INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: acute respiratory distress syndrome; fractals; lung injury; mechanical ventilation

Mechanical ventilation is still the mainstay of therapy for acute respiratory distress syndrome (ARDS), a common disease with high morbidity and mortality. Different modes and techniques of ventilation have been suggested to improve patient outcome and reduce ventilator-associated lung injury (VALI). Low tidal volume (VT < 10 ml/kg, usually 4 to 7 ml/kg) ventilation with permissive hypercapnia has been shown to be superior to conventional ventilatory strategies with larger VT (10 to 15 ml/kg) (1). A recently published study from the ARDS Network (ARDSnet) showed improved outcome, particularly lower mortality with the use of low VT ventilation (2). The use of lower VT is presumed to reduce pulmonary epithelial disruption caused by excessive stretch, and to prevent the release of proinflammatory cytokines by avoiding high airway pressure (Paw) and subsequent overdistention of aerated areas of the lungs (3, 4). Despite these potential advantages, derecruitment remains a risk with low VT strategies (5).

In a series of animal studies, biologically variable ventilation (_bv), which features a fractal delivery pattern for both respiratory rate (f) and VT, has been shown to improve Pa_O2 in a porcine model of ARDS (6, 7). Evidence suggests that _bv can recruit atelectatic lung units as well as prevent the collapse of aerated units (8). Oleic acid (OA) lung injury is an excellent model to study derecruitment/recruitment (9) and correlates well to clinical examples of ARDS (5).

In this prospective randomized controlled trial using the same porcine model as in previous studies, we measured changes in Pa_O2, lung compliance, and proinflammatory cytokines to compare mechanical ventilation using the ARDSnet low VT ventilation protocol with and without biologic variability.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

See the online data supplement for a full description of Methods.

Experimental Preparation

Twenty-one pigs were studied following the Canadian Council on Animal Care Guidelines. The experimental preparation has been described before (7); differences are noted. After surgery, animals were ventilated with an Esprit ventilator (Respironics Inc., Vista, CA) with VT 12 ml/kg, f 20 breaths/min, fraction of inspired oxygen (FI_O2) 0.5, and positive end-expiratory pressure (PEEP) 4 cm H₂O. An intravenous loading dose and infusion of sodium thiopental/midazolam at 16/ 0.1 mg/kg/h maintained anesthesia.

All animals received a continuous infusion of dopamine (5 to 12 µg/kg/min) to maintain mean arterial pressure () > 50 mm Hg. The infusion was adjusted to the lowest possible dose.

OA Lung Injury

OA was infused until Pa_O2 decreased to  65 mm Hg for two consecutive measurements, 5 min apart. All animals were placed on an additional 4 cm H₂O PEEP (total 8 cm H₂O).

Ventilation Protocol after Lung Injury

VT was initially set at 6 ml/kg, and f increased to 30 breaths/min. Animals were randomly allocated to either control mode ventilation (_c) or _bvvariable f at the same mean f and VT and minute ventilation (_e). If _bv was chosen, a laptop computer activated the controller. Ventilation continued in either _c or _bv mode for the duration of the experiment. In both groups, f was increased to a maximum of 35 breaths/ min, and VT increased by 0.5 ml/kg increments to control Pa_CO2 if arterial pH was less than 7.20, according to the ARDSnet protocol (2).

Static Pulmonary Pressure-Volume (P-V) Curves

P-V curves were generated after the 5-h data collection. A randomized sequence of VT values (1 to 50 ml/kg) was delivered and plateau pressure measured 0.5 s after end inspiration.

Plasma and Tracheal Aspirate Cytokine Concentrations

Arterial and mixed venous blood was collected at baseline (pre- and post-OA), and then each hour for 5 h. Heparinized blood samples were spun at 400 g for 15 min and the serum supernatant collected. Tracheal aspirates were obtained at 5 h. Samples were frozen and kept at 80°C until analysis. Analyses in duplicate were done to determine the concentrations of tumor necrosis factor-alpha (TNF-), interleukin-6 (IL-6), IL-8, and IL-10 by sandwich ELISA. The analyses were done in a blinded fashion at the McDonald Research Laboratories at St. Paul's Hospital, University of British Columbia.

Wet:Dry Lung Weight Ratios

This methodology has been described previously (7).

Post hoc Analysis

Measurements of data files of Paw and flow have been previously described (7). Because of the variability in f and VT with _bv, a lengthy data collection was performed to more accurately assess f, VT, and Paw with _bv (range 298 to 1,010 breaths) at 1 and 5 h after OA. Assessment of VT for fractal behavior was based on relative dispersion analysis described by Glenny and coworkers (10).

Statistical Analysis

Data were analyzed by repeated-measures analysis of variance (ANOVA) as previously described (7). The relationship between proinflammatory cytokine concentrations (IL-8 and IL-10), mode of ventilation, and other relevant observations such as wet:dry weight ratios was examined by analysis of covariance (ANCOVA). Where significant, evaluation of slope and intercept between groups was compared (p  0.05 considered significant).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

After establishing the model and protocol, we conducted 21 experiments. Two animals died during OA infusion before randomization, and one animal died 120 min after randomization to Group _c. A fourth animal did not sustain an adequate lesion after 90 min of OA infusion and was excluded. Thus, data were analyzed on 17 experiments (n = 9 in Group _bv and n = 8 in Group _c).

There was no difference in body weight between groups 25 ± 2 kg in Group _bv and 26 ± 2 kg in Group _c. The volume of OA infused did not differ (0.16 ± 0.04 ml/kg and 0.16 ± 0.07 ml/kg in Group _bv and _c respectively). Dopamine dose was 5.8 ± 1.9 µg/kg/min in Group _bv and 7.5 ± 2.5 in Group _c (not significant [NS] between groups). The measured PEEP in Group _bv was 8.5 ± 0.2 cm H₂O and 8.6 ± 0.2 cm H₂O in Group _c. As well, both groups developed the same level of auto-PEEP (0.5 to 0.6 cm H₂O).

Hemodynamics

Data are shown in Table 1. The groups were not different at baseline, and similar changes in systemic, pulmonary, and filling pressures, pulmonary vascular resistance (PVR), and cardiac output () occurred immediately after OA lung injury. decreased in both groups after OA infusion but remained higher in Group _bv at all time periods after OA. Mean pulmonary artery pressure () increased with infusion of OA, and then improved slightly in the last 3 h, with no difference between the groups. Filling pressures, pulmonary artery occlusion pressure (Ppao), and central venous pressure (Pcv), were stable in both groups. Both groups showed a modest increase in core temperature after OA administration. Temperature remained stable in Group _bv, but increased significantly in Group _c (group × time interaction [G × T]; p < 0.0001). Four animals ventilated with _c required active cooling to maintain core temperature below 38° C. 

Respiratory Gases

Pa_CO2, mixed venous partial pressure oxygen (Pv_O2), arterial pH (pHa), dead space ventilation (VD/VT), and arteriovenous O₂ content difference data are shown in Table 2. Baseline values were similar between groups, and Pa_CO2 increased by the same magnitude in both groups at all time periods after OA. Pa_CO2 and pHa were in range in both groups according to the protocol (G × T; p = 0.988). Sodium bicarbonate was not needed to treat significant acidosis in either group. Pv_O2 was significantly reduced in both groups after OA. Pv_O2 in Group _bv showed significant recovery within the first hour, whereas it remained decreased with _c (G × T; p = 0.0038). The two groups showed no significant differences in the arteriovenous O₂ content difference and dead space ventilation.

Pa_O2, alveolar-arterial O₂ (A-a o₂) gradient, and shunt fraction  (S/T) are shown in Figures 1A, 1B, and 1C, respectively. Pa_O2 was slightly higher in Group _c both pre-OA and post-OA lung injury (p = 0.07 and 0.049, respectively). The two groups showed progressive improvement in Pa_O2 by the first hour, although the improvement was more marked and statistically significant with _bv. By 2 h post-OA, Pa_O2 was significantly higher with _bv. This improvement in Pa_O2 persisted for the remainder of the experiment. With _c, Pa_O2 stabilized at approximately 120 mm Hg (a mean increase in Pa_O2 of 37 mm Hg at 8 cm H₂O of PEEP at 5 h). With _bv, Pa_O2 showed a gradual improvement over time, such that at 5 h post-OA Pa_O2 was 173 ± 30 mm Hg (a mean increase in Pa_O2 of 95 mm Hg at 8 cm H₂O of PEEP). The G × T was p < 0.0001. Both groups had significant deterioration in shunt fraction after OA. At 4 cm H₂O, the shunt increased to 30 to 40% in both groups. With the additional 4 cm H₂O PEEP, shunt was significantly higher in Group _bv 18 ± 6% versus 13 ± 4%. Both groups showed progressive improvement in S/T with greater improvement in Group _bv (G × T; p = 0.0026). Similar changes were seen in the A-a O₂ gradient in the two groups. OA injury resulted in significant increases in A-a gradient in both animal groups. The gradient showed progressive recovery with _bv throughout the experiment and less so with _c (G × T; p < 0.0001).

View larger version (18K): [in this window] [in a new window]   View larger version (19K): [in this window] [in a new window]   View larger version (17K): [in this window] [in a new window]   Figure 1.   Indices of oxygenation between groups; bv (n = 9), c (n = 8). Changes from baseline (B/L) before administration of OA, then over the next 5 h. (A) PaO2 versus time for the two groups. PaO2 was significantly greater by 2 h post-OA with bv G × T; p < 0.0001. (B) A-a O2 gradient for the two groups. The gradient was significantly lower with bv by 2 h post-OA, G × T; p < 0.0001. (C) s/ t for the two groups. Significantly higher shunt fraction was seen after OA in Group bv. A crossover then occurred with s/ t lower by 3 h with bv, G × T; p = 0.0026.

Airway Pressure and Respiratory System Compliance (Crs)

Airway pressure (mean and mean peak), VT, and Crs are listed in Table 3. The values for Paw, peak Paw, and VT in Group _bv were obtained by collecting at least 10 min of data at 1 and 5 h to control for the variability in VT with _bv. There were no differences between groups at baseline for any of the variables. Both groups had significant increases in Paw and peak Paw after OA. Mean VT was similar in the two groups (p = 0.36) at all time periods. Significantly higher peak Paw was seen in Group _c at 5 h (p = 0.034; G × T).

Both groups had an equivalent reduction in Crs after OA of almost 50%, with a slight further reduction within the first hour after randomization with no significant difference in Crs within or between groups over time.

Serum and Tracheal Cytokines

No measurable concentrations of TNF- or IL-6 were obtained from either serum or tracheal aspirate at any time point. Serum IL-8 levels were very low. When measured in arterial blood in five animals, levels exceeded the limit of detection only twice at 12 pg/ml. In contrast, IL-8 concentrations were very high in tracheal aspirates, from approximately 1,600 to over 7,000 pg/ml. Although mean tracheal concentrations of IL-8 were not different between groups, ANCOVA revealed an inverse correlation between wet:dry weight ratio and IL-8 concentration in tracheal aspirate (p = 0.011). A difference in y-intercept was seen between modes of ventilation (p = 0.022). No difference was seen for slope (p = 0.229). The equation defining the relationship between IL-8 concentrations and wet:dry weight ratio for _c was [IL-8] = 1,004(ww:dw) + 15,340. For _bv the equation was [IL-8] = 1,815(ww:dw) + 19,560 (Figure 2). A right shift of the relationship between IL-8 and wet:dry weight ratio was seen with _c; for instance, the IL-8 concentration at a wet:dry weight ratio of 10 is calculated to be 5,300 pg/ml with _c and 1,410 pg/ml with _bv, a 3.75-fold difference. IL-10 serum levels were detected in most experiments, but did not differ between groups. ANCOVA failed to reveal a relationship between IL-10 concentrations in tracheal aspirate and wet:dry weight ratios (p = 0.203), although the slopes again showed an inverse relationship. Correlations were not seen between IL-8 or IL-10 tracheal aspirates and temperature, although the ANCOVA temperature × group effect for IL-8 was p = 0.081. 

View larger version (12K): [in this window] [in a new window]   Figure 2.   Relationship between IL-8 concentration in tracheal aspirates and wet:dry weight (ww:dw) ratios between groups. ANCOVA revealed an inverse correlation between IL-8 concentration in the tracheal aspirates and ww:dw ratios (p = 0.011). Concentrations of IL-8 were right-shifted in Group c as compared with that seen in Group bv.

P-V Curves and Mode of Ventilation

A representative example of a static P-V curve for _bv and _c is shown in Figure 3. By way of illustration, the ranges of delivered VT with _bv or _c are shown on the static P-V curves in Figures 3A and 3B, respectively.

View larger version (11K): [in this window] [in a new window]   View larger version (11K): [in this window] [in a new window]   Figure 3.   P-V curves for the two groups. Quasi-static curves were generated at the end of the experiment as described in METHODS. In both groups, ventilation was occurring in the linear portion of the P-V curve. (A) The variation of VT in Group bv is shown by the large open squares. The middle square is the mean value of VT (210 ml); the upper and lower squares are the highest (394 ml) and lowest (107 ml) VT seen in this experiment and are shown for illustrative purposes. (B) The variation in VT in Group c is shown by the large open squares, as in (A). The middle square is the mean value of VT (219 ml); the upper and lower squares are the highest (221 ml) and the lowest (214 ml) VT seen in this experiment; shown for illustrative purposes. A much greater variation in VT is seen with bv.

VT Variability and Power Law Analysis

The complete variability file of VT from a representative experiment is shown in Figure 4. VT changes were a consequence of alterations in f used to program the ventilator. The VT and f were inversely related as their product remained constant. A power law analysis of the percent relative dispersion (RD% = standard deviation/mean × 100) versus number of lumped averages of VT (V) from the first 256 separate measures of VT (V₀) in Figure 4 is shown in Figure 5. As an example, a V/V₀ equal to 128 means that the first 128 independent measures of VT were averaged and the second 128 measures of VT were averaged to give two values for VT over the observation period (see online data supplement for a fuller description of methodology). These two values were subsequently averaged to give a mean and standard deviation to calculate the smallest RD%. When plotted in log-log space, the RD% decreased in a power law fashion as the VT values were lumped into data sets of VT from 1, 2, 4, ...128 in size before averaging. An inverse power law is demonstrated such that RD%  v/v₀ with  = 0.39, with correlation coefficient (R²) = 0.98 confirming fractal behavior. The fractal dimension (D) is defined as 1-slope = 1.39. 

View larger version (25K): [in this window] [in a new window]   Figure 4.   The complete data set for VT delivered in an experiment using bv. There were 376 breaths in this file.

View larger version (9K): [in this window] [in a new window]   Figure 5.   Log-log power law analysis of the VT data from the first 256 data points in Figure 4 based on the relative dispersion technique (10). An inverse power law was demonstrated for these data fitting an equation y = x with  = 0.39. The fractal dimension (D) is 1 - slope or 1.39 over 2.1 decades (log 128) with R2 = 0.98 (see online data supplement for a fuller description). A dimension of 1.0 defines a completely homogeneous data set, and 1.5 defines a completely random data set. In between these two boundary conditions the data has fractal characteristics.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The use of lower VT to ventilate patients with acute lung injury has been advocated to reduce lung stretch and the release of inflammatory mediators, the latter considered to contribute to VALI (2). The risk of lower VT ventilation in injured lungs is respiratory acidosis and decreasing arterial oxygenation. We have shown that the addition of a fractal signal, which restores breath-to-breath variability to a low VT ventilation strategy, improves arterial oxygenation, shunt fraction, and A-a O₂ gradient at lower peak Paw in a porcine model of ARDS without affecting adversely or Pa_CO2. Also, was better maintained, associated with greater Pv_O2, suggesting improved hemodynamic stability.

The details of the computer technology that can incorporate a fractal signal into the output of a conventional volume cycled ventilator have been published previously (6). This study attempted to ventilate animals in the _c group as closely as possible to the protocol used in the low VT group in the ARDSnet trial (2). Initial settings were VT = 6 ml/kg at f of 30 breaths/min with PEEP of 8 cm H₂O. Increases in f and VT were required in both groups to control for acidosis (pH < 7.20), but no differences were seen between groups for either parameter. Breathing frequency was increased to 35 breaths/ min and VT increased to 7.2 and 7.5 ml/kg in Group _bv and _c, respectively. By 5 h, neither Paw nor Crs differed between groups. In the ARDSnet trial, no difference was seen in Crs using VT of either 6 or 12 ml/kg; thus, it is not surprising that Crs differences would not be seen for animals ventilated at the same mean VT.

We were able to maintain acceptable levels of both ventilation and oxygenation using a conventional low VT strategy in this ARDS model. However, the higher O₂ tensions seen with _bv may mitigate against the deterioration in arterial oxygenation that can occur with the use of low VT over longer time periods or may permit the use of even lower levels of FI_O2 or PEEP to maintain acceptable levels of arterial oxygenation without aggravating other determinants of VALI.

The superior oxygenation with this study is similar to previously published work using the same ARDS model and higher VT with or without 10 cm H₂O PEEP (7). The animals in the current study sustained comparable degrees of lung injury with similar volume of OA infused, wet:dry lung weight ratios, postinjury Pa_O2, postinjury Crs, requirement for inotropic support, and death rate during the study. The improvements in oxygenation, shunt fraction, and A-a O₂ gradients using lower VT and PEEP of 8 cm H₂O are remarkably similar to those obtained using VT of 12 to 15 ml/kg in the earlier studies. Higher VT was associated with a modestly greater Pa_O2 at 4 h (200 mm Hg as compared with 175 mm Hg in the present study). These similarities attest to the robustness of _bv in improving systemic oxygenation in severely injured lungs irrespective of VT chosen, and confirm the lack of significant benefit to ventilation with higher VT. The advantage of _bv at lower VT was the associated reduction in both from 14 to 12 cm H₂O and mean peak Paw from 40 to 21 cm H₂O.

After OA injury, gas exchange and S/T were modestly worse in the _bv group (Pa_O2: 78 mm Hg versus 96 mm Hg; S/T: 13% versus 8%). Subsequently, a crossover occurred for both of these variables with significantly greater Pa_O2 by 2 h with _bv and lower shunt fraction by 3 h. After OA injury, both groups had a deterioration in Pv_O2. Over the following 5 h, Pv_O2 was significantly lower with _c. The work of Sandoval and colleagues (11) indicates that with this difference in Pv_O2 alone, shunt fraction would be calculated to be 25 to 30% greater with _bv, as higher Pv_O2 decreases the denominator of the shunt equation. However, calculated _bv shunt was 30% lower than in animals ventilated with _c (Figure 1C), indicating that the numerator of the equation decreased to an even greater extent. This suggests better matching of alveolar ventilation to cardiac output (_A/) with _bv supported by the associated increase in Pa_O2 and lower A-a O₂ gradients seen.

The improved results seen here and in other studies from our group have not been confirmed in a canine study by Nam and colleagues (12). We have commented on their study, outlining the differences (13). An additional difference relates to how Crs was determined. We continued to ventilate with _bv during compliance measurements. In contrast, Nam and coworkers (12) returned to _c to measure Crs. We calculate from their methodology that for 10 to 17% of the 4-h study period animals did not receive variable ventilation. Pelosi and coworkers (9) have recently shown the propensity for derecruitment in OA lung injury models. Thus, it is possible that derecruitment occurred during return to _c when measuring Crs.

Examination of Figures 3A and 3B demonstrates that ventilation in the present study occurs on the linear portion of the P-V curveabove the lower inflection point and well below the upper inflection point with both _c and _bv. Even the uppermost large VT with _bv is within the plateau pressure of 30 cm H₂O in this model. The advantage of _bv is evident with ventilation over a broader range of airway pressures along the maximal alveolar recruitment phase of the P-V curve without an overall increase in .

Very high concentrations of the specific proinflammatory cytokines, IL-8 and IL-10, were seen in the tracheal aspirates at 5 h in both groups, whereas levels of TNF- and IL-6 were undetectable. An inverse correlation between concentrations of IL-8 and IL-10 and wet:dry weight ratios was present, likely due to dilutional effects from increasing edema fluid with increasing lung water. Despite relatively low power owing to the small number of samples analyzed, when dilutional effects were controlled for in this manner, significantly greater IL-8 concentrations were seen with _c as assessed by ANCOVA. This was manifest by a significant rightward shift of the curve, such that similar wet:dry weight ratios were correlated to more than 3 times higher concentrations of IL-8 in the _c group. A similar effect was not evident for IL-10.

IL-8 is considered to be the major neutrophil chemoattractant cytokine in lung diseases such as ARDS (14). Activated neutrophils play a major role in the pathogenesis of ARDS. Unfavorable outcome is associated with an initial exaggerated pulmonary inflammatory response that persists unabated over time. Sustained high levels of serum IL-8 are also associated with poor outcome (15). We did not observe high serum levels of IL-8. Others have demonstrated increased IL-8 concentrations in pulmonary edema fluid of patients with ARDS from sepsis and low levels in plasma (16). The high concentrations of IL-8 in tracheal aspirates in association with low plasma levels suggest that the lung was the primary source of IL-8 in patients with ARDS (17). In our study, the results suggest that IL-8 concentrations may be decreased by the mode of mechanical ventilation, independent of a low VT strategy.

IL-10 markedly inhibits lymphocytic and phagocytic function, essential for an adequate immune response to invading microbes (18). Circulating IL-10 levels are increased in some clinical studies of ARDS, but do not predict the development of the syndrome in at-risk patients (19). However, initial IL-10 concentrations have been shown to be significantly higher in patients who died as compared with survivors (20).

The precise role of specific proinflammatory mediators in the development of VALI is unclear. A recent study showed that preexisting lung damage induced by surfactant depletion and exposure to hyperoxia may also contribute to cytokine production independent of mechanical stress (21).

The complete file of VT variation for a single experiment is shown in Figure 4. There are 376 measurements in this file. Figure 5 shows the power law analysis for the first 256 of these observationsthe largest power of 2 for this data set (see Glenny and coworkers [10] and the online data supplement for an explanation). The slope of the relationship is 0.39; an inverse power law yielding a fractal dimension of 1.39. Using this analysis method, a slope of 0 indicates completely homogeneous VT and a slope of 0.5 indicates that VT is random. A slope between these two extremes indicates fractal behavior if the curve fit is good (R² = 0.98 in this instance). Using a different analysis technique, we have previously shown that f has such characteristics. Fractal behavior defines many respiratory patterns as outlined by Barabási, Frey, and colleagues (22, 23) and Que and colleagues (24).

Why Could Programming with Fractal Sequences Improve Mechanical Ventilation?

Suki and coworkers (25) have shown that the noisy end-inspiratory pressure signal seen with _bv can account for improved gas exchange. Recruitment of collapsed regions results in a greater than anticipated effect when delivery pressure exhibits biologic noise. Previously, it was suspected that alveoli had been recruited if distending pressure was maintained above the lower inflection point. Recent work suggests that recruitment occurs throughout the linear portion of the P-V curve (26, 27). Therefore, nonlinear pressure profiles seen with alveolar opening can be expected to occur all along the P-V curve (25). Recruitment has also been demonstrated to follow power law or fractal characteristics when measuring the relative jumps in terminal airway resistance and the time interval between jumps (28). A noisy signal that has fractal characteristics may be optimal to recruit alveoli under such circumstances, because rare eventsboth high and low end-inspiratory pressuresare more frequent with a power law distribution than in a Gaussian distribution (28). With fractal ventilation, maximal recruitment could occur with large VT, but Paw would not be increased over time because large VT values are offset by small VT breaths with their low end-inspiratory pressure especially so if VT values are on the linear portion of the P-V curve. Thus, a fractal signal can potentially maximize recruitment without an increase in either Paw or peak Paw. In fact, in this study, peak Paw with _bv was significantly lower than that seen with conventional _c. In contrast, increases in Paw would occur with various sigh protocols (29, 30). We have previously demonstrated that sighs programmed at the same time and magnitude as the large VT values with _bv did not improve gas exchange (31). In addition, Ingenito and colleagues (32) have recently shown that noisy ventilation increases surfactant production. Thus, alveoli opened through noisy ventilation would tend to remain open because of increased surfactant levels. Finally, the relative increase in alveolar area recruited with a noisy signal is potentially great because of the extremely efficient fractal packing of the lung (33).

In other work on biologically variable life support, we have shown that blood flow delivered by roller pump in a fractal manner can improve cerebral oxygenation during rewarming from hypothermic cardiopulmonary bypass (34, 35) and prevent deterioration in diastolic elastance after administration of cardioplegia solution (36). The presumed mechanismsupported by mathematical modelingsuggests greater microvascular flow with fractal perfusion. Analogous to the current study, the only difference between conventional and biologically variable management was the addition of a fractal input signal to the standard roller pump.

In conclusion, this animal study shows improved oxygenation, lower shunt fraction, lower peak Paw, and decreased IL-8 concentrations in tracheal aspirate when _bv is added to the ARDSnet protocol featuring a low VT strategy. Return to fractal transmission with _bv offers the possibility of additional improvement to mechanical ventilation over that seen with conventional low VT strategies in patients with ARDS.