BIOLOGIC VARIABILITY IN MECHANICAL VENTILATION IN A CANINE OLEIC ACID LUNG INJURY MODEL

To the Editor :

We read with interest the article by Nam and colleagues that finds no advantage with biologic variability in mechanical ventilation in a canine oleic acid lung injury model (1). We have developed biologically variable ventilation (BVV) and have published a series of articles (2) showing its efficacy in experimental modelstherefore, we have analyzed the article by Nam and colleagues and would like to report the following:

1. Statistical analysis. Nam and colleagues report no difference in Pa_O2 between groups in their study. Using longitudinal analysis (an ambiguous description in methodology) they report the p value for Pa_O2 between groups as 0.77. In their study, at 240 min, mean Pa_O2 is 79% greater with BVV. In our work we showed a Pa_O2 difference of 87%, favoring BVV at 240 min. We have examined their raw data as presented in their Figure 5 time course of Pa_O2 for the nine individual animals in each of the CV and BV groups. When analyzed using the same statistical approach as in our publicationsa split plot ANOVA with repeated measureswe show a statistical significance for Pa_O2 between groups over time for their data. The group × time interaction is p = 0.0125. At 240 min, by least squares means test to examine for effect of time for Pa_O2 between groups, p = 0.0031. Even with a Bonferroni's adjustment this difference would be significant. Thus, a benefit for oxygenation is seen with BVV in the study by Nam and coworkers when the statistical analysis is done as described in our studies. The approach to statistical analysis represents the fundamental difference in interpretation between our findings and those reported by Nam and colleagues. However, other important differences between the two studies are present to account for the negative findings by Nam and colleagues and include:

2. Differences in severity of ARDS. The canine model by Nam and coworkers was associated with a total mortality of 33% versus 0% in our porcine model. It is highly questionable that valid conclusions can be made from animals so sick that end Pa_O2 is below normal venous O₂ in 5/10 animals reported in their Table 5.

3. Misinterpretation of stochastic resonance model. Nam and colleagues have erroneously interpreted the work of Suki and colleagueswork that may offer an explanation for the efficacy of BVV(6). The model by Suki and coworkers assumes two populations of alveolione population aerated and the other collapsed. They do not base their model on P_flex as suggested by Nam and colleagues. An important part of the model by Suki and colleagues is that an optimal variability or noise occurs with stochastic resonance. The noise associated with variable f in our study was 11.5% (coefficient of variation) and in the study by Nam and colleagues it was 24%. Increased noise can decrease the optimal signal-to-noise ratio with stochastic resonance, and the 100% higher noise in the study by Nam and colleagues may be too great in this situation to maximize benefit.

In conclusion, the work by Nam and coworkers is seriously flawed with methodologic and interpretative errors. A more carefully designed study is necessary to independently prove or disprove the efficacy of BVVin fact our analysis of Nam and colleagues' data for Pa_O2 indicates that they have, in fact, shown efficacy for BVV. If, as we suggest, efficacy for BVV has been shown, the experiment performed by Nam and colleagues is a strong endorsement for the robustness of BVV, given the marked differences from our initial study.

University of Manitoba, Winnipeg MB, Canada

1. Nam AJ, Brower RG, Fessler HE, Simon BA. Biologic variability in mechanical ventilation rate and tidal volume does not improve oxygenation or lung mechanics in canine oleic acid lung injury. Am J Respir Crit Care Med 2000; 161: 1797-1804 [Abstract/Free Full Text].

2. Lefevre GR, Kowalski SE, Girling LG, Thiessen DB, Mutch WAC. Improved arterial oxygenation after oleic acid lung injury in the pig using a computer-controlled mechanical ventilator. Am J Respir Crit Care Med 1996; 154: 1567-1572 [Abstract].

3. Mutch WAC, Eschun GM, Kowalski SE, Graham MR, Girling LG, Lefevre GR. Biologically variable ventilation prevents deterioration of gas exchange during prolonged anaesthesia. British Journal of Anaesthesia 2000; 84: 197-203 . [Abstract/Free Full Text]

4. Mutch WAC, Harms S, Graham MR, Kowalski SE, Girling LG, Lefevre GR. Biologically variable or naturally noisy mechanical ventilation recruits atelectatic lung. Am J Respir Crit Care Med 2000; 162: 319-323 [Abstract/Free Full Text].

5. Mutch WAC, Harms S, Lefevre GR, Graham MR, Girling LG, Kowalski SE. Biologically variable ventilation increases arterial oxygenation over that seen with positive end-expiratory pressure alone in a porcine model of acute respiratory distress syndrome. Crit Care Med 2000; 28: 2457-2464 [Medline].

6. Suki B, Alencar AM. Sujeer MK, Lutchen KR, Collins JJ, Andrade Jr JS, Ingenito EP, Zapperi S, Stanley HE. Life-support system benefits from noise. Nature 1998; 393: 127-128 [Medline].

From the Authors :

Dr. Mutch and associates question our statistical methods (3), and their analysis of our data finds a significant benefit of BV on Pa_O2. However, it is unclear whether they account for the data missing at later time points due to mortality. In contrast, we utilized longitudinal analysis, which is both well documented (1) and implemented in major statistical packages including Stata, SAS, and S-plus. This analysis includes all data from all animals, accounting for both temporal trends and correlations between measurements within each animal. The mean values at later time points are strongly influenced by the absence of data from the animals that died from hypoxemia, as well as the high Pa_O2 in two BV outliers. Indeed, a simple inspection of Figure 5 shows no qualitative difference between the groups, with the exception of these two BV animals.

With regard to the differing severity of the oleic acid models, our initial postinjury Pa_O2 was virtually identical to that in the original report of Lefevre and colleagues (2). In our discussion, we speculated that the fact that we reported worse outcomes despite similar initial injuries could reflect a species difference, with dogs demonstrating a more progressive type of injury that is unresponsive to BV.

Finally, we have not misinterpreted the work of Suki and colleagues (4). The whole lung pressure-volume (P-V) curve is the summation of P-V curves of all alveoli. This is simplified in the Suki and coworkers model to two groups, aerated units and collapsed but recruitable units. In this model, the equivalent of "Pflex" is determined by the opening pressure of the collapsed unit curve, and peak inspiratory pressures must be distributed about this value to benefit from stochastic resonance. We discuss in detail the differences in the "variability" between the studies. In an attempt to use a relevant "biological" variation, our respiratory pattern was recorded from an awake, spontaneously breathing dog. It had a coefficient of variation of 24%. We cannot speculate as to how much variability is optimal. We maintain, however, that there is no theoretical need for a "biological" source of variation. Perhaps as Suki and colleagues suggest, a breathing pattern with gaussian or other distribution with a different coefficient of variation might be more effective.

In summary, we agree with Dr. Mutch and colleagues that further study is necessary to independently prove or disprove the efficacy of BVV. However, our carefully done study is not, as they suggest, a "strong endorsement for the robustness of BVV."

The Johns Hopkins Medical Institutions, Baltimore, MD

1. Diggle PJ, Liang KY, Zeger SL. Analysis of longitudinal data. New York: Oxford University Press; 1995.

2. Lefevre G, Kowalski S, Girling L, Thiessen D, Mutch W. Improved arterial oxygenation after oleic acid lung injury in the pig using a computer-controlled mechanical ventilator. Am J Respir Crit Care Med 1996; 154: 1567-1572 .

3. Nam AJ, Brower RG, Fessler HE, Simon BA. Biologic variability in mechanical ventilation rate and tidal volume does not improve oxygenation or lung mechanics in canine oleic acid lung injury. Am J Respir Crit Care Med 2000; 161: 1797-1804 .

4. Suki B, Alencar AM, Sujeer MK, Lutchen KR, Collins JJ. Life-support system benefits from noise. Nature 1998; 393: 127-128 .