) (high V T , no ILA) and ventilation with 5 ml/kg V T and active ILA (low V T , ILA) was detected. Panel B: Ventilation difference scans of the same animal. Red color indicates increase in regional ventilation, blue color shows the decrease in ventilation compared with baseline. doi:10.1371/journal.pone.0072675.g002 regional C RS between ventilation with 5 or 10 ml/kg BW at the four measurement time points with PEEP increased to 2 cm H 2 O above LIP (Figures 3 and 4).
End-expiratory lung volume. Regardless of the applied V T , EELV did not change in the healthy lung ( Figure 5). ALI led to a pronounced loss in EELV despite PEEP of 10 cm H 2 O. This volume loss could only partially be regained when increased PEEP was set according to the pressure-volume curve with high variability among the animals ( Figure 5).
Stability of the model. There were no significant differences in gas exchange between the initial measurement after induction of ALI and the final measurements at the end of the experimental protocol (Table 2).

Pilot patient study
Respiratory and hemodynamic data are given in Table 3. LIP was identified at 862 cm H 2 O. Regional C RS differences at the respective PEEP levels are shown in Figure 6. At the lowest PEEP of 361 cm H 2 O we found a pronounced increase in C RS in the dorsal regions and a small decrease in the ventral regions with high V T . At PEEP of 1061 cm H 2 O, high V T led to a smaller increase in C RS in the dorsal regions and more pronounced reduction in the ventral parts than at the lowest PEEP. At PEEP of 1562 cm H 2 O, the C RS difference between high and low V T was even smaller.

Discussion
Our study examined regional ventilation, C RS and EELV using EIT at different V T and PEEP in lung healthy animals and after 3. Regional ventilation and respiratory system compliance. Ventrodorsal profiles of regional tidal volume (V T ) (A), regional respiratory system compliance (C RS ) (C) and differences in regional V T (DV T ) (B) and regional C RS (DC RS ) (D) with respect to baseline (mean values 6SD in 10 animals). Panel A (V T [%]) shows the relative distribution of V T in 32 horizontal layers in% of overall V T in the chest cross-section. Panel B (DV T [%]) indicates the respective differences in regional V T compared with baseline. It shows a shift in ventilation toward the dorsal regions already with higher V T at time point 1 in normal lung and more pronounced shifts at points 3 through 6 with a PEEP set 2 cm H 2 O above the lower inflection point in acute lung injury (ALI). Panel C (C RS [ml/kg H 2 O]) shows regional C RS in the same 32 layers with the respective differences to baseline provided in panel D (DC RS [ml/kg H 2 O]). The differences in regional C RS indicate slightly lower and higher C RS in the ventral and dorsal regions at time point 1.
Significantly lower values were found in layers 7 to 12 and significantly higher ones in layers 16 to 25 (layers counting from 1 to 32 in the ventrodorsal direction). ALI (time point 2) resulted in significant decrease in C RS in all layers. A small increase in C RS in the dorsal regions (significant in layers 23 and 24) with the decreased C RS in the ventral regions (significant in layers 1 to 17 and 27 and 28) at time point 3/5. At all other three time points, the findings were comparable. High V T , ventilation with 10 ml/kg BW, low V T , ventilation with 5 ml/kg BW. doi:10.1371/journal.pone.0072675.g003 induction of ALI. The protocol was designed to reflect clinical decision-making regarding the choice of PEEP and V T in ARDS patients. It tested the hypothesis that a variation of tidal volume at a preset PEEP could be used to assess tidal recruitment and therefore guide the choice of adequate (optimum) PEEP. We subsequently applied this EIT-based approach developed in the   experimental study in a small pilot study in mechanically ventilated patients.
In the experimental study, we found that an increase in V T from 5 to 10 ml/kg BW led to an increase in regional C RS and fractional ventilation in the dependent parts of the healthy lung. This is in accordance with our hypothesis that the application of different V T at a distinct PEEP can be used in combination with EIT to assess the recruitment potential of the lung. The increase in C RS in the dependent parts of the lung with high V T is hereby interpreted as an increase in ventilated volume in these parts of the lung ( Figure S1). With identical mechanical properties, an increase in ventilated volume leads to an increase in compliance. ALI induced a profound reduction in EELV accompanied by low global and regional C RS despite the application of a PEEP of 10 cm H 2 O. Surprisingly, the distribution of ventilation was similar to the healthy lung. The subsequent further PEEP increase according to the identified LIP of the pressure-volume curves restored C RS to near baseline values in the dependent lung regions. However, C RS in the ventral regions of the lung was markedly decreased implying overinflation. The fact that there was no difference in regional C RS and distribution of ventilation between V T of 5 and 10 ml/kg BW suggests that no tidal recruitment occurred. The fact that EELV did not reach the level observed in the normal lung even with the high PEEP has to be interpreted with caution, since the induction of ALI may have changed the electrical conductivity of the lung tissue [35].

Recruitment and derecruitment
Our finding of dorsally directed shift in ventilation and increase in regional C RS in the dependent lung regions at a PEEP of 5 cm H 2 O and high V T in normal lungs is in accordance with previous findings. Sinclair et al. who examined the effect of different PEEP (0 and 8 cm H 2 O) and V T (6 and 12 ml/kg) on cyclic airway collapse and recruitment using aerosolized fluorescent microspheres in a rabbit model revealed that cyclic tidal recruitment occurred with low PEEP in the healthy lung [36]. Improved ventilation in the dorsocaudal lung regions with high PEEP in that study was attributed to local recruitment with a postulated increase in regional C RS .
In our study, a PEEP of 10 cm H 2 O was not sufficiently high to prevent derecruitment after ALI induction as reflected by overall decrease in regional C RS . This derecruitment could be reversed after the pressure-volume maneuver and subsequent application of higher PEEP of about 21 cm H 2 O. At the same time, however, regional C RS fell in the non-dependent regions when compared with baseline. Thus, although PEEP had the beneficial effect of recruiting the lung in the dependent regions and thereby avoiding tidal recruitment it also lead to regional overdistension. This phenomenon was also identified by Grasso et al. in three different pig models of ALI, including the lavage model, where overinflation was present in the baby lung despite recruited lung areas [6]. The concomitant existence of lung regions exhibiting recruitment and overinflation was also determined in patients with ALI [37], which renders the selection of adequate PEEP so difficult in individual patients.

End-expiratory lung volume
In our study, ALI led to a pronounced reduction in EELV, which could not be offset by PEEP of 10 cm H 2 O. By further increasing PEEP according to the pressure-volume curve, EELV increased.
Several studies have proven the ability of EIT to detect PEEPdependent changes in EELV by analysis of end-expiratory impedance values [19,[38][39][40] as used in our analysis. The fall in end-expiratory impedance values associated with ALI development has previously been described, although in an oleic acid ALI model [33]. Previous EIT studies using either two [41] or three electrode planes [19] demonstrated that EIT-based evaluation of EELV requires a cautious selection of the electrode plane. Therefore, we chose the midthoracic plane for our EIT measurements. With this approach, most of the lung tissue was included in the analysis because the examined chest slice was approximately 10 to 15 cm thick [42].

Ventilation distribution
To characterize regional ventilation distribution by EIT, we have generated ventrodorsal ventilation profiles derived from the functional EIT scans and calculated the centers of ventilation [30,32,43]. This is a relatively simple but sensitive procedure that was previously applied to determine the effects of ventilation mode, PEEP, recruitment maneuvers or surfactant administration on ventilation distribution [30,31,43].
The redistribution of ventilation occurring between the individual measurement time points resulted in shifts of the centers of ventilation in the ventrodorsal direction. The dorsal shift identified in the animals ventilated with high V T compared with low V T before ALI implied tidal recruitment in the dependent lung regions. The centers of ventilation exhibited the dorsal most locations during ventilation with high PEEP set according to the pressure-volume curve after ALI. This was consistent with reduced ventilation in the non-dependent regions caused by overinflation accompanied by an increase in ventilation in the dependent regions caused by recruitment.
To better visualize the changes in regional ventilation induced by the study interventions (i.e., ALI, PEEP and V T changes), we also calculated ventrodorsal profiles showing the differences in regional V T at individual time points in comparison with baseline. These profiles highlighted the ventilation changes identified by the centers of ventilation by showing the respective changes in 32 chest layers.

Regional respiratory system compliance
Although the topographical distributions of regional V T and C RS have to be identical by virtue of the underlying calculation of regional C RS , the absolute values of regional C RS reflect the changing respiratory system mechanics in the course of the experiment. This was detected especially after the induction of ALI (measurement time point 2) where a dramatic loss in regional C RS was observed in all analyzed lung layers, whereas regional tidal volumes remained fairly unchanged (Figure 3).
The profiles of differences in regional C RS detected the changing C RS when compared with baseline: After the induction of ALI during ventilation at 10 cm H 2 O of PEEP (time point 2), a marked decrease in C RS was found. Higher regional C RS in the dependent lung regions at high PEEP in the later phase after ALI could be attributed to recruitment of lung tissue as shown by Sinclair et al. [36]. The simultaneous decrease in regional C RS in the ventral, non-dependent regions reflected regional overinflation. These results indicate that the distribution of regional C RS and regional differences in C RS are crucial for the interpretation of the PEEP and V T effects and that the threshold PEEP, where tidal recruitment begins or ceases, might be the optimal PEEP to achieve best possible recruitment and minimal overinflation.
The importance of regional C RS has also been highlighted by two recent EIT studies. Bikker et al. calculated C RS in four horizontal chest layers at 15, 10, 5 and 0 cm H 2 O of PEEP during a decremental PEEP trial [41]. They found that high PEEP led to an increase in regional C RS in the dependent part of the lung indicating recruitment but also to a decrease in C RS in the nondependent part suggesting overinflation. Dargaville et al. examined regional C RS in three horizontal layers of the lungs during an incremental/decremental PEEP trial using a total of 11 PEEP steps [44]. Regional recruitment, derecruitment and overinflation could be detected and the PEEP value identified at which the most homogeneous C RS distribution was achieved during the deflation limb of the maneuver. Our results show that a change in regional C RS with different V T can be used to determine recruitment potential, implying that tidal recruitment occurs and the choice of a higher PEEP could be advantageous.

Interventional lung assist
Based on former studies [28,45], we presumed that ventilation with low V T of 5 ml/kg BW would lead to CO 2 accumulation in our animal model of ALI which would not allow us to maintain the ventilatory pattern constant. Since we focused on the measurement of lung mechanics we considered it essential to keep the pattern constant throughout the whole experiment. The technique of interventional lung assist is easily available in our animal laboratory, therefore, we used it for CO 2 removal in the phases of ventilation with low V T . When we designed the experimental protocol, we expected from our previous experience that the ventilation with high V T during ALI would result in a P a co 2 of about 40 mmHg. During the experiments we saw it was slightly higher, nevertheless, we decided to adhere to our original protocol. Pilot patient data In the animal experimental phase we could not identify any significant differences in C RS between the low and high V T after ALI because the lungs were already maximally inflated by the applied high PEEP. Therefore, we could not test our EIT-based approach of identifying tidal recruitment and end-inspiratory overinflation in the injured lungs of the studied animals. However, our pilot patient EIT data acquired at three PEEP levels, allowed us to apply this analysis. At the lowest PEEP, tidal recruitment in the dependent regions could clearly be identified. At the two higher PEEP values, progressive reduction in tidal recruitment was seen. End-inspiratory overinflation in the non-dependent regions was present already at the PEEP level of 2 cm H 2 O above LIP. At the highest PEEP, regional overdistention at the higher of the two V T values was blunted by the already present PEEP-induced overinflation. We postulate that the individual optimum PEEP could be derived from similar EIT examinations at the bedside in the future: the variation of V T at different PEEP levels could identify the settings with minimum tidal recruitment and minimum overinflation.

1)
We chose a lavage model of ALI well aware of the fact that it does not closely reflect the clinical situation because it is recruitable with PEEP and high V T [46]. However, this was a desired feature of the model in our study in order to evaluate the ability of EIT to detect V T -dependent tidal recruitment. We could show that our ALI model was stable during the experiment and did not exhibit spontaneous recruitment in the course of time. 2) We limited the experiment to the crucial measurement time points to exclude the influence of time and, therefore, we applied only the high but not the low V T after ALI induction  Data are shown as mean values 6 standard deviation. V T : tidal volume, F I o 2 : fraction of inspired oxygen, PEEP: positive end-expiratory pressure, PIP, peak inspiratory pressure, C RS : respiratory system compliance, P a co 2 : end-expiratory partial pressure of carbon dioxide, HR: heart rate. *The measurement at LIP+7 and high V T was not conducted in patient 1 due to excess of peak inspiratory pressure limit of 40  with a PEEP of 10 cm H 2 O. We did not expect lacking tidal recruitment during ventilation at high PEEP set according to the pressure-volume curve, otherwise, we would have studied both V T at the lower PEEP of 10 cm H 2 O. Since the data analysis was performed offline it was too late to change the protocol. However, our pilot patient data acquired at lower PEEP values than in animal experiments could show that tidal recruitment could be reliably assessed by EIT-derived regional C RS . 3) EIT measurements were not compared with another established radiological imaging modality like computed tomography. This might be regarded as a limitation, however, the feasibility of EIT to assess regional ventilation has been previously validated with multiple standard imaging techniques [47][48][49]. 4) EIT does not measure absolute lung volumes and thus we were only able to report relative changes in EELV. The validity of using relative instead of absolute lung volumes was previously confirmed by using the nitrogen washout technique [40].
Conclusions 1) With a PEEP of 5 cm H 2 O, tidal recruitment was determined by EIT in the normal lung implying recruitment potential at this PEEP value. 2) PEEP set according to the pressure-volume curve at 2 cm H 2 O above LIP proved to be too high in the experimentally injured lung since no tidal recruitment was detected but pronounced regional overinflation was present in the nondependent lung regions. 3) Regional tidal recruitment and end-inspiratory overinflation was identified in patients with ARDS with EIT by calculation of regional C RS differences from measurements acquired at different V T and PEEP. 4) Concomitant analysis of regional V T , EELV and C RS using EIT holds substantial potential to titrate lung protective ventilation by facilitating choice of adequate PEEP to avoid tidal recruitment and adequate V T to prevent overdistension.

Supporting Information
Figure S1 Explanation of the model. Schematic presentation of postulated changes in regional lung ventilation and regional respiratory system mechanics during different phases of the study protocol. Each large circle symbolizes ventilated lung volume. The small blue and large red circles represent normally aerated and overdistended lung regions, respectively. The oval dark grey symbols indicate atelectatic lung regions. The transparent grey bars show schematically two of the 32 regions of interest (ROI) used in our EIT analysis. (The sizes of these representative ROIs were enlarged to enable better visual perception.) The effect of an intervention is displayed from left to right showing the compliance change in the pressure (P)-volume (V) coordinates in the respective ROI and the assumed differences in regional compliance (DC) in the whole lung. Upper panel: An increase in tidal volume (V T ) at a given constant positive end-expiratory pressure (PEEP) increases the ventilation in the dependent parts of the lung by recruiting atelectatic lung regions (reduction of the dark grey oval symbols).
Overdistension occurs in the non-dependent regions (increasing number of red circles). On the right, the decrease in compliance in the non-dependent ROI and its increase in the dependent ROI is explained in a P-V diagram. Additionally, the observed changes in the distribution of regional DC is shown. Middle panel: The effect of acute lung injury (ALI) with an increase in atelectatic lung (higher number of dark grey oval symbols) and the decrease in regional compliance is shown. Lower panel: Applying high levels of PEEP after ALI results in reduction of atelectasis (reduction of the number of dark grey symbols) along with an increase in compliance in the dependent ROI but also leads to a higher degree of overdistension in the non-dependent ROI (higher number of large red circles).