Respiratory compliance but not gas exchange correlates with changes in lung aeration after a recruitment maneuver: an experimental study in pigs with saline lavage lung injury

Acute lung injury was induced through repeated lung lavage as described previously [11] and allowed to stabilize until the arterial blood partial oxygen pressure (PaO₂) had been below 100 mmHg for 60 minutes. The following measurements were obtained before and 10 minutes after a recruitment maneuver was performed.

All data are reported as mean ± SD. To correlate the parameters under investigation with the CT measurements, the Pearson's coefficient (r) was calculated. Where appropriate, multiple linear regression was used. The validity of the model was verified by a Durbin-Watson statistic. Because correlations of parameters with end-inspiratory or end-expiratory CT measurements exhibited equal results, only the end-expiratory data are presented. To determine the parameter with the strongest influence, the dimensionless standardized beta coefficient (beta_S) was calculated. Pre- and post-recruitment maneuver (RM) values were compared using Wilcoxon's signed ranks test. In the case of parameters exhibiting a significant difference, the dimensionless fractional change for any parameter 'X' was then calculated as fractional change (X) = X_postRM/X_preRM - 1 and correlation analysis performed as explained above. Fractional change values are expressed as percentages. Statistical significance was accepted at p < 0.05 (SPSS 11.0, SPSS, Chicago, USA).

We investigated parameters used to indicate the amount and the change of aerated and nonaerated lung in acute lung injury. We chose the lavage model in pigs for this because it is known to be easily recruitable. This model has been shown to cause lung inflammation [20], ventilation-perfusion mismatch equal to other models [21] and an increase in extravascular lung water and excess tissue [22]. Furthermore, the preferential distribution of atelectasis to the dependent lung could also be demonstrated in patients with ARDS by use of CT scanning [12]. The number of experiments is in line with recent studies investigating respiratory mechanics in acute lung injury [23, 24]. Increasing the power may have resulted in more subtle correlations, although we have found some correlations to be significant (certain effect) and others not (possible effect).

Our definition of recruitment may be questioned, because what we measured really is a density scale proportional to gas-tissue distributions. Thus, the decrease in a portion of HU labeled 'atelectasis' does not necessarily mean opening of alveoli. Instead, edema fluid could be squeezed out of the lung and pushed into poorly aerated lung; however, we did not find changes in extravascular lung water [22] or lung tissue after the recruitment maneuver. Therefore, the observed changes in differently aerated lung volumes could have been caused by transformation of completely collapsed lung into partly opened lung or by an increased homogeneity in the distribution of alveolar fluid [25]. Importantly, the observed changes in aerated lung volume were relatively small 10 minutes after the recruitment maneuver and do not support the usefulness of such a maneuver, which has also been demonstrated in clinical studies [26]. Possibly higher levels of PEEP could have enhanced recruitment, but to avoid possible influences of PEEP on the physiological parameters studied we maintained the same level of PEEP (10 cmH₂O).

Although impaired oxygenation is the main symptom in acute lung injury [27] correlated with atelectasis [28, 29], our study suggests that PaO₂ is less related to the amount of atelectatic lung than to the aerated lung that remains for ventilation. These studies suggested that there was a linear correlation between PaO₂ or shunt and atelectasis formation, especially if atelectasis was below 5% of total lung [28]. Lung healthy subjects were studied, however, and only one slice of the lung close to the diaphragm was analyzed, representing the area where most atelectases occur. So atelectasis as a fraction of the whole lung was probably much lower. Furthermore, there seems to be a difference in the characteristic of atelectasis formation between otherwise healthy lungs and injured lungs with high proportions of instable alveolar units that are poorly ventilated. Poorly aerated lung has been considered as low regions. Because we found a correlation between the PaO₂ and poorly aerated lung, it is possible that the regional blood flow through these regions was considerably high. Therefore, intrapulmonary shunt does not only happen in totally collapsed, but also in low , units. What the clinician wants to know is whether a certain improvement in oxygenation can predict the amount of recruitment. Improvements in gas exchange after recruitment are attributed mainly to two basic mechanisms: first, by redirection of blood flow from nonaerated to aerated lung regions and reduction of venous admixture, which we observed; and second, which we did not observe, through an increase in alveolar ventilation, leading to a reduction in PaCO₂. In several clinical studies that have failed to demonstrate a benefit for active recruitment [26, 30, 31], oxygenation parameters, but not mechanical parameters, were used for decision making. Because we could not find the PaO₂ changes representative of recruitment, even in a very recruitable model, this could have important implications on the interpretation of these studies. It seems that the amount of oxygenation improvement is not so much determined by the reduction of nonaerated lung, but by the blood flow through these regions.

The plateau pressure and static lung compliance correlated equally with nonaerated and poorly aerated lung volumes. It appears that in lung injury, V_POOR and V_NON are the main determinants in overall lung compliance. Following the argument of Barnas et al. [32] that the elastance (E) of the rib cage compartment is parallel to the elastance of the diaphragm-abdomen compartment, the elastances of the differently aerated lung compartments could behave similarly and thus be described by the equation 1/E_LUNG = k₁/E_HYP + k₂/E_NORM + k₃/E_POOR + k₄/E_NON, where the constants k_1–4 depend on their fraction of total lung volume. Thus in healthy lungs, E_L is mainly dependent on E_NORM, because it has the highest fraction of lung volume. But with increasing fractions of E_POOR and E_NON (with much higher values than E_NORM) they will become increasingly determinant for lung compliance. This hypothesis is supported by multiple regression analysis, showing that the fractional change of C_INF was most dependent on V_POOR (beta_S 0.550) and V_NON (beta_S -0.331).

The PV-curve has been used to obtain information about diseased lungs [33‐36]. Although the calculated curve may not equally fit all data [37], the mathematical analysis of the PV-curve is objective and the best available algorithm so far [38]. Because the PV-curve characteristics reflect a dynamic investigation of the lung, they have been used to set the parameters of ventilation [39]. We did not investigate whether the point of maximum compliance increase really reflects the lower inflection point (LIP). We were surprised that the inflation point of maximum compliance increase actually increased after recruitment in a nonlinear way (Fig. 5d), with a sharp increase beyond an increase in aerated lung >20%. If the point of maximum compliance increase truly represented the commencement of alveolar recruitment, it should be lower in conditions with less atelectasis. An explanation for this phenomenon could be that recruitment happens throughout the inflation curve [36], making the existence of a singular threshold opening pressure unlikely. Also, inflation LIP has been shown to only poorly represent the pressure at which recruited lung stays open [33, 40]. But since we did observe an increase in the LIP with recruitment, the logical consequence would be to increase PEEP after the recruitment maneuver.

Another parameter of the PV-curve, V_REC has been used as an indicator of recruited volume in several investigations [36, 41, 42], but it had never been validated with actual CT measurements. Especially in ventilation with FiO₂ 1.0, the V_REC represents unstable lung units prone to collapse. In our results, there was a significant increase in V_REC after the recruitment maneuver, which correlated with the observed changes in V_POOR and V_NON. This means that a significant portion of the recruited lung still collapsed endexpiratory, probably because we did not increase PEEP after the recruitment. Therefore, V_REC could not only serve as a measurement for recruited lung, but also for the lung in danger of being de-recruited.