After approval by the local Animal Care Committee, 36 pigs (23.8–37.0 kg) were anesthetized, mechanically ventilated, and instrumented. After induction of ALI, animals were randomly assigned to the ARDSnet protocol or to OLA with or without noisy ventilation (n = 9/group). Lung mechanics, gas exchange, hemodynamics, and distribution of pulmonary blood flow (PBF) were determined at intervals. After 6 hours, animals were killed, and the lungs were extracted for postmortem analysis (see below).

Mechanical ventilation was performed in volume controlled mode using the EVITA XL 4 Lab (Dräger Medical, Lübeck, Germany). Ventilator settings are summarized in Table 1.

ALI was induced by surfactant depletion (17) until Pa_O2/Fi_O2 < 200 mm Hg for 30 minutes.

In ARDSnet and ARDSnet+noisy groups, PEEP was set at 12 cm H₂O and Fi_O2 at 0.7 (2). In OLA and OLA+noisy groups, a RM (40 cm H₂O for 30 s) was performed and followed by a decremental PEEP trial to achieve the minimal elastance of the respiratory system (Ers) (18).

Noisy ventilation was applied on a breath-to-breath basis as sequence of randomly generated Vt values (n = 600; mean, 6 ml/kg) (7, 8). The coefficient of variation was 40%, corresponding approximately to the variability in healthy spontaneously breathing subjects (6) and capable of maximizing oxygenation (19). The breath-by-breath f was maintained constant. The flow rate was 30 L/min and active inspiratory time was adjusted at each cycle to achieve the target Vt. Because the inspiratory/expiratory ratio was fixed at 1:1, the inspiratory pause varied. The minute volume for one cycle of 600 breaths, but not for single breaths, was constant. After completion of the sequence, the system looped itself.

Respiratory mechanics, gas exchange, and hemodynamics were assessed at baseline, at injury, and at 1-hour intervals thereafter (time 1 to 6). PBF was marked with intravenously administered fluorescent microspheres at baseline, at injury, and at time 6 (20). Samples were obtained from the first whole lung bronchoalveolar lavage (BAL) and from a lavage performed immediately before killing animals. Lungs were removed at continuous airway pressures equal to PEEP. The left and right lungs were used for microspheres and tissue analysis, respectively. Diffuse alveolar damage (DAD) was evaluated by an expert blinded to the groups (21). Gene expression of IL-6, IL-8, transforming growth factor-β (TGF-β), amphiregulin, and tenascin-c (TNC) was analyzed using quantitative real-time polymerase chain reaction. Plasma, BAL, and lung tissue cytokine levels were measured by commercially available ELISA kits.

Data are presented as mean ± SD unless indicated otherwise. Student's t test, one-way analysis of variance (ANOVA), repeated-measurements ANOVA, Mann-Whitney U tests, and Wilcoxon's test were used as appropriate. Effects of noisy ventilation were tested separately in ARDSnet and OLA groups. Associations between two variables were determined with Spearman's rank correlation. Tests were performed using the SPSS software package (SPSS version 15.0, Chicago, IL), multiple comparisons adjusted according to the Bonferroni procedure, and statistical significance was accepted at P < 0.05.

Body weight, number of lavages (Table E1), and functional parameters at baseline (Figure 1, Table 2) did not differ between groups. Adjusted PEEP values in the OLA and OLA+noisy groups were identical in both groups (15.1 ± 1.5 cm H₂O). PEEP in the ARDSnet and ARDSnet+noisy groups were fixed at 12 cm H₂O according to a previously published table (2).

Figure 1. Effects of variable tidal volumes (Noisy) combined with the ARDS Network protocol (ARDSnet) and the open lung approach (OLA) on Pa_O2 (A and B) venous admixture (Q_VA/Qt) (C and D) and Pa_CO2 (E and F). Values are given as mean and standard deviation. Group and Time × Group effects were assessed by general linear model statistics. Statistical significance was accepted at P < 0.05.

Download Figure | Download Powerpoint

[More] [Minimize]

Mean Vt, f, and minute ventilation did not differ between groups during the experiment (Table 2). The coefficient of variation of Vt was higher in the groups with noisy ventilation.

Noisy ventilation led to lower mean and peak airway pressures (Pmean and Ppeak, respectively) and to lower Ers as compared with the ARDSnet and OLA groups (Figure 2) while not affecting the resistance of the respiratory system.

Figure 2. Effects of variable tidal volumes (Noisy) combined with the ARDS Network protocol (ARDSnet) and the open lung approach (OLA) on mean airway pressure (Pmean) (A and B) mean peak airway pressure (Ppeak) (Q_VA/Qt) (C and D) and elastance of the respiratory system (Ers) (E and F). Values are given as mean and standard deviation. Group and Time × Group effects were assessed by general linear model statistics. Statistical significance was accepted at P < 0.05.

Download Figure | Download Powerpoint

[More] [Minimize]

Noisy ventilation improved Pa_O2/Fi_O2 compared with the ARDSnet protocol and OLA (Figures 1A and 1B) while decreasing intrapulmonary shunt when compared with ARDSnet protocol (Figure 1C) but not OLA (Figure 1D) and without affecting Pa_CO2 (Figures 1E and 1F).

Heart rate, mean arterial blood pressure, and cardiac output did not differ between groups (Table E4). Mean pulmonary arterial blood pressure was lower with noisy ventilation compared with ARDSnet. Other variables did not differ between groups.

Distribution of PBF in one representative animal per group is illustrated in Figure 3. Data on angular coefficients of relative PBF are presented in Table 3.

Figure 3. Regional distribution of pulmonary blood flow in four representative animals at baseline (upper row), after injury (middle row), and after 6 hours of protective mechanical ventilation according (A) the ARDS Network protocol (ARDSnet) and (B) the open lung approach (OLA) with and without variable tidal volumes (noisy) (lower row: Therapy). Blue represents lowest and red highest relative pulmonary blood flow (Q̇rel), respectively, in each condition. Straight lines in the scatter plots represent linear regression of (Q̇rel) along the central-peripheral (X), dorsal-ventral (Y), and caudal-cranial (Z) axes.

Download Figure (Part 1) | Download Figure (Part 2) | Download Powerpoint

[More] [Minimize]

In all groups, ALI led to redistribution of PBF from dorsal to ventral, from caudal to cranial, and from central to peripheral lung zones. At time 6 compared with injury, all mechanical ventilation modes led to a redistribution of PBF toward dorsal regions, whereas a redistribution of PBF toward cranial areas was observed with OLA and noisy ventilation combined with both protective strategies. ARDSnet and OLA redistributed PBF toward central areas, whereas noisy ventilation did not. Comparisons at time 6 showed that noisy ventilation did not affect dorsal-ventral distributions of PBF. However, noisy ventilation redistributed PBF toward caudal areas compared with ARDSnet and toward peripheral zones compared with OLA.

Noisy ventilation reduced the overall DAD score compared with both protective ventilation strategies (Figure 4). Compared with ARDSnet, noisy ventilation reduced the DAD score in dependent regions. In the ARDSnet group, noisy ventilation reduced interstitial edema, hemorrhage, and epithelial destruction mainly in dependent zones (Table 4). In contrast, noisy ventilation reduced overdistension mainly in nondependent regions in OLA.

Figure 4. Cumulated diffuse alveolar damage score (DAD) overall (left column) and in dependent (middle column) and nondependent (right column) lung regions after protective mechanical ventilation according the ARDS Network protocol (ARDSnet) and the open lung approach (OLA) with and without variable tidal volumes (noisy). Values are shown as medians, interquartiles, minima, and maxima. Mann-Whitney U Test was used to compare groups. Statistical significance was accepted at P < 0.05.

Download Figure | Download Powerpoint

[More] [Minimize]

In ARDSnet and OLA, noisy ventilation did not increase gene expression of amphiregulin, TNC, IL-6, IL-8, or TGF-β in dependent and nondependent lung regions (Table 5). Independent from the mode of ventilation, gene-expression of IL-8, and mechanical stress markers were increased in nondependent compared with dependent lung regions (IL-8, 0.88 [0.69–1.74] vs. 0.20 [0.07–0.86], P = 0.002; amphiregulin, 0.62 [0.36–1.43] vs. 0.56 [0.15–0.97], P = 0.008; TNC, 0.79 [0.58–1.72] vs. 0.32 [0.10–0.56], P < 0.001). The cytokine levels of IL-6 in plasma (see Fig. E8 in the online supplement), as well as of IL-6 and IL-8 in lung tissue (Tables E5 and E6, respectively) and in BAL (Table E7), were not influenced by noisy ventilation in ARDSnet and OLA groups, but tissue levels of IL-6 and IL-8 were higher in dependent than nondependent lung zones (P < 0.001).

Pa_O2/Fi_O2 was negatively correlated with the angular coefficients of the distribution of PBF along the caudal-cranial axis (r² = 0.47, P < 0.001) and positively correlated with the angular coefficients of the distribution of PBF along the peripheral-central axis (r² = 0.11, P = 0.045). Ers was negatively correlated with Pa_O2/Fi_O2 (r² = 0.26, P = 0.002) and positively correlated with the mean gene-expression of amphiregulin as well as of mean protein levels of IL-6 and IL-8 in lung tissue (r² = 0.13, P = 0.047; r² = 0.19, P = 0.012; r² = 0.26, P = 0.003). Pmean and Ppeak were correlated with the mean gene expression of TNC (r² = 0.14, P = 0.035 and r² = 0.12, P = 0.048, respectively).

Our data confirm previous observations suggesting beneficial effects of noisy ventilation to improve respiratory function. In a first report on natural noisy ventilation, Lefevre and colleagues (22) showed an improvement in oxygenation, respiratory system compliance, and lung water content through variation of Vt and f, as compared with conventional ventilation. However, significant differences in Pa_CO2, pH, and delivered Vt between ventilation groups, as well as the absence of PEEP, could have influenced their results. In a porcine oleic acid model of ALI, Boker and colleagues showed that natural noisy ventilation improved oxygenation but not respiratory system compliance when using low tidal volumes (13). In a similar model of ALI, Funk and colleagues reported improvements of oxygenation and respiratory system compliance when comparing natural noisy ventilation and OLA (23). In contrast, in a canine oleic acid model of ALI, Nam and colleagues could not show any beneficial effect of natural noisy ventilation (14). However, differently from other studies, those authors used higher Vt (approximately 15 ml/kg) and no PEEP. Moreover, the mortality of animals was high, indicating more severe lung injury. Our study differs from previous ones in the following respects: (1) Only Vt was modulated, whereas other authors (8, 11, 13) varied simultaneously f and Vt, maintaining a constant minute ventilation on a cycle-by-cycle basis. Therefore, our results suggest that Vt variability is the main determinant for improvement in respiratory function. (2) We used a stable surfactant depletion model of ALI, and all animals survived the observation period without need for vasoactive drugs, which are usually required during oleic acid injury (24). (3) f and Vt were strictly controlled, and the combination of PEEP and Fi_O2 were selected according to the ARDSnet recommendations. (4) In OLA, PEEP was selected to minimize Ers, contributing to keep the lung open over time. Thus, in our study, protective ventilatory strategies with and without lung recruitment were optimized according to current standards in mechanical ventilation.

Independent from the protective ventilation strategy we used, noisy ventilation improved arterial oxygenation and Ers at a lower mean Ppeak. Several mechanisms can explain such findings: (1) The occasional proportionally higher Vt occurring during noisy ventilation may recruit atelectactic zones (9, 11), (2) redistribution of PBF toward better aerated lung areas may improve ventilation/perfusion matching (25), (3) increased release of surfactant (26), (4) stochastic resonance behavior of the respiratory system (noise enhancement of an input signal, e.g., Vt [7]), and (5) enhanced respiratory sinus arrhythmia (27).

The improvement in Ers and Ppeak with ARDSnet and OLA suggest that recruitment of atelectactic zones played an important role. In ALI induced by oleic acid or surfactant depletion, improved arterial oxygenation and compliance reflect recruitment of atelectactic zones as assessed by computer tomography (28, 29).

Redistribution of PBF toward caudal and peripheral lung regions likely reflected better aeration/ventilation resulting from recruitment in those areas as shown by the correlation analysis. In oleic acid and surfactant depletion–induced ALI, Karmrodt and colleagues (30) have shown that aeration in caudal lung zones increased as a function of recruitment. Moreover, mainly in ARDSnet, noisy ventilation reduced Pmean and mean pulmonary arterial pressure, likely favoring redistribution of PBF.

It has been shown that mechanical ventilation with high Vt leads to ventilator associated lung injury (VALI) (31). When the global applied force on the lung parenchyma is excessive or the fibers near the diseased regions experience excessive mechanical stress, mechanical rupture and/or biological activation of inflammation may occur (32, 33), activating or worsening lung injury. In line with these claims, in our study, we found that increased gene expression was associated with higher Ppeak and Pmean, whereas decreased Ers was correlated with increased gene-expression of inflammatory mediators. Although much has been learned regarding the ability of pulmonary cells to sense and integrate information from mechanical distortion during monotonic controlled ventilation (34), little is known about the effects of variable mechanical ventilation.

To our knowledge, this is the first study showing that noisy ventilation is able to attenuate lung histological damage and to demonstrate that it does not increase gene expression or release of proinflammatory markers of lung injury. Although Funk and colleagues (23) reported that natural noisy or fractal ventilation did not worsen lung histological lung appearance compared with conventional monotonic ventilation, in the present study animals developed less interstitial edema, hemorrhage, and epithelial destruction, especially in the dependent lung regions, compared with ARDSnet. On the other hand, compared with OLA, noisy ventilation resulted in reduction of DAD, mainly due to decreased overdistension in nondependent regions.

Our data suggest that noisy ventilation can reduce VALI by different mechanisms, depending on the protective mechanical ventilation strategy it is used in combination with. In ARDSnet it seems likely that dependent lung zones were not fully recruited, leading to cyclic opening/closing of atelectactic areas and peripheral airways. This could explain our findings of prevalent histological damage in the dependent lung areas in this group. In this situation, noisy ventilation may recruit and stabilize the lungs at lower PEEP, thereby reducing injury. On the other hand, when OLA was used, the lungs were possibly fully recruited, with overdistension of nondependent lung regions. Under these conditions, the reduced mean Ppeak during noisy ventilation probably acted to limit overdistension in OLA, with less effect on other features of lung injury. Furthermore, there was no evidence that noisy ventilation increased gene expression of inflammation mediators or their release into BAL, plasma, or lung tissue, which is in agreement with other studies (13, 23).

Although not directly related to the use of noisy ventilation, it is worth noting that protein levels of IL-6 and IL-8 were higher in dependent than nondependent zones, while gene expression of inflammation markers showed the opposite pattern. There are two possible explanations. First, different time dynamics may have played a role. Dependent areas may have developed inflammation immediately after lavage, but because they developed atelectasis and/or edema, they were less ventilated. Accordingly, gene expression was more closely related to the effects of VALI in nondependent zones, which occurred later on the course of injury. Second, cytokines may have cumulated preferentially in dependent zones due to the gravity gradient.

It could be argued that mechanical stress could still be higher with noisy ventilation despite similar inflammatory activation. However, the analysis of gene expression of amphiregulin and TNC, which in contrast to IL-6, IL-8, and TGF-β are activated selectively by mechanotransduction (35), largely rules out this hypothesis.

Different mechanisms could explain the finding that noisy ventilation reduced VALI and did not influence mechanical stress: (1) Alveolar recruitment resulted in more homogenous distribution of ventilation and improved mechanical properties; (2) transpulmonary pressure was decreased as a result of lower Ers and Ppeak at comparable mean Vt; (3) the nonnormal distribution of the respiratory pattern during noisy ventilation led to an increased number of respiratory cycles with Vt <6 ml/kg (Fig. E5). However, Vt values that are too low can also be injurious (36); (4) monotonic and variable lung straining may have different impacts on VALI.

Our study has several limitations. First, the surfactant depletion model does not reproduce all features of the more complex human ALI/ARDS and is highly recruitable. Second, we did not directly assess regional lung aeration, although the improvement in functional parameters and redistribution of PBF suggests recruitment with noisy ventilation. Third, the short observational period of 6 hours does not allow extrapolation of our results to long-term mechanical ventilation. Fourth, it can be argued that other combinations of lower PEEP and Fi_O2 would probably lead to different results. However, the use of lower PEEP would have resulted in worse respiratory function and possibly increased injury due to cyclic alveolar collapse/reopening. On the other hand, higher PEEP does not match the recommendations of the ARDSnet and would have likely resulted in increased overdistension, as observed in the OLA group. Fifth, we used only one distribution of Vt and did not vary f. Thus, we cannot exclude that other Vt distribution patterns and/or f variation would lead to different results. However, our findings show that variation of Vt alone is enough to achieve beneficial effects.

In a surfactant depletion model of ALI, the use of random variable Vt improves respiratory function and reduces histological damage during mechanical ventilation according to the ARDSnet protocol and OLA without increasing lung inflammation and mechanical stress.