Time to Generate Ventilator-Induced Lung Injury Among Mammals

Caironi et al., Electronic Supplementary Material, page 1

Time to generate ventilator-induced lung injury among mammals

with healthy lungs: a unifying hypothesis.

Pietro Caironi, Thomas Langer, Eleonora Carlesso, Alessandro Protti

and Luciano Gattinoni

Electronic Supplementary Material

Additional Methods

We performed a Medline research,updated until November 2009, employing “mechanical ventilation AND lung injury” or “ventilator-induced lung injury” as keywords. As no unique categorization on articles related to VILI has been commonly applied, further articles were also selected starting from references of pre-selected articles.

Data selection

From Medline research, articles reporting data on healthy animals ventilated until the achievement of a pre-terminal lung injury were selected. From the selected articles we defined as “VILI groups” the groups of animals ventilated until the achievement of such injury, and as “control groups” the groups of animals ventilated without the achievement of such injury. Of note, in order to avoid confounding factors, we excluded from control groups those animals in which no injury was observed as a potential consequence of an intervention other than a different ventilatory settings (such as genes expression in knockout mice, pharmacological interventions, change in positioning, and hypercapnia).

Morphometric, ventilatory setting and respiratory function data

From the selected articles, data on morphometry of the different animal species employed were obtained for both injured and non injured groups. Similarly, data on the type of ventilation applied (either volume-controlled or pressure-controlled mode), as well as data on VT, airway pressure, PEEP, respiratory rate, inspiratory to expiratory ratio, PaO2, FiO2 and respiratory system compliance were also obtained. Every time parameters were reported as ranges, we arbitrarily considered the mean values of the two extreme values as the value applied. Moreover, when FiO2 values were not reported, and further articles were available in which the same well-standardized model had been described more extensively by the same group of investigators, we obtained such values from those articles.When not reported, data on inspiratory to expiratory ratio were obtained from technical information on the ventilator employed. For data regarding alveolar morphology, we selected five articles, in which alveolar anatomical morphology of the species included in the present analysis was studied [1-4]. Values of alveolar diameter were therefore derived as a mean of the values reported. Values of alveolar membrane thickness were then calculated for each species from values of alveolar diameter and applying the equation reported by Mercer and colleagues [4].

Calculation of maximal lung stress

Lung stress was estimated based upon data on the airway pressure recorded (or applied) and measurements of partitioned respiratory mechanics for each mammal species, according to the following formula [5]:

Lung stress = Airway pressure * (Elastance Lung / Elastance Total) [species]1.1

where “Airway pressure” denotes the peak airway pressure resulting from the VT applied (or the plateau airway pressure applied, during ventilation in pressure-controlled mode), and “(Elastance Lung / Elastance Total) [species]” denotes the ratio between the lung elastance and total elastance of the respiratory system characterizing each species. As no data on these parameters were reported, we arbitrarily applied for each species values on partitioned respiratory mechanics as reported by different articles in which healthy animals have been studied. In particular, for sheep, rabbits, rats and mice data reported by Davey and colleagues [6] and Crossfill and colleagues [7] were respectively employed, whilefor pigs, data reported by Fisher and colleagues were applied [8]. Of note, data on partitioned respiratory mechanics for sheep and pigs have been reported as standardized on body weight[6, 8], while data on portioned respiratory mechanics of rabbits, rats and mice have been reported as absolute values [7].

Limitations of estimated maximal lung stress

The estimation of maximal stress applied at the end of inspiration to the lung parenchyma according to equation 1.1 may present some limitations.

First, as the lung stress has been estimated based upon the ratio between the lung elastance and the elastance of the total respiratory system (Elastance Lung / Elastance Total), its value may theoretically be underestimated due to the non-linear and the downward shape of the pressure-volume relationship commonly characterizing mammal species at high levels of volume and pressure. Although we cannot completely exclude such inaccuracy, we think that it may have an impact only in smaller mammal species (rabbits, rats and mice). In fact, while data on Elastance Lung / Elastance Total characterizing larger animals have been reported to be obtained at a range of pressures and volumes near total lung capacity[6, 8], no specific details have been reported for data obtained in smaller species[7]. Nonetheless, as these species are generally characterized by a Elastance Lung / Elastance Total[7]very close to 1, we think that such approximation may have had a minimal impact in the estimation of the lung stress applied.

Second,as no direct data on pleural pressure (or esophageal pressure) have been reported in the articles selected for the analysis, transpulmonary pressure had to be estimated based upon measurements of airway pressure and the estimate of the “transmission factor”, i.e., Elastance Lung / Elastance Total, characterizing each mammal species, as shown in Equation 1.1. Although such estimate is conceptually not based on the estimated variation of pleural pressure between end-expiration and end-inspiration, but rather on its absolute value, we think that the possible inaccuracy derived by such approximation may be considered as limited at a minimal degree, especially when considering the main perspective of our analysis, i.e., to individuate a possible rule explaining the great variability observed among different mammal species. In fact, the main difference observed among the respiratory mechanics of the mammal species considered has been reported to be related precisely on the different ratio between Elastance Lung and Elastance Total, which appeared to be quite constant within the same mammal species, rather than on other factors potentially affecting the estimation of lung stress.

Third, as sometimes data on the actual plateau airway pressure have not been directly reported (especially during volume-controlled ventilation and in smaller animals), in these cases values of lung stress have been derived from values of peak airway pressure, not taking into account the possible effect of airway and tissue resistance on the calculation of the actual transpulmonary pressure applied.Consequently, the lung stress estimated according to Equation 1.1 might slightly overestimate the actual lung stress applied. As this case applied mainly for small mammal species (see Table 1 in the main manuscript and Table 4 in the Online Data Supplement), based upon the position of these groups of animals within the relationship reported (see Figure 1 in the Online Data Supplement), we think that such overestimation may minimally affect the overall association between the lung stress applied and the time for the achievement of a pre-terminal VILI.

Calculation of maximal lungstrain

Maximal lung strain, i.e., the lung strain resulting at the end of tidal-inflation, was estimated as previously reported [5] according to the following formula:

Lung strain = (VT + PEEP volume) / FRC1.2

where VT denotes the VT applied, and “PEEP volume” the inflated lung volume due to the PEEP applied. As no data on FRC were reported in studies performed in pigs, rabbits, rats and mice, we arbitrarily selected four articles available in the literature for each of those species, in which values of FRC in healthy animals were reported, and we calculated the average values of FRC per kg of body weight for each species. Subsequently, we estimated for each group of animals the baseline values of FRC according to the body weight reported and we included them in Equation 1.2 (see Table 1 in the Online Data Supplement for complete references). Similarly, as no data on the inflated lung volume due to the applied PEEP were available, “PEEP volume” was estimated based upon the level of PEEP applied and the standard characteristics of respiratory mechanics (i.e., respiratory system compliance) of the species considered, according to the following formula:

PEEP volume [species] = PEEP * Respiratory system compliance [species]1.3

where “PEEP volume [species]” denotes the inflated lung volume due to the applied PEEP in a specific mammal species, “PEEP” the PEEP values specifically applied, and “Respiratory system compliance [species]” denotes the respiratory system compliance characterizing healthy animals of that particular species. For such parameters, values recorded in each articles at baseline (i.e. in healthy conditions) were employed. For mice, as only one selected articles reported data on respiratory system compliance [9], we arbitrarily selected five articles available from the literature [7, 10-13], and we employed the mean value of respiratory system compliances there reported.Moreover, when data on FRC, VT and respiratory system compliance were not directly reported but were showed on graphs, values were graphically derived by careful visual inspection of the graphs [14-17]. Finally, for article in which pressure-controlled ventilation was applied [14-20], lung strain was estimated as the ratio between mean values of both VT and FRC recorded at the beginning and at the end of ventilation, as it could have varied during the development of lung injury (due to both VT and FRC variation).

Finally, to take into account the actual time of exposition of the lung to the strain applied during either the inspiratory (VT and PEEP) and the expiratory phase (just PEEP) of each tidal breath, the average strain applied to the lung during each breath was calculated according to the inspiratory-expiratory time ratio:

Weighted lung strain = [(VT + PEEP volume) / FRC * Tinsp + (PEEP volume / FRC) * Texp] /

(Tinsp + Texp)1.4

where VT denotes the VT applied, PEEP volume the inflated lung volume due to the applied PEEP, Tinsp the inspiratory time of each tidal breath, and Texp denotes the expiratory time of each tidal breath.

Limitations of estimated lung strain

The calculation of lung strain as defined according to Equation 1.2 represents an estimate rather than the actual lung strain applied to the lung parenchyma during mechanical ventilation, and for this reason is based upon some assumptions and presents some limitations. A full discussion of this topic has been previously reported elsewhere [21]. Firstly, Equation 1.2 is based upon the assumption of a mono-alveolar lung model. Although such assumption may theoretically have an influence in its calculation, we think that the potential errors derived from it may be reasonably considered at a low degree in our analysis. In fact, although during the inflation of the lung a portion of alveoli will unfold and will be added to those already open at lung resting volume (i.e., FRC)[22], such phenomenon is usually considered be relatively limited in healthy lungs, as compared to what may occur in a diseased and collapsed lung. Moreover, although the real lung strain actually applied to the lung should be theoretically derived from the calculation of the lung strain applied to each ventilated alveolus, its simple estimation based upon ventilator parameters which may be commonly derived or measured in a clinical setting (such as VT and FRC) according to Equation 1.2 has appeared to be empirically very close to the actual lung strain applied [21]. Secondly, lung strain as defined in our analysis may not take into account the possible dishomogeneity occurring during mechanical ventilation, especially at high volumes and pressures. Nonetheless, although we cannot exclude a priori a possible influence on the results observed, we think that the strict correlation observed between the duration of mechanical ventilation and the lung strain applied as compared to the worse correlations observed with the other parameters investigated further suggest the likely low impact of such limitations on our analysis.

Impact of alveolar recruitment in the calculation of maximal lung strain

Since the initial development of a lung injury associated to mechanical ventilation may lead to the occurrence of lung edema and consequent alveolar collapse, some degree of intra-tidal lung recruitment may have characterized the final phase of the manifestation of VILI in animals included in the VILI groups. As the occurrence of lung recruitment may increase FRC, therefore affecting the estimate of lung strain (as previously reported [21, 23]), it is likely that values of lung strain actually applied onto the lung parenchyma at the end of the experiment (both during pressure-controlled and volume-controlled mechanical ventilation) have been lower than those calculated according to Equation 1.2. Unfortunately, no data were reported on such process in the articles selected, and no method for its estimation is available, to the best of our knowledge. Nonetheless, we reasoned that lung recruitment might have potentially affected the calculation of lung strain at a relatively quantifiable degree only in the very last period of exposition to mechanical ventilation, during the overt and “avalanche” manifestation of VILI. It is therefore conceivable that the calculation of the average lung strain applied according to Equation 1.2 overestimated the actual value applied only at a relatively low degree.

Additional statistical analysis

Comparison between larger (sheep and pigs) and smaller species (rabbits, rats, and mice) was performed by Mann-Whitney Rank Sum test, as data did not appear normally distributed. Although rabbits presented a body weight similar to that of pigs (at least for the studies selected in our analysis), we classified rabbits as “smaller species” based upon the strict similarity of the respiratory system mechanics of this species to that of other rodents (especially mice and rats), as previously reported [4, 7, 24].

To elucidate the possible rule regulating the variation of several respiratory parameters (i.e. FRC, respiratory mechanics, etc.) as a function of body mass, we employed the methodology of allometry, as it has been previously reported [25, 26]. Briefly, allometry allows the determination of the power function describing the modification in magnitude of the structural parameters analyzed (i.e., FRC) along the variation of body mass, by applying the following equation:

y = a * (body mass)b2.1

where y denotes the respiratory parameters analyzed, “body mass” the body weight of the animal species considered, “a” denotes a coefficient of the relationship (indicating the y-intercept of the relationship when the two variables analyzed – the respiratory parameters considered and the body mass – are transformed in logarithmic scale, as shown below), and “b” denotes the scaling factor indicating the nature of the allometric relationship. In details, when coefficient b equals 1, it indicates that the respiratory parameters analyzed linearly increases with the increase of body mass;when coefficient b > 1, it indicates that the respiratory parameters analyzed is relatively smaller in smaller animals as compared to larger animals, proportionally to their body mass;finally, when coefficient b < 1, it indicates that the respiratory parameters analyzed is relatively greater in smaller animals as compared to larger animals, proportionally to their body mass. Graphically, the allometric relationship between a respiratory parameter and body mass may be figured as a linear regression when the two variables (i.e., both the respiratory variable and body mass) are transformed in logarithmic scale, according to the following formula:

log y = b * log (body mass)b + log a2.2

where y, a and b denotes the same mathematical variables and parameters as shown in Equation 2.1, while “b” (i.e., the scaling factor) represents the slope of the linear regression, and “log a” represents the y-intercept of the same regression. Of note, the transformation ofboth the respiratory variable analyzed and body mass in logarithmic scale allows a better and clearer visualization of the nature of the allometric relationship, i.e., the value of coefficient b (as shown in Figure 4A and 4B, as well as in Figure 6 of the Online Data Supplement).

Statistical analysis was performed using SigmaStat 3.1 (Systat Software, Inc., Chicago, IL, USA) computer software.

Additional Results

Differences in mammal species: thickness of the alveolar membrane

In animals included in the VILI group, after normalization for the specific thickness of the alveolar membrane of each single species, maximal lung stress appeared to be highly and better correlated with the duration of mechanical ventilation (r2 = 0.77, p<0.0001; see Figure 2A in the Online Data Supplement) than the rough value of the estimated maximal lung stress. Similarly, maximal lung strain as normalized for the thickness of the alveolar membrane was confirmed to be inversely and highly correlated to the time of achievement of VILI (r2 = 0.85, p<0.0001; see Figure 2B in the Online Data Supplement).

Mammal species and lung strain

Duration of mechanical ventilation appeared to be closely and inversely correlated to the value of maximal lung strain applied (r2 = 0.85, p < 0.0001; Figure 2 of the main manuscript). Moreover, when considering lung strain as weighted for the actual time of application during each tidal breath (see Table 4 in the Online Data Supplement), a similar relationship was observed (r2 = 0.83, p < 0.0001; Figure 3), although it pointed out a greater separation between animals ventilated with a lung strain greater than 2.0, in which the development of the pre-terminal injury appeared to be almost immediate, and animals ventilated with a lung strain below 1.5, in which the time of achievement of VILI exponentially increased with the decrease of the lung strain applied. Of note, larger species appeared to be mostly located in the left and upper part of the relationship (lower weighted lung strain and longer duration of mechanical ventilation), while smaller species appeared to be mainly located in the right and lower part of the relationship (higher weighted lung strain and shorter duration of mechanical ventilation). Although we cannot completely exclude a more precise and effective “clinical” treatment during the study protocols in larger animals, we think that the consistency of the data observed from different group of investigators within the same mammal species may suggest the minimal role of such confounding factors. Moreover, the constant observation of a lower weighted lung strain applied in larger species, as compared to that applied in smaller species, further underlines the main differences in the characteristics of the respiratory system between larger and smaller mammal species (a relatively lower chest wall compliance and a relatively higher FRC in larger species), indicating the actual impossibility of applying in large species a higher lung stress and strain.