INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recruitment of atelectatic lung units and maintenance of alveolar patency is an integral goal of mechanical ventilation. We have recently developed a new mode of mechanical ventilation called biologically variable ventilation (bv). This computer-controlled ventilator mimics the normal spectrum of breathing by incorporating breath-to-breath variability in respiratory rate (f) and tidal volume (VT). Using bv we have demonstrated improved arterial oxygenation (Pa_O2) without an increase in mean airway pressure () in a porcine model of severe lung injury (1). Suki and colleagues (2) postulate that bv improves Pa_O2 owing to recruitment of collapsed alveoli, which open in bursts or avalanches (3). They speculate that bv is an example of stochastic resonancethe addition of noise to an input signal (variable peak airway pressure [Ppaw]) to amplify output (Pa_O2) in a nonlinear system (4). With the noisy input signal seen with bv, the volume gained at higher Ppaw greatly exceeded the volume lost at lower pressures over time, the net result being improved oxygenation without an increase in airway pressure ().

To test the hypothesis that bv enhances recruitment of collapsed alveoli, we developed a porcine model of stable unilateral lung collapse and compared lung reinflation over 5 h using bv versus conventional control mode ventilation (c) at similar minute ventilation.

We also examined if the addition of sigh breaths to c was effective in recruiting collapsed alveoli. We programmed the sigh breaths to occur at the same interval with the same delivered volume as the largest breaths with bv. If sighs programmed in this manner were as efficacious as bv, then the more complicated variability programmed with bv would be unnecessary.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The Committee for Animal Experimentation at the University of Manitoba approved the study. When depth of anesthesia was adequate (isoflurane 1.5 minimal alveolar concentration [MAC] in 100% O₂), a tracheostomy was done and a double-lumen endotracheal tube was placed in the airway. Correct positioning was confirmed by fibreoptic bronchoscope. Mechanical ventilation by c was instituted with an Ohio 7000 anesthesia ventilator (Ohio Instruments, Madison, WI) with f approximately 15 breaths/min and minute ventilation adjusted to maintain the end-tidal CO₂ at approximately 35 mm Hg. Catheters were placed for blood sampling and pressure measurements. Airway pressures and volumes were measured by pneumotachograph (Hans Rudolph, Kansas City, MO). After baseline measurements, the right side of the double-lumen endotracheal tube was opened to air to allow the right lung to collapse. A minithoracotomy (pleural opening 2.5 cm) permitted complete collapse and observation of the lung. The lung remained collapsed for 1 h. At this point, the double-lumen tube was removed and replaced with a cuffed tracheostomy tube for reexpansion of the right lung. Animals were randomly allocated to continue with c or switched to bv. Conceptually, this is equivalent to flipping a switch: on = biological variability; off = no variability. Before the switch was flipped in those animals receiving bv, f and measured tidal volume (VT_m) were the same in each group. The delivered minute ventilation remained unchanged from baseline and continued with either mode for the next 5 h. Blood gases and O₂ contents (arterial and mixed venous) and expired gas samples were measured (Radiometer ABL3 and Radiometer OSM3, Copenhagen NV, Denmark). Static respiratory system compliance (Crs) was measured by transiently clamping the expiratory limb of the ventilatory circuit at end inspiration. Calculated indices included shunt fraction (S/T), and Crs (V/P).

Computer-controlled Ventilation

The computer-controller and software for the ventilator have been previously described (1). Data for the modulation file were obtained from an awake, spontaneously breathing animal. The variability file used with bv is shown in Figure 1.

View larger version (31K): [in this window] [in a new window]   Figure 1.   Modulation data file used to control f with bv. Instantaneous f (in breaths/min) versus time (s). There were 369 different f values over 24.7 min before the file looped to repeat itself. The mean programmed f was 15 breaths/min ( thick line).

Computer-controlled Sigh Ventilation

Examination of Figure 1 reveals four instantaneous breaths below 10 breaths/min. To deliver computer-controlled sighs, a modulation file was written so that all instantaneous breaths were set to 15 breaths/ min except at the four time periods where the instantaneous breaths were < 10 breaths/min. At these times the instantaneous breath rate was programmed as shown in Figure 1. Thus, each of these low breath rates resulted in delivery of a large VT or sigh breath at the same interval and magnitude as those programmed for bv. In addition, the duration of the modulation file was the same as with bv before it looped to repeat itself. Experiments were done post hoc in this group. Every attempt was made to ensure that this group did not differ at baseline or with one-lung ventilation from the other two groups.

Post hoc Analysis

Data acquisition files of airway pressure and flows were processed to integrate the area under the pressure-time and flow-time curves to give and volume. Mean Ppaw was also calculated.

Statistical Analysis

Data were analyzed by repeated measures analysis of variance (ANOVA). A p value  0.05 was considered significant for group × time interactions or differences between groups. Comparisons between and within groups were based on generated least-squares means matrices with Bonferroni's correction applied when multiple comparisons were made. Data are presented as mean ± SD unless otherwise noted.

Inverse power law analysis was done as follows: mean instantaneous f was determined, then each instantaneous f was subtracted from mean f, this value was squared, then log transformed. These data were partitioned into incremental bins of equal size to determine their frequency distribution. The probability of each frequency was determined by Ni/N where Ni = number of observations in a given frequency bin and N = total number of observations. A log transform of the probability distribution was derived. The log probability distribution versus log f variation was plotted. The confidence interval and correlation coefficient were derived by regression analysis.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Data were analyzed on 22 experiments (n = 7 with bv and c and n = 8 with s). Measured mean f was the same in all groups at baseline values and during the period of one-lung ventilation. At these times, there were only minor differences seen for any measured parameter between groups (Table 1). Importantly, at baseline there were no differences in Pa_O2, Pa_CO2, S/T, Ppaw, VT or Crs between groups. With one-lung ventilation, the Pa_O2 decreased significantly with approximately a 4-fold increase in S/T. The Ppaw nearly doubled. The Crs decreased to approximately 40% of baseline values. After 60 min both lungs were again ventilated. In the group receiving bv and s, the computer-controller was activated. In these two groups, the delivered minute ventilation was not changed from its baseline settings with c. Mean f was scaled to 15 breaths/min, the same mean rate as in the control groupmeasured rate with bv remained unchanged from baseline at 13.9 breaths/min; coefficient of variation 18%. At 5 h, Pa_O2 was significantly higher with bv than in the other two groups [group × time interaction (G × T); p < 0.0001]). Carbon dioxide clearance was superior with bv such that Pa_CO2 was significantly lower at 5 h than for the other two groups (G × T; p < 0.0001). Mean Ppaw was lower at 5 h with bv than with c or s (G × T; p < 0.0001). VT was significantly greater with bv at 5 h (G × T; p < 0.0001). Crs was much greater with bv by 5 h than in the other two groups (G × T; p < 0.0001).

Figure 2 shows the changes in Pa_O2 over time for each group from Time 0 (end one-lung ventilation) to experiment completion at 5 h. The Pa_O2 increases more rapidly with bv and reaches a higher asymptote (p < 0.0001 group × time interaction) than in the other two groups.

View larger version (19K): [in this window] [in a new window]   Figure 2.   PaO2 versus time for the 3 experimental groups. Group × time interaction by ANOVA, p < 0.0001. With bv, Pa O2 increases more quickly and approaches a higher asymptote.

At baseline, during c, minute ventilation, the product of f and VT_m (f × VT_m) was not different for bv versus c (186 ± 11 ml/kg with bv and 187 ± 10 ml/kg with c). With one-lung anesthesia with both groups on c, measured VT decreased as airway pressure increased. This is compatible with the volume lost because of the compliance of the anesthesia circuit (circuit compression volume; 2 ml/cm H₂O/L in this case). With switch to bv, the measured minute ventilation product increased despite unchanged ventilator settings. The f × VT_m product increased solely as a consequence of an increase in measured VT. Measured f remains unchanged (> 75 instantaneous breath intervals and VT measured per observation with bv). Measured VT was approximately 11% greater with bv at 5 h compared with c, such that Vbv = Vc + Vc with  = 0.11.

In Figure 3 we have plotted log probability distribution versus log variability in f. An inverse power law frequency distribution was seen with slope 1.6 ± 0.3. 

View larger version (12K): [in this window] [in a new window]   Figure 3.   Log-log plot of probability distribution versus f variability used to program the biologically variable ventilator. A y  1/fa plot is obtained with a = 1.6 ± 0.3. The solid box shows the probability distribution with c. The probability is 1 with no f variability. The f variability is shown as 1 because the log of zero variability is undefined. The difference in the behavior of the two ventilatory modes is clear.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this simple experimental model of reversible atelectasis deflation of one lung in the pigbv resulted in more rapid and greater recruitment of collapsed lung. Our results are compatible with the theoretical model that "noisy" Ppaw would better recruit atelectatic lung units (2). With bv, effective VT increased 11% under these experimental conditions. Suki and coworkers showed the probability functions for alveolar recruitment, which occur in avalanches (3, 5). These functions follow inverse power law frequency distributions (y  1/f^a; examples of noise in natural phenomena) (6), with slopes of 1.1 to 2.5. We have plotted the probability distribution of the variability in f used with bv (Figure 3). This function also follows an inverse power law frequency distribution with a negative slope of 1.6.

Suki and coworkers suggest that "both the magnitude and timing of pressure excursion applied at the airway entrance during artificial ventilation may be critical in triggering the avalanche process of alveolar recruitment" (3). As such, variable f and VT with bv presumably facilitates this avalanche process. The increase in the f × VT_m product is the fundamental change with bv. The increase in measured VT is compatible with the improved Crs seen at 5 h with bv. Gunnarsson and coworkers have shown a positive correlation between shunt fraction and the area of atelectasis as measured by computed tomography during anesthesia (shunt = 1.6 × atelectatic area + 1.7) (7). If such results apply to this experiment, bv would have resulted in a net 3% greater atelectatic area recruited than with c. Although of small magnitude, this change would represent 3 to 4 times greater area if such lung were aerated (8) (in this instance a difference in area of 9 to 12%). At 5 h, S/T was returned to baseline with bv but remained elevated at almost 160% of normal with c and 330% of normal with s, suggesting near complete recruitment of atelectatic lung with "noisy" ventilation. The lower Pa_CO2 in the bv group also suggests better matching of ventilation to perfusion (A/).

Examination of Figure 1 reveals 4 low-frequency rates in the modulation file/cycle with instantaneous f approximately 7/min. As mean V T in this group was 14.7 ml/kg, the calculated volume of the lowest breath rate would be (13.9/7 × 14.7) ml/ kg or 29 ml/kg. Concern was raised that the large VT with the low-frequency breaths represented sighs and that volume recruitment by this mechanism alone could account for the documented improvement in gas exchange during bv.

To address this issue, we examined eight additional animals (s group), ventilated in the same manner as the original c group, but submitted to programmed "sighs" of identical magnitude and frequency as the low-frequency, large VT breaths in the bv group. Thus, this group of animals was exposed to the same inflation stress as the bv group but without the biologically variable noise of the bv group. Despite essentially identical measurements of gas exchange and respiratory mechanics at baseline values and during one-lung ventilation, s was not associated with any improvement in these parameters as seen with bv after 5 h of mechanical ventilation and conferred no advantage compared with c alone (see Table 1). Pa_O2 and shunt fraction were, in fact, worse with this ventilatory mode, with the proviso that this was a post hoc comparison.

Although the issue of sighs as an effective volume recruitment maneuver during prolonged ventilation is controversial, examination of the literature suggests that sighs are only effective when administered as a sustained inflation with high pressures under specific circumstances. Sighs of the magnitude seen during bv have not been shown to produce beneficial effects on Crs or gas exchange (9). Balsys and coworkers have shown that even larger sighs of 46 ml/kg resulted in unsustained increases in compliance and insignificant increases in Pa_O2 in healthy lungs in mechanically ventilated dogs (10). Bond and coworkers demonstrated that sustained inflation increased respiratory compliance only during conventional mechanical ventilation using low VT (7 ml/kg) and low end-expiratory pressure. Sustained inflation was of no benefit during conventional ventilation with high VT (14 to 17 ml/kg)the range of the current studyor with low VT and high end-expiratory pressure (11). Tusman and coworkers showed that a recruitment strategy (VT up to 18 ml/kg coupled with increasing levels of positive end-expiratory pressure [PEEP] up to 15 cm H₂O) can improve arterial oxygenation after 40 min of anesthesia (12). Pelosi and coworkers demonstrated that gas exchange and respiratory mechanics were improved with 3 consecutive sighs/min at 45 cm H₂O plateau pressure over 1 h in patients with acute respiratory distress syndrome (ARDS) ventilated with a lung-protective strategy. The improvements seen were lost within 1 h after return of ventilatory parameters to baseline values (13). Thus, improvement in gas exchange occurs at the expense of increases in mean and Ppaw with sustained alveolar recruitment and it is not surprising that the relatively modest sighs delivered during s resulted in no benefit in the current study.

Biologically variable ventilation confers the advantage of improved gas exchange at an unchanged Paw and a lower Ppaw than either c or s. Exclusive to bv, the animals also received many small VT/cycle with the potential for alveolar derecruitment.

Ongoing atelectasis is of significant concern during general anesthesia with mechanical ventilation (8). The near complete recovery of Pa_O2 and S/T to baseline values with bv indicates better A/ matching over time during a lengthy period of anesthesia in the present study. As such, bv may be of clinical relevance for control mode ventilation during anesthesia (14).

All variability files used to date in the laboratory have demonstrated inverse power law frequency distributions and all have been obtained by lengthy collections of physiological signals such as heart rate (1), respiratory rate, and blood pressure (15). Variability in these physiological signals is ubiquitous in mammals (16). Further clarification is necessary to determine if biological variability is representative of stochastic resonance as defined by Suki and coworkers (2). If such turns out to be the case, then biological variability to recreate normal variation in VT and f may be an example of tuned noise to enhance an output (Pa_O2) in a nonlinear system (4). We have demonstrated that the programmed variability follows an inverse power law frequency distribution. This "noisy" behavior may explain the effectiveness of bv. It is important to realize that signals with inverse power law frequency distributions are not random (a = 0 or white noise). Others have suggested that such biologically variable noise demonstrates deterministic behavior (19, 20). This experiment provides strong support for the theoretical proposal of how noisy Ppaw can increase recruitment of collapsed alveoli (2). It is uncertain if  = 0.11 has optimized the benefits that can be obtained with bv in this context. Whether or not such a "noisy" mechanical ventilator has clinical utility must await further study. However, it is entirely possible that clinical life support systems may be improved by programming them for biologically variable or natural noise.