Ten pigs [32 (31–33) kg] were positioned supine and premedicated with Zoletil Forte (tiletamine and zolazepam) 6 mg kg⁻¹ and xylazine 2.2 mg kg⁻¹ i.m. and anesthetized with i.v. infusion of ketamine 30 mg kg⁻¹ h⁻¹, midazolam 0.1 mg kg⁻¹ h⁻¹, and fentanyl 3.75 μg kg⁻¹ h⁻¹. After assessment of the adequacy of anesthesia—defined as unresponsiveness to painful stimulation between the front hooves—muscle relaxation was obtained with a continuous i.v. infusion of rocuronium 0.3 mg kg⁻¹ h⁻¹. After a bolus of fentanyl 10–20 μg kg⁻¹, an 8 mm inner diameter endotracheal tube (Mallinckrodt Medical, Ireland) was positioned through a surgical anterior neck access. An arterial line was placed in the right carotid artery, and a central venous line was placed in the right jugular internal vein through surgical access. A pulmonary artery catheter (Criti Cath, Ohmeda Pte Ltd., Singapore) was introduced via an external jugular vein, and its correct position was verified by pressure monitoring. Cardiac output (CO) was measured in triplicate with the thermodilution technique (Siemens SC 9000XL, Dräger, Germany). During the experiment, airway pressure (Paw) and flow (f) were continuously recorded at the airway opening. Esophageal pressure (Pes) was measured with an esophageal balloon (CooperSurgical, Inc., Trumbull, USA) inserted through the oral route; its correct position was verified according to Baydur et al. (1982). Paw and Pes were measured by pressure transducers (Digima Clic pressure transducers; Special Instruments, Nördlingen, Germany). Flow was recorded by a Fleisch pneumotachograph (Laminar Flow Element type PT; Special Instruments, Germany) positioned between the endotracheal tube and the ventilator and connected to a differential pressure transducer (Diff-Cap pressure transducer; Special Instruments, Nördlingen, Germany). Signals from the transducers were forwarded to an analog-to-digital converter card (DAQ-card AI-16XE50; National Instruments, Austin, TX, USA) with a sampling frequency of 200 Hz, using the BioBench Software (National Instruments, Austin, TX, USA). A 32-electrode electrical impedance tomography (EIT) belt was placed around the shaved chest of the animal and connected to a dedicated monitor (Enlight 1800, Timpel, Brazil). The signal quality was checked at the beginning and constantly monitored during the experimental protocol. During instrumentation, the animals were ventilated using volume-controlled ventilation (VCV) mode with a Servo-I ventilator (Getinge, Solna, Sweden) with tidal volume (VT) 6 ml kg⁻¹, PEEP 5 cmH₂O, inspiratory to expiratory ratio (I:E) 1:2, respiratory rate (RR) 30 bpm, and FiO₂ 0.4.

After instrumentation, a two-hit lung injury was used to create an ARDS model: the first hit was warm saline lung lavage (20 mkg⁻¹), repeated until PaO₂/FiO₂ (P/F) ratio was <150 mmHg. Thereafter, 120 min of injurious mechanical ventilation (PEEP 2 cmH₂O, inspiratory pressure 35 cmH₂O, I:E 1:2, and RR 20 bpm) was started (Borges et al., 2015). Once the injurious model was established, the animals were stabilized for 30 min in VCV (VT 6 ml kg⁻¹, PEEP 5 cmH₂O, and RR 20 bpm).

The protocol was divided into the following three consecutive steps: (1) PEEP titration, (2) gas exchange and respiratory mechanics, and (3) lung imaging. During the whole protocol, 5 min was allowed to elapse before measurements after each change of position.

The EIT images were analyzed trough a purposely made MatLab script. To allow inter-animal comparisons, in each EIT image, the variation of impedance related to ventilation (ΔZ_V) was normalized to the highest pixel level ΔZ_V, yielding values ranging from 0 to 1. To assess lung perfusion, a bolus of hypertonic saline (NaCl 10%, 10 ml) was injected intravenously using the distal tip of the central vein catheter during an expiratory hold maneuver. The variation of impedance related to the saline passage through the pulmonary circulation was recorded and graphically reconstructed in a 32 × 32 pixel EIT image, as previously described (Borges et al., 2012). The relative pixel variation of impedance was normalized to the maximum pixels' value to allow inter-animal comparisons. The global inhomogeneity index of ventilation (GIv) was calculated using the following formula (Zhao et al., 2014):

To evaluate heterogeneity in perfusion assessed by EIT, impedance maps related to perfusion were analyzed using the same process. The variation of impedance related to perfusion (ΔZp) was normalized to the highest pixel level ΔZp, yielding values ranging from 0 to 1. The global inhomogeneity index of perfusion (GIp) was therefore calculated as follows:

The EIT derived regional compliance was calculated both for the respiratory system and for the lungs, assuming that ΔZv = ΔV, as previously described (Spadaro et al., 2018). Region of interest division was made using the mass content derived from the corresponding CT scans. For regional analysis of EIT ventilation and perfusion data, we assumed that the vertebral–dorsal gradient observed in the EIT slice was representative of the whole lung. For the analysis of ventilation at the ROI level, we computed for each ROI the estimated specific ventilation, measured in milliliters of tidal volume per minute per gram of lung tissue, using the following formula:

where ΔZ_V is the variation of impedance related to ventilation, V_T the tidal volume, and RR the respiratory rate. Similarly, we computed for each ROI the estimated specific perfusion, measured in milliliters of blood flow per minute per gram of lung tissue, using the following formula:

where ΔZ_p is the variation of impedance related to perfusion and CO is the cardiac output in milliliters per minute. Finally, we computed the ventilation perfusion ratio for each ROI as follows:

The sample size calculation was based on a previously published study (Retamal et al., 2016), from which we expected a P/F ratio of 115 ± 5.5 mmHg in a double-hit model of ARDS pigs. In this crossover study design, we expected a strong within-subject correlation (R > 0.6) between the P/F ratio in prone and supine positions, and we considered clinically relevant an increase of 10%. This increase corresponds to an effect size of 1.41; therefore, accounting for the asymptotic relative efficiency of non-parametric tests, we needed to include at least eight animals to achieve 90% power (1 – β) to detect such difference at an α level of 0.05. To compensate for a potential drop-off rate of 20%, we aimed to include 10 pigs. Values are expressed as median (first to third quartiles). We compared continuous variables using the Friedman test (non-parametric analysis of variance for repeated measurements) with the Dunn post-hoc test (prone vs. supine position using the same PEEP titration technique, two comparisons). For continuous variables with missing measurements, to maximize the use of data, we used a mixed-effects model comprising positioning as a dichotomous fixed effect with a fixed interaction (positioning within PEEP titration mode) and animal as a random effect with a random intercept. The correlation between variables was assessed using the Spearman correlation for non-parametric data. The statistical analysis was carried out using SPSS 19 and GraphPad Prism 6. Statistical significance was set at p < 0.05.

Owing to unavailability of EIT machine or major artifacts in perfusion assessment, we had 30% (12/40) missing data points for ventilation maps and 53% (21/40) for perfusion maps.

An overview of gas exchange, hemodynamics, and respiratory mechanics in the different positioning/PEEP combinations is provided in Table 1. PaO₂/FiO₂ increased in prone vs. supine position with BP_ARDS [difference 54 (19–106) mmHg, p = 0.04], but not with BP_DPL [difference 17 (−24 to 68) mmHg, p = 0.99]. The respiratory system driving pressure decreased during prone positioning with both BP_ARDS [difference −1.0 (−2.0 to 0.0) cmH₂O, p = 0.03] and BP_DPL [difference −1.8 (−2.0 to 1.00) cmH₂O, p = 0.03]. The transpulmonary driving pressure significantly decreased during prone position with both BP_ARDS [difference −0.9 (−1.5 to −0.9) cmH₂O, p = 0.009] and BP_DPL [difference −0.55 (−1.6 to −0.4) cmH₂O, p = 0.04] (Figure 1). Change in PaO₂/FiO₂ did not correlate with change in respiratory system driving pressure from supine to prone position, neither with BP_ARDS (ρ = 0.28; p = 0.47) nor with BP_DPL (ρ = 0.08; p = 0.83). The MP was not reduced with prone positioning (p = 0.60 and p = 0.11 with BP_ARDS and BP_DPL, respectively).

Lung aeration data derived from CT scan analysis are reported in Table 2. With BP_ARDS, and hence at same PEEP, prone position increased normally aerated lung tissue (p = 0.019) and decreased strain (p = 0.019). With BP_DPL, prone position did not increase normally aerated tissue (p > 0.99), increased non-aerated tissue (p < 0.001) and did not reduce strain (p = 0.77). There was no correlation between the PaO₂/FiO₂ variation and the strain variation from supine to prone position with BP_ARDS (ρ = 0.30; p = 0.43) or BP_DPL (ρ = 0.47; p = 0.17). Ventilation and perfusion EIT data are reported in Table 3. Prone positioning shifted the center of ventilation toward vertebral direction (p = 0.006 and p < 0.001 with BP_ARDS and BP_DPL, respectively) and reduced the GIv (p = 0.001 and p < 0.001 with BP_ARDS and BP_DPL, respectively) but had no effect on the GIp or on the center of perfusion.

Detailed data on regional CT and EIT analyses are reported in Supplement Tables 2, 3. As illustrated in Figure 2, with BP_ARDS, prone position shifted normally aerated tissue from sternal to vertebral ROIs and decreased poorly aerated tissue in the two most vertebral ROIs (p < 0.001). With BP_DPL, prone position increased normally aerated tissue in the vertebral ROIs (p = 0.004), whereas it increased non-aerated tissue in mid-sternal (p = 0.006) and sternal (p < 0.001) ROIs. The regional compliance of the respiratory system improved in the vertebral ROI in prone vs. supine position with both BP_ARDS (p = 0.003) and BP_DPL (p = 0.043), whereas it was unchanged elsewhere. As shown in Figure 3, strain was distributed more homogeneously between ROIs in prone position with both BP_ARDS [coefficient of variation from 52 (49–56)% in supine to 21 (12–42)% in prone, p < 0.001] and BP_DPL [from 55 (52–63)% to 19 (12–35)%, p < 0.001]. With BP_ARDS, in prone positioning, a tidal recruitment increase was observed in the sternal ROI (p = 0.03), with a symmetric reduction in the vertebral ROI (p = 0.019). With BP_DPL, prone positioning increased tidal recruitment in the sternal ROI (p = 0.011). In prone position, regional ventilation assessed with EIT was redistributed from sternal to vertebral regions in both BP_ARDS and BP_DPL (Figure 4A), whereas perfusion was marginally affected only with BP_DPL (Figure 4B). The ventilation/perfusion ratio increased in vertebral and decreased in sternal regions (Figure 4C). With BP_ARDS, the ROIs with a ventilation/perfusion ratio below 1 were 2/4 (50%) and 1/4 (25%) in supine and prone positions, respectively. With BP_DPL, 1/4 (25%) ROIs had a ventilation/perfusion ratio below 1 in both prone and supine positions. Prone position did not reduce the coefficient of variation across ROIs of ventilation, perfusion, and ventilation/perfusion ratio regardless of the PEEP titration technique (see Supplementary Table 4).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.