The animal preparations were described in detail elsewhere (Wittenstein et al., 2021). Briefly, intravenous anesthesia and muscle paralysis were induced and maintained with midazolam (1 mg kg^-1 h^-1), ketamine (15 mg kg^-1 h^-1), and atracurium (3 mg kg^{− 1} h^-1) in sixteen female pigs (German landrace, weighing 35–49 kg, Danish Specific Pathogen Free Certification (Health Status Management, 2022)). The intravascular volume was maintained with a crystalloid solution at a rate of 5 mL kg^-1 h^-1. Norepinephrine was used to maintain a mean arterial pressure of at least 60 mmHg throughout the experiments. Lungs were ventilated in volume-controlled mode: fraction of inspired oxygen (F_IO₂) of 1.0, tidal volume (V_T) of 6 mL kg^-1, positive end-expiratory pressure (PEEP) of 5 cmH₂O, inspiratory: expiratory (I:E) ratio of 1:1, the constant gas flow of 25 L min^-1, and respiratory rate (RR) adjusted to achieve arterial pH > 7.3 (Evita XL, Drägerwerk AG & Co. KGaA, Lübeck, Germany).

All skin incisions were preceded by infiltration with 2–5 mL lidocaine 2%. A 20 cm PiCCO catheter was inserted in the right internal carotid artery. A 7.5 Fr. pulmonary artery catheter was advanced through an 8.5 Fr. sheath, placed in the right external jugular vein until typical pulmonary arterial pressure waveforms were observed. Urine was collected with a bladder catheter inserted through a median mini-laparotomy.

Regional pleural pressures (P_Pl) were measured using three custom-made pressure sensors, calibrated and placed as previously described by our group (Kiss et al., 2019). Briefly, pressure sensors were made from a glued (Silicon-based clue, MBFZ toolcraft GmbH, Georgensgmünd, Germany) double layer of thin latex to build a sealed air-filled chamber (30 × 30 × 3 mm³). A silicon tube (50 cm length, 500 µm inner diameter, 1.2 mm outer diameter) was introduced into the chamber using seldinger technique and connected to the respective pressure transducers (163PC01D48-PCB, FirstSensors AG, Berlin, Germany) via Luer-lock. Before implantation, each sensor was inflated with 0.05 mL of room air and calibrated within an air-sealed chamber using pressures in the range of −50 to 50 cmH₂O within 30 s, yielding a pressure rate of 6.7 cmH₂O s^-1. For calibration, a linear correction was used.

For placement of the sensors, lungs were separated introducing a left-sided double-lumen tube (39 Fr., Silbroncho Fuji, Tokyo, Japan) through a tracheotomy, with the bronchial tip placed into the left main bronchus under fiberoptic control (Ambu aScope 3 and Ambu aView, Ambu GmbH, Bad Nauheim, Germany). Animals were positioned in the right lateral decubitus position, and OLV of the dependent lung (V_T = 5 mL kg^-1, RR = 35 min^-1) was initiated. Using video-assisted thoracoscopy (Thoracoscopy Set, Karl Storz, Tuttlingen, Germany), three pressure sensors were attached to the parietal pleura by staples in the following regions of the left hemithorax: 1) ventral (4–5th rib parasternal); 2) dorsal (4–5th rib paravertebral); and 3) caudal (8–9th rib paravertebral) (Figure 1). A thoracic drain was placed, and surgical wounds were sutured and sealed.

To mimic thoracic surgery conditions, a right-sided thoracotomy was performed between the medial-clavicular and the anterior axillary line in the fourth-fifth intercostal space, and a rib spreader was placed. Furthermore, systemic inflammation was induced by continuous administration of 0.5 μg kg^-1 h^-1 Lipopolysaccharides (LPS) from Escherichia coli O111:B4 (SIGMA Aldrich, St. Louis, United States) through the central venous line.

According to the protocol of the primary study, animals were randomized to either normo- or the hypovolemia group. For induction of hypovolemia, 25% of the calculated blood volume, estimated as 70 mL kg^-1. However, in this substudy all animals were analysed independently from their volume status. Animals were randomly submitted to one of four sequences during OLV according to a Latin square cross-over design: 1) a-b-c-d, 2) b-d-a-c, 3) d-c-b-a, and 4) c-a-d-b (with a = supine; b = left semilateral; c = left lateral; d = prone position; 30 min each). For protective low tidal volume OLV, volume-controlled mode was used (V_T = 5 mL kg^-1; F_IO₂ = 1.0, PEEP = 5 cmH₂O, I:E = 1:1, RR = 30–35 min^-1 to achieve arterial pH > 7.25, and gas flow = 25 L min^-1). To reset lung history between the interventions, animals were placed in the supine position and disconnected from the ventilator (20 s), TLV resumed with baseline settings until normalization of gas exchange, and an alveolar recruitment maneuver performed before the start of OLV in each position.

P_L and respiratory variables were first collected after completing instrumentation (Baseline). Measurements were then repeated 1 h after starting the LPS infusion (TLVsupine) and 30 min after turning the animal into each body position during OLV (OLVsupine, OLVsemilateral, OLVlateral, and OLVprone, 30 min each).

All animals of the primary study were included in this analysis. Data are presented as mean and standard deviation (SD) if not stated otherwise. The statistical analysis was conducted with SPSS (Version 27, IBM Corp, United States). Significance was accepted at p < 0.05. Differences between TLVsupine and OLVsupine were tested using a t-test for paired samples. Differences between the respective body position during OLV and sequences of interventions were compared using a repeated measures linear mixed-effects model with OLVsupine, OLVsemilateral, OLVlateral, and OLVprone as within-subjects factor and with latin-square sequence as between-subjects factor. Pairwise post hoc multiple comparisons were performed when appropriate and corrected for multiple comparisons according to Šidák. Differences of the binary outcome lung instability, defined as end-expiratory P_L < 2.9 cmH₂O (Hedenstierna et al., 1981) were tested using the χ² test and pair-wise binomial post hoc test corrected according to Bonferroni.

Global transpulmonary pressures are depicted in Table 1. Global end-expiratory, end-inspiratory and ΔP_L were lower during TLVsupine than during OLVsupine. During OLV, end-expiratory P_L was lower in lateral as compared to supine and prone position, while end-inspiratory and ΔP_L did not differ among body positions.

Regional end-expiratory P_L is depicted in Table 2. During TLVsupine, ventral and dorsal end-expiratory P_L were lower than during OLVsupine, while caudal end-expiratory P_L was not different.

During OLV, dorsal end-expiratory P_L was not different among the body positions during OLV. Ventral end-expiratory P_L was higher in OLVsupine as compared to all other positions and higher in OLVsemilateral as compared with OLVlateral and OLVprone. Caudal end-expiratory P_L was lower in the semilateral position as compared to OLVlateral and OLVprone. Lung instability was detected more often in semilateral (26 out of 48 measurements; p = 0.012) and lateral (29 out of 48 measurements, p < 0.001) as compared to supine position (15 out of 48 measurements) and more often in lateral as compared to prone position (19 out of 48 measurements, p = 0.027).

Regional end-inspiratory P_L is depicted in Table 2. Ventral, dorsal, and caudal end-inspiratory P_L were lower during TLVsupine as compared with OLVsupine. During OLV, ventral, dorsal and caudal end-inspiratory P_L were not different between the different positions. During TLVsupine, ΔP_L was lower in all regions as compared with OLVsupine (Figure 2). During OLV, ΔP_L was not different among the body positions.

Respiratory variables are represented in Table 3. Respiratory rate, plateau, and mean airway pressure, as well as elastance and resistance, were lower during TLVsupine compared with OLVsupine, while there were no differences between positions during OLV. Respiratory system MW and MP were lower during TLVsupine as compared with OLVsupine. During OLV, MW and MP were not significantly different between the different body positions.

Our finding that end-expiratory P_L increased when switching from TLV to OLV in supine position might be explained by the fact that the pressure difference over the ribcage is reduced during pneumothorax as well as a shift of the mediastinal content towards the side of pneumothorax, reducing intrathoracic pressure. During OLV, ventral and caudal end-expiratory P_L differed depending on body position, with the lowest values in lateral and semilateral position, suggesting increased lung instability in this position. This finding is line with previous investigations in healthy volunteers during two lung ventilation (Ferris et al., 1959; Milic-Emili et al., 1964; Washko et al., 2006) and in thoracic surgery patients (Chiumello et al., 2020) using esophageal balloon manometry. Our finding can be explained by the fact that in the lateral position, the ventilated lung is compressed by the weight of the mediastinum, which shifts to the side of ventilation following gravity once the thorax is opened. More pronounced lung compression may increase the risk of end-expiratory lung collapse. As a consequence, potentially higher PEEP values are necessary to keep the lung open in lateral position. However, it is a matter of ongoing debate on which PEEP should be chosen during OLV (Battaglini et al., 2020). In addition to the set PEEP, auto-PEEP can influence end-expiratory P_L. In our study however, auto-PEEP did not differ among the investigated body positions.

End-inspiratory P_L, as well ΔP_L, markedly increased in all regions of the lung during OLV as compared to TLV, theoretically indicating increased lung stress and tidal overdistension. High alveolar stress and overdistension are the driving forces of VILI (Güldner et al., 2016; Bates and Smith, 2018). We used a low V_T during OLV, which is considered to be protective. However, even this low V_T resulted in high end-inspiratory P_L and high ΔP_L, independently of the body position, putting the ventilated lung at risk of VILI during OLV. Our results might explain why the risk of postoperative pulmonary complications is increased after thoracic anesthesia (Uhlig et al., 2020).

The fact that, during OLV, MP was almost twice as high as during TLV is in line with the literature and likely explained by increased elastance. In a prospective, observational, single-center study in 30 patients, MP increased in the lateral position with OLV as compared to TLV (Chiumello et al., 2020). Interestingly, MP exceeded 12 J/min during OLV in our animals, a threshold that was associated with formation of lung edema during TLV in pigs (Cressoni et al., 2016). Furthermore, OLV was accompanied by an increase of elastance in our study, which means that the stress exerted on the parenchymal tissue increased, potentially leading to increased distension of the septal walls and VILI (Bates and Smith, 2018). The fact that elastance did not differ significantly among body positions contrasts with results reported during TLV elsewhere (Tanskanen et al., 1997). A likely explanation for this discrepancy is that, in our animals, elastance almost doubled during OLV compared to TLV, possibly masking less pronounced effects of body position. Furthermore, V_T adjusted to body weight during left lung ventilation was almost twice when adjusted for the relative size of the ventilated lung (assuming that the left lung accounts for approximately 45% of the total lung size), which may explain the increased elastance, despite the pneumothorax.

The findings of this study suggest that the added risk of VILI during of OLV, as compared to TLV, is explained by increased MP and lung distending pressures. Importantly, this was not amenable by protective low tidal volumes that are commonly used in clinical practice. It is worth noting that lateral and semi-lateral decubitus positon, the most common positions used during thoracic surgery with OLV, led to the highest risk of VILI, since they were accompanied by increased lung instability.

The present study knows several limitations. First, the study was conducted as an exploratory sub-study of another trial and lacked formal sample size estimation. Second, we did not investigate the effects of right lateral and semilateral decubitus positions. A previous study reported lower end-expiratory P_Pl, as estimated by oesophageal manometry, in the right lateral position compared to the left lateral position during OLV, which was explained by the lower volume of the left lung, thus leading to a smaller decrease in P_Pl during right decubitus compared to left decubitus position due to less traction of the smaller left lung on the pleural surface (Chiumello et al., 2020). Third, the cross-over design precluded assessment of lung injury. Therefore, we do not know whether lower end-inspiratory P_L during OLV in the lateral position increases the risk of VILI. Fourth, the thoracotomy limited the direct comparison of respiratory mechanics between TLV and OLV since the latter did not include assessment of chest wall mechanics. Fourth, the different shape of the rib cage and anatomical distribution of intrathoracic organs in pigs may lead to different amplitudes and distribution of pressures in humans.

In anesthetized pigs, protective low tidal volume OLV, as compared to TLV, increased lung stress and respiratory system MP and impaired respiratory system mechanics. Body position did not affect lung stress, or respiratory system mechanics and MP during OLV, but end-expiratory P_L was lowest in semilateral and lateral decubitus position.