| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Department of Biomedical Engineering, Boston University; and 2CelluTraf Sci., Boston, Massachusetts
Submitted 20 September 2007 ; accepted in final form 11 March 2008
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
acute respiratory distress syndrome; lung injury; elastance; gas exchange
In 1996, Lefevre et al. (35) proposed an alternative approach to mechanical ventilation by introducing "biologic variability" of natural breathing to mechanical ventilation, whereby the breathing rate (f) and VT were varied on a cycle-by-cycle basis. They showed that incorporation of such variability into mechanical ventilation improved oxygenation and lung compliance in an oleic acid-induced porcine model of ALI. Subsequently, Suki et al. (53) proposed a mechanism to explain how addition of variability, which they called noise, to peak inspiratory pressure can improve arterial PO2 (PaO2). Several beneficial physiological effects of this "noisy ventilation," which is also termed "variable ventilation" (VV), have since been shown to occur in healthy and ARDS/ALI models in pigs (11, 40, 41), dogs (42), guinea pigs (7, 8), sheep (10), and, more recently, normal human subjects (12). In addition to the beneficial effects on respiratory function, VV has also been demonstrated to favorably influence surfactant secretion in healthy guinea pigs (8).
The mouse has been a popular and often preferred species in medical research because of the relative ease with which various models of lung diseases can be generated. Indeed, mice have increasingly been used in respiratory research, including the lung injury model (2, 5). The chest wall of the mouse is relatively compliant (31, 51), resembling that of preterm or newborn infants (45). Consequently, the alveoli of the mouse lung, even the uninjured lung, are prone to collapse during mechanical ventilation. It is thus conceivable that the mouse lung can easily develop VILI via atelectrauma.
Therefore, the purpose of the present study was to investigate the physiological effects of VV during short-term ventilation of healthy and injured mice. Our preliminary results suggested that the current form of VV used in larger animals (7, 8, 10) is unable to recruit the lung and keep it open. Hence, we developed a new VV (VVN) method and optimized its parameters to recruit the lung and keep it open in normal mice and mice with HCl-induced injury. We then tested the effects of VVN on mechanics, gas exchange, and IL-1β, an inflammatory cytokine known to accumulate in lung injury (21). The performance of VVN was systematically compared with that of the following ventilation modes: conventional ventilation (CV) with fixed VT and f, original VV (VVO) as developed by Arold et al. (8), and CV with added periodic large breaths (CVLB) with amplitudes that matched the largest VT in VVN. The latter ventilation mode was included to test the hypothesis that the beneficial effects of VV are due to the variability in VT as characterized by its distribution, and not just by the presence of a few large breaths.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
In a pilot study, we found that to prevent the progressive collapse of alveoli during ventilation of normal mice with CV, we had to deliver unexpectedly large VT to sufficiently recruit the lung. Briefly, several mice were ventilated with CV at 8 ml/kg superimposed on 3 cmH2O PEEP. Respiratory mechanics were measured, and respiratory elastance increased 6% in the first 5 min. A deep inspiration (DI) always returned the lung to baseline, characterized by a low elastance, suggesting that the increase was due to gradual lung collapse (2, 3, 5, 31). For determination of the smallest large VT that could recruit the lung, the animals were ventilated with CV for 1 min and given a large (e.g., 16 ml/kg) test VT, and elastance was measured to assess recruitment. The size of the test VT was then progressively increased to several times the baseline VT. We found that the smallest large VT that could return elastance to within 10% of its baseline was three to four times the baseline VT of 8 ml/kg.
The distribution of VT in VVO developed for guinea pigs (Fig. 1) was based on a uniform distribution of peak airway pressures, which was shown to result in an inverse power-law distribution of VT values over a relatively narrow range (8). Consequently, the VT distribution in VVO may not be able to provide adequate recruitment and/or keep the recruited regions of the mouse lung open. Thus, in order for VV to recruit the lung, larger VT values would have to be incorporated in the VV by extension of the tail of the VT distribution. However, if the shape of the distribution, a single power law as in VVO, and the mean VT (VTmean) were maintained, stretching of the tail to high VT values would also require an extended low-VT region that would not be able to sustain adequate ventilation of the animal. Therefore, the distribution of VT needs to be redesigned by stretching the tail while limiting its minimum value.
|
 | (1) |
is a parameter and p0 is the normalization constant. Using the shape defined by Eq. 1, one can conveniently control the minimum and the maximum, as well as the range, of "low"-VT values by adjustment of Vmin, Vp, and Vmax. The sequence of VT values from this distribution ventilates the animals mainly in the lower range of VT values while occasionally larger VT values are also delivered. The mean and standard deviation (SD) of the distribution are given by Eqs. 2 and 3
 | (2) |
 | (3) |
, determine the shape of the distribution, as well as VTmean and SD of the distribution. To compare the effects of the various ventilation protocols, we set VTmean to be the same as VT in the CV group chosen on the basis of the weight of the animal. Vmin was fixed at 0.7 VTmean to avoid too-low VT values during ventilation as well as to match Vmin in the VVO distribution. On the basis of preliminary computer simulations, we found that in order for a substantial increase in Vmax and constant VTmean and Vmin, VP and were constrained within a small range. Thus we had to choose VP = 0.9 VTmean, a value slightly smaller than VTmean, while was set to 5.1, which resulted in a much faster decrease of p(VT) in VVN than in VVO (Fig. 1). This latter feature of the VVN distribution allowed us to achieve much higher values of Vmax than in VVO while maintaining VTmean and Vmin. To determine the optimum Vmax, a group of mice was ventilated with VVN with several sequences of VT values with similar shape and identical mean but different Vmax. Animal Preparation
Male C57BL/6 mice (22–26 g body wt; Charles River Laboratories, Wilmington, MA) were used throughout the studies. The protocol was approved by the Animal Care and Use Committee of Boston University. The mice were anesthetized with pentobarbital sodium (70 mg/kg ip), tracheostomized with an 18-gauge metal cannula, and placed on a heated pad to maintain a constant body temperature (37°C) throughout the experiment. Extra doses of pentobarbital sodium (20 mg/kg) were administered every 20 min to keep the animal in a deeply anesthetized state. The tracheal cannula was later connected to the outlet of a small animal ventilator-oscillator system (flexiVent, SCIREQ, Montreal, PQ, Canada).
Acid Aspiration Lung Injury
A mouse model of lung injury was obtained by intratracheal induction of 0.1 M HCl (pH 1.25) in 1 µl/g increments, each separated with a bolus of air, for a total of 3 µl/g. Lung collapse was prevented by connecting the mice to the ventilator, immediate administration of two DIs, defined here as a ramp increase in volume to 1 ml in 4 s, and administration of two additional DIs at 5 and 10 min after treatment. The animals were then ventilated using a constant VT of 8 ml/kg for a stabilization period of 20 min at 3 cmH2O PEEP.
Optimization of the Parameters in VVN
As described by Eq. 1, the distribution of VT values in VVN consists of a flat region between Vmin and VP followed by a power-law decay up to Vmax. A group of normal mice (n = 4) and a group of HCl-injured mice (n = 2) were ventilated with CV (VT = 8 ml/kg and f = 240 breaths/min) at 3 cmH2O PEEP for 15 min followed by a series of 15-min sessions of VVN with identical Vmin, VP, , and VTmean but different Vmax. For the normal mice, the tested Vmax levels were 4, 3.75, 3.5, or 3.25 times VTmean of 8 ml/kg. For the injured mice, however, we found that it was sufficient to test lower Vmax levels, including 2.5, 2.25, 2, and 1.5 times VTmean. These sessions with different Vmax levels were randomized in order. After each set of trials, a DI was given to fully recruit the lung. Respiratory mechanics were measured at the beginning, before, and during (at 5, 10, and 15 min) each VVN session.
Comparison of the Effects of Ventilation Modes
Volume history was standardized by administration of two DIs to normal (n = 32) and injured (n = 32) mice. Mice were randomly selected for 60 min of CV, VVO, CVLB, or VVN (for each mode n = 8 each for normal and injured mice). PEEP was always set at 3 cmH2O. In a previous study, we found that the VT delivered by the flexiVent significantly depended on the load impedance (55). Hence, under all ventilation modes, VT was monitored by the flexiVent and, when necessary, manually adjusted to ensure that the proper VT was delivered to the animal independent of its mechanical impedance.
CV mode. Mice were ventilated with constant VT at 8 ml/kg and f of 240 breaths/min.
VVO mode. VTmean was 8 ml/kg (SD 1.6), and f was adjusted on a cycle-by-cycle basis to obtain a minute ventilation equal to that during CV. This VVO was based on the original design by Arold et al. (8). An example of the corresponding VT sequence is shown in Fig. 2A.
|
VVN mode.
With use of Eq. 2, VTmean was fixed to be 8 ml/kg and f was adjusted on a cycle-by-cycle basis to obtain a minute ventilation of 240 breaths/min. The parameters were as follows: Vmin = 0.7 VTmean, VP = 0.9 VTmean, and = 5.1. Vmax was set to 3.5 and 2.25 times VTmean for normal and injured mice, respectively, with the corresponding distributions shown in Fig. 1. An example of the VT sequence is shown in Fig. 2C. The SD values of VT were calculated from Eq. 3 (3 and 2 ml/kg for healthy and injured mice, respectively).
At the conclusion of the experiment, arterial blood from the carotid artery was collected and analyzed for arterial PO2 (PaO2), PCO2 (PaCO2), blood pH, and percent O2 saturation with an I-STAT blood gas analyzer (Abbott Laboratories, Abbott Park, IL). A lavage sample was also collected from each mouse at the conclusion of the ventilation session by instillation of 1 ml of warm saline via the tracheal cannula and slow retrieval of 
0.9 ml. The lavage sample was then centrifuged, and the cell-free supernatant was frozen at –20°C until further analysis. After the animals were killed, their lungs were perfused, removed, and homogenized in 2 ml of PBS (pH 7.2) and centrifuged. The supernatant was transferred to microcentrifuge tubes and stored at –20°C.
A group of C57BL/6 mice (n = 4, 20–22 g body wt) were anesthetized with pentobarbital sodium (70 mg/kg ip), tracheostomized, and immediately lavaged as described above. This group served as an unventilated control group. Another group of C57BL/6 mice (n = 4, 20–22 g body wt) was treated with HCl as described above and ventilated using a constant VT of 8 ml/kg for an additional 20 min at 3 cmH2O PEEP. The animals were lavaged immediately after completion of the injury induction steps. This group served as a baseline injury group without the additional 60-min ventilation session. The lavage samples and lung homogenate samples from both groups were centrifuged, and the supernatants were stored as described above.
Impedance Measurements
Respiratory mechanics were determined using the optimum ventilation waveform (OVW), which is a broadband waveform designed to estimate lung impedance via forced oscillations while simultaneously delivering tidal-like volume excursions (37). The energy in the OVW was distributed among five discrete frequencies from 2 to 31 Hz, which were selected to eliminate harmonic distortion and minimize cross talk according to a nonsum-nondifference principle (54). The phase angle for each frequency component was adjusted to minimize the peak-to-peak pressure of the oscillatory waveform. For each mouse, the peak-to-peak amplitude of the oscillatory waveform was adjusted to match the VT delivered during mechanical ventilation to minimize the disruption of ventilation during impedance measurements. Every 5 min, two complete 1-s cycles of OVWs were delivered. The piston displacement and pressure in the ventilator cylinder were recorded by the flexiVent, and the data were corrected for the effects of gas compression in the valves and the resistance of the cannula. Calculations were done offline using software developed in the laboratory. Fast Fourier transform was used to obtain the cross- and auto-power spectra of pressure and flow from overlapping data segments. The respiratory input impedance (Zrs) was calculated at each input frequency as the ratio of the cross-power spectrum of pressure and flow to the auto-power spectrum of flow.
Modeling
The Zrs spectra were fit by a model consisting of an airway compartment, characterized by the airway impedance (Zaw), and a tissue compartment, described by a tissue impedance (Zti), connected in series. Thus Zrs is calculated as Zaw + Zti. Zaw is further partitioned to airway resistance (Raw) and inertance (Iaw)
 | (4) |
is the circular frequency. Zti was modeled as a constant-phase impedance, which was introduced by Hantos et al. (24) and has been widely used to estimate Zti in animal models (22–24) and humans (33, 34). The constant-phase model characterizes Zti by a tissue damping coefficient (G) and a tissue elastance coefficient (H) as follows
 | (5) |
n = /0 is the normalized circular frequency with 0 = 1 rad/s (13). The exponent 
characterizes the frequency dependence of tissue resistance [Rti = G/(n)] and tissue elastance [Eti = H(n)1–]. All parameters were estimated using a global optimization algorithm, which minimized the root-mean-squared (RMS) error between data and model (16). Airway Pressure Measurement
A separate pressure transducer (World Precision Instruments, Sarasota, FL) attached to the tracheal cannula was used to monitor and record the airway pressure throughout the experiments. The mean peak airway pressure was calculated for every 5-min period.
Western Blot Analysis
To test whether the ventilation mode had an effect on lung injury, Western blot analysis was carried out for IL-1β as follows. Each lavage sample was mixed with the corresponding homogenized lung sample at a ratio of 1:1. The mixed samples for each lung were further diluted at a ratio of 1:5. Equal volumes from the mixed samples were processed by electrophoresis and Western blot analysis. A polyclonal anti-rabbit anti-IL-1β antibody (Chemicon, Temecula, CA) at a dilution of 1:1,000 in Tris-buffered saline + Tween 20 at room temperature was used to identify the presence of IL-1β in the samples. A chemiluminescent agent (Bio-Rad Laboratories, Hercules, CA) was applied to produce luminescence in proportion to the amount of protein on a photographic film. The relative intensity was quantified by densitometry after removal of the background intensity using software developed in the laboratory.
Statistical Analysis
The statistical differences among parameters under different conditions were tested using ANOVA (1-way ANOVA, 1-way repeated-measure ANOVA, 2-way ANOVA, and Tukey's pairwise comparison) and t-test (SigmaStat, SPSS, Chicago, IL). When the data were not normally distributed, one-way ANOVA on ranks was used to analyze statistical differences. P < 0.05 was used to establish statistical significance.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
During CV, H increased linearly; during VVN, H eventually reached a plateau level for certain values of Vmax. The plateau in H and the time required to reach the plateau strongly depended on the value of Vmax. For the normal mouse, within 15 min of ventilation, H reached a plateau when Vmax was 
3.5 times VTmean. For the injured mouse, H reached its plateau level when Vmax was 2.25 times VTmean. On the basis of similar results in four normal and two injured mice, the final Vmax was set to 3.5 and 2.25 times VTmean for normal and injured mice, respectively, for the rest of the study.
Effects of Ventilation Mode on Respiratory Mechanics
The respiratory mechanical parameters in normal mice (Fig. 3) show that DIs can successfully recruit the collapsed regions and standardize volume history, because at time 0, H had similar low values in all animals. The RMS error was similarly low, <0.04 cmH2O·s·ml–1, among all ventilation modes. Iaw was small and did not change consistently; hence, it is not reported. During the 60-min ventilation period, all other parameters (Raw, G, and H) for animals in the CV, VVO, and CVLB groups significantly increased with time (P < 0.001). For animals in the VVN group, only Raw and H depended significantly on time (P < 0.001). All parameters in the CV and VVO groups increased linearly throughout the experiment, whereas in the CVLB and VVN groups, they reached a plateau within 30 min of ventilation. Thus only the latter two ventilation modes were able to stabilize the mechanical condition of the lung within 60 min of ventilation.
|
The mechanical parameters during ventilation of HCl-injured mice are shown in Fig. 4. The RMS errors for the CV and VVO groups were higher than those of the normal mice (P < 0.05). Nevertheless, the model was still able to fit the Zrs spectra relatively well. All parameters (Raw, G, and H), including the RMS error for all ventilation groups, significantly increased with ventilation time (P < 0.001). At baseline, there was no difference in any parameters among the groups. On average, the baseline Raw was similar between the normal and injured groups, whereas the baseline G and H were 20% and 25% higher, respectively, than the baseline levels in the normal mice (P < 0.001). All respiratory mechanical parameters in the CV group increased linearly with time; during VVO, CVLB, and VVN, they reached a plateau level, indicating that variability in VT during VVO and VVN, as well as periodic large breaths in CVLB, were able stabilize lung mechanics in injured mice.
|
Effects of Ventilation Mode on Airway Pressures
The mean peak airway pressure and its percent increase averaged over 5-min intervals are shown in Fig. 5 for the normal and injured mice. For each mode of ventilation and in normal and injured mice, mean peak airway pressure was significantly affected by ventilation time (P < 0.001).
|
Mean peak airway pressure was only 8% higher in the injured groups than in the normal groups (10.4 vs. 9.3 cmH2O, P < 0.001). In period 1 (corresponding to 0–5 min), there was no significant difference among the groups. Between 25 and 30 min, period 6, mean peak airway pressure was significantly higher in the VVO group than in the CVLB and VVN groups (P < 0.05). However, percent mean peak airway pressure in the VVN group increased at a significantly slower rate than in the CV and VVO groups (P < 0.01 and P = 0.05, respectively). During the last period, 55–60 min, mean peak airway pressure and percent mean peak airway pressure significantly depended on the mode of ventilation (P < 0.001). Pairwise comparisons showed higher mean peak airway pressure in the CV and VVO groups than in the CVLB and VVN groups (P < 0.005). Also, percent mean peak airway pressure was significantly higher during CV than during any other ventilation mode (P < 0.05 for VVO and P < 0.001 for CVLB and VVN). Finally, from period 8 in the normal and injured mice, mean peak airway pressure was consistently lower in the VVN group than in any of the other groups and percent mean peak airway pressure nearly doubled during CVLB compared with VVN, a statistically highly significant difference (P < 0.001).
Effects of Ventilation Mode on Blood Gases
PaCO2 and PaO2 are shown in Fig. 6 for the normal and injured mice. For the normal mice, PaO2 and PaCO2 were significantly better in the CVLB and VVN groups than in the CV and VVO groups (P < 0.01). Moreover, the alveolar-arterial gradient was significantly lower in the VVN and CVLB groups than in the VVO group (Table 1; P < 0.02). Direct comparison of the alveolar-arterial gradient between the VVN and CVLB groups showed that the alveolar-arterial gradient in the latter group was statistically significantly higher (P < 0.05). For the injured mice, PaO2 was significantly higher in the CVLB and VVN groups than in the CV group (P < 0.05). However, PaCO2 and the alveolar-arterial gradient were not different among the various ventilation groups. Additionally, the ratio of PaO2 to fraction of inspired O2 (P/F) was significantly higher in the CVLB and VVN groups than in the CV group in normal and injured animals (P < 0.05) but significantly higher in the CVLB and VVN groups than in the VVO group only in normal mice (P < 0.02).
|
|
Western blots for IL-1β in the normal and injured mice are shown in Fig. 7. For statistical analysis, the amounts of IL-1β in the normal group were normalized with the mean value of the unventilated normal group (Fig. 8). In baseline injury samples, IL-1β expression was on average five times higher than in the unventilated normal group. One-way ANOVA showed no effect of ventilation mode on IL-1β levels in the lungs of normal mice. After injury, the dependence of IL-1β levels on ventilation mode was statistically significant (P < 0.01). Pairwise comparisons showed significantly higher IL-1β levels in the CV group than in the baseline injury, VVO, and VVN groups (P < 0.02, P < 0.05, and P < 0.01, respectively). However, interestingly, the IL-1β level in the CVLB group was not different from that in the CV group.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Design of a New VV Distribution
The study by Arold et al. (8) in normal guinea pigs showed little difference in H over 3 h of ventilation, despite the improvement in oxygenation and increase in surfactant release, suggesting a dissociation between these indexes of lung function. Our preliminary results in normal mice and in a mouse lavage model of lung injury showed that H progressively increased during CV and VVO, when variability in VT was 25% around the mean, and no difference between the two ventilation modes (data not shown). In normal mice, there was no difference in mechanical parameters during CV and VVO, even at 60% variation around VTmean. This led us to a new design, VVN, in which the tail of the distribution of VT was stretched to accommodate larger VT values (Fig. 1).
We optimized the parameters in VVN for maximum recruitment and stability of lung elastance using VVN with similar shape and VTmean but with different Vmax values. To reduce interanimal variability, we used the same animal to test the effect of Vmax on respiratory mechanics. Within 15 min of ventilation, H plateaued at Vmax >3.5 times VTmean for normal mice and Vmax >2.25 times VTmean for injured mice. These are surprisingly large VT values compared with the modest 40% variability in VT during VVO in guinea pigs (8). However, the alveoli of the mouse lung are smaller (36); hence, surface forces at the air-liquid interface are expected to be larger than in other animals. When the larger surface tension forces are combined with an extremely floppy chest wall in the supine mouse (32), the result is a strong tendency of the lung to collapse at low and even moderate PEEP levels. Another possibility is that the surface film itself stiffens over time. Horie and Hildebrandt (28) observed that dynamic lung compliance (inverse of H) of isolated air-filled lungs fell by 10–30% during 20 min of ventilation. Since the drop in compliance was significantly attenuated in saline-filled lungs, they concluded that the mechanism responsible for these observations was related to viscoelastic adaptation of the surface film. However, they also acknowledged that when a small VT was superimposed on a low end-expiratory lung volume, which also produced the largest drop of 30% in compliance, the results could be influenced by small airway closure. Allen et al. (2, 3, 5) found trends in H similar to those in our study during CV, and they interpreted their results in terms of derecruitment, which is also supported by recent in vivo microscopy of the periphery of the lung during ventilation (4, 49). Using computerized tomographic scanning, Gattinoni et al. (20) demonstrated that the pressure-volume (P-V) curve in human patients with ARDS is significantly affected by recruitment-derecruitment processes. Therefore, since a significant increase in surface tension would ultimately also lead to collapse, we will assume that any increase in H is primarily due to derecruitment of alveoli.
The injured regions in the ARDS/ALI lung are heterogeneously distributed, including almost-normal regions, mildly injured regions, and severely injured regions (20). Therefore, it is possible that the part of the injured lung that is collapsed cannot be recruited. This explains why the baseline H values were higher after the injury (Figs. 3 and 4). An intermediate VT in VVN could be sufficient to recruit the normal and perhaps part of the mildly injured regions of the lung. The largest VT values could perhaps also recruit most of the mildly injured regions. The presence of collapsed regions may result in slight overinflation of the neighboring open regions in the injured lung. Thus the Vmax necessary to achieve this may be smaller in the injured than in the normal lung. It is important to point out that the Vmax used in the normal mice was 3.5 times VTmean, which was generally between 0.2 and 0.22 ml. Therefore, Vmax was 0.7–0.8 ml, which is still lower than our DI, defined as 1-ml ramp increase in lung volume. In the injured mice, Vmax was 0.5–0.6 ml, which was well below the DI level.
Effects of Ventilation Mode in Normal Mice
Because of the significant stretch of the tail of the VT distribution in VVN, especially for the normal mice, it is crucial to ensure that the effects of VVN are due to the variability in VT and not simply a few large VT values. Therefore, we also implemented CVLB, in which CV was accompanied with large breaths that had the same large VT values delivered exactly at the same time points as in VVN (Fig. 2).
Respiratory mechanics. Raw and G barely increased over 60 min of ventilation during CVLB and VVN, whereas during CV and VVO, these parameters increased linearly (Fig. 3). H increased in all groups until 25 min. Although H continued to increase in the CV and VVO groups, suggesting no stable equilibrium, H during CVLB and VVN developed a plateau. In the plateau region (at 30–60 min), %H was significantly smaller during VVN than during CVLB (P < 0.001). Thus, although the large breaths in CVLB and VVN are needed to reach a stable dynamic equilibrium (the plateau in H) that is similar to a fixed point in a nonlinear dynamic system, the presence of intermediate VT values in VVN leads to a different dynamic equilibrium with a larger number of open alveoli. The lung behaves nonlinearly (38), especially after injury (30). Thus differences in the stable dynamic equilibrium are expected, since in a nonlinear system the output depends not only on the mean, which is matched between CVLB and VVN, but the higher-order moments of the input signal, in this case, the shape of the VT distribution. Since the application of a DI at the conclusion of the experiment was able to return all mechanical parameters to their baseline values, we conclude that changes in mechanics were due largely to derecruitment of alveoli.
The mean peak airway pressure showed a trend similar to H, but with smaller increases in percent mean peak airway pressure (5% and 16%) than in %H (20–80%). These results imply that dynamic elastance is a far more sensitive indicator of physiological changes than airway pressure. The reason is that airway pressure is influenced by viscous components of airways (Raw) and tissues (G), whereas H is selectively sensitive to tissue viscoelasticity and the number of open alveolar units. The smallest percent mean peak airway pressure was seen during VVN and was significantly smaller in the plateau region than during CVLB. This further strengthens the conclusion that variability leads to a different dynamic equilibrium, which brings about additional improvements in lung function over regularly delivered large breaths in CVLB.
Gas exchange and IL-1β. Blood gas measurements at the conclusion of the experiment showed a significant improvement in PaCO2, PaO2, and P/F in the VVN and CVLB groups compared with the CV and VVO groups. This improvement was due to the presence of more and better-aerated regions in the VVN and CVLB groups. Furthermore, the alveolar-arterial gradient during VVN was about half of that during CVLB (9.8 vs. 18.5), and this difference was statistically significant (Table 1).
Western blot analyses showed no significant difference in IL-1β in any of the ventilation groups compared with the unventilated group. This is consistent with the study of Allen et al. (5), who showed no significant difference in the mean tissue mRNA levels for IL-1β between conventional low VT with and without periodic recruitment over 2 h of ventilation. Thus, despite the increase in H and the decline in blood oxygenation, after 60 min of ventilation, there was no detectable sign of IL-1β-related inflammation in normal mice due to any of the ventilation modes.
Effects of Ventilation Mode in HCl-Injured Mice
Respiratory mechanics.
Our mouse model of acid aspiration adapted from Nagase et al. (43) was used to mimic the conditions that prevail during ARDS/ALI. Nagase et al. found an almost threefold increase in elastance compared with the control group. We also found similar increases in H immediately after the application of HCl. However, with subsequent DIs, H dropped to a value only 25% higher than in normal mice. Thus the threefold increase in elastance (43) may have been due to fluid blockage as a result of incomplete recruitment. In our HCl-induced injury, we allowed 30 min for the HCl to take effect. Our results showed that, despite the close proximity between normal and injury baseline levels, H increased significantly faster in the injured mice (Fig. 4 vs. Fig. 3), suggesting that some fundamental mechanical properties of the lung, most likely related to surfactant and subsequent development of edema, had already been altered as a result of injury.
In contrast to normal mice, the mechanical parameters in the VVO group during injury were different from those of the CV group, in that they reached a stable dynamic equilibrium marked by the plateau in %H (Fig. 4D). Since after HCl injury the level of variability in VVN needed to stabilize the lung was only 2.25 times VTmean, the variability in VVO appeared to be sufficient to achieve such a stable dynamic equilibrium in 60 min, albeit one with a much higher %H (Fig. 4D). Since HCl causes direct injury to the parenchyma on contact, it is conceivable that the largest breaths in VVO were sufficient to prevent collapse of parts of the lung compared with the single VT in CV but were not as efficient as the VVN method. Raw and %H were the smallest during VVN, and, in the plateau region, H and %H were significantly lower during VVN than during CVLB. Thus, similar to normal mice, VVN was able to reach a stable dynamic equilibrium with a higher recruited lung volume than CVLB. Furthermore, percent mean peak airway pressure nearly doubled during CVLB compared VVN, which was statistically highly significant (P < 0.001). This again may be a potentially important improvement over CVLB, especially if ventilation was continued beyond 60 min.
Gas exchange and IL-1β. PaO2 and P/F were significantly higher in the VVN and CVLB groups than in the CV group, but not in the VVO group. Also, there were no signs of any differences during CVLB or VVN. However, in contrast to the consistent differences in mechanical properties, the variation in PaO2 (Fig. 6) and alveolar-arterial gradient (Table 1) was quite high. Although the reason for this is unclear, it is possible that drawing a small amount of blood (100 µl) was inherently accompanied by large interanimal variability.
In patients with ARDS, elevated IL-1β was associated with poor survival rate (39). In the present study, Western blot analyses showed that IL-1β levels were significantly higher than baseline in the CV group (Fig. 8). This may be due to the overstretching of the aerated regions during CV. Interestingly, IL-1β levels were not elevated compared with baseline in the VVO group. Thus the variability in VVO and VVN was able to further minimize IL-1β-related injury compared with CV. In fact, although not significant, the level and variability of IL-1β were smallest in the VVN group. Finally, it is striking that IL-1β in the CVLB group was not different from that in the CV group. It is possible that the variability in VVO and VVN helps prevent development of unphysiological shear stresses, because the same alveolar regions do not undergo repeated opening and closure as in CV. However, the large breaths in CVLB alone are unable to eliminate such shear stresses when the lung is ventilated with CV between these large breaths, and consequently inflammation-related injury can develop.
Possible Mechanism
More than half of the breaths were smaller during VVN than during CVLB. However, these small breaths in VVN are mixed with the intermediate breaths. The amount of lung that can be recruited by these intermediate breaths depends on the size of the breath as well as the degree of collapse at the end of the previous expiration. Suki et al. (53) proposed a simple nonlinear mathematical model of the P-V curve to show how variability in peak pressures during VV can increase lung volume.
Briefly, during CV, airway pressure increases from end-expiratory pressure to end-inspiratory pressure, which is equal to peak airway pressure. The corresponding regional lung volume increases to a value V0. However, during VV, peak airway pressure varies on a cycle-by-cycle basis around its mean with a given SD. The corresponding regional recruited volume also varies. Because of the nonlinearity of the regional P-V curve, the recruited volume following a larger-than-average peak airway pressure will be significantly amplified; hence, the average recruited volume over a sufficiently large number of breaths will increase from V0 to V1, while the average peak airway pressure is maintained. Suki et al. (53) showed that, depending on the regional P-V curve, V1 can be >200% larger than V0. Recently, Bellardine Black et al. (9) used CT images to measure the regional P-V curves in a sheep saline lavage-induced ARDS model. They found a nearly linear P-V curve at the top of the lung, which was more aerated, but a strongly nonlinear P-V curve at the bottom of the lung, which was much less aerated. Thus the regional P-V curves can be very different in the injured lung, with highly nonlinear shapes reminiscent of the curve used in the modeling study of Suki et al.
Although we could not acquire regional P-V curves, we measured global P-V curves in several mice at three different time points (Fig. 9A). The "open" P-V curve was obtained during baseline, when the animal's lung was intact and fully recruited. The "stiff" P-V curve was obtained after 30 min of ventilation following induction of HCl injury. At this stage, a significant portion of the lung was collapsed. The "intermediate" P-V curve was obtained immediately after the stiff P-V curve, which also served to recruit the animal's lung. To determine the average recruited lung volumes, we selected a series of measured peak airway pressure values from the VVN and CVLB groups and used these time series as inputs to the three measured P-V curves. The corresponding volumes obtained from each P-V curve as a function of peak airway pressure are shown in Fig. 9, B and C, and the average recruited volume from each curve is summarized in Table 2. These simulations show that although the average peak airway pressure was lower in the VVN group (10.9 cmH2O) than in the CVLB group (11.6 cmH2O), the average recruited volume was consistently higher, by 8%, during VVN, in reasonable agreement with the lower %H (Figs. 3D and 4D). This is a consequence of the nonlinearity of the P-V curves. Therefore, the lung was more protected from overstretching, as well as repetitive closing and opening, because of many small VT values during VVN than during CVLB. Furthermore, the presence of the intermediate breaths in VVN also resulted in higher average recruited lung volume. These simulations provide some limited mechanistic insight as to how the variability in VVN makes it a better ventilation mode than CVLB and, hence, support the original mechanism of stochastic resonance (53).
|
|
|
We developed a new mode of VV, VVN, and compared its performance in mice with that of the constant low-VT ventilation, CV, the original design of VV, VVO, and a low-VT ventilation mode accompanied with large breaths, CVLB. We found that VVN and CVLB significantly outperformed CV, the current clinical method, as well as VVO. VVN and CVLB lead to a stable dynamic equilibrium in alveolar recruitment, but during VVN, this equilibrium corresponds to an improved mechanical lung condition. Taken together, our results imply that, during ventilation of normal and injured mice at a moderate PEEP, lung function significantly benefits from periodic large breaths, whereas variability provides an additional improvement in mechanics and airway pressures and protects the lung from heterogeneities and inflammation-related injuries. These results may have clinical implications for the ventilation of infants with premature lung conditions.
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. P. Arold, E. Bartolak-Suki, and B. Suki Variable stretch pattern enhances surfactant secretion in alveolar type II cells in culture Am J Physiol Lung Cell Mol Physiol, April 1, 2009; 296(4): L574 - L581. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. B. Massa, G. B. Allen, and J. H. T. Bates Modeling the dynamics of recruitment and derecruitment in mice with acute lung injury J Appl Physiol, December 1, 2008; 105(6): 1813 - 1821. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
