INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recent clinical trials have indicated that ventilatory strategies aimed at limiting ventilator-induced lung injury (VILI) may have a positive impact on the outcome of the acute respiratory distress syndrome (1, 2). These strategies impose limits on tidal volume (VT) and inflation pressure, and may result in hypercapnic acidosis. Until recently, the latter condition was mainly viewed as a side effect to be either tolerated (permissive hypercapnia) or treated by increasing the rate of mechanical breathing or by giving a bicarbonate infusion, or with techniques such as tracheal gas insufflation or extracorporeal CO₂ removal (3).

However, hypercapnia and acidosis have myriads of effects at the cellular level, several of which might influence the development of VILI. First, it is now believed that mechanical stretch induces lung injury in part by activating various intracellular signaling cascades (4), some of them having demonstrated a dependence on pH or partial pressure of CO₂ (PCO₂) (5). Second, surfactant production or function may be modulated by acid-base status (9). Third, hypercapnia appears to downregulate the endogenous production of nitric oxide (NO) in the lung (12), an effect of potential relevance, considering the dual cytoxoxic/cytoprotective potential of NO (13). Additionally, respiratory acidosis protects rabbit lungs ex vivo and in vivo against ischemia-reperfusion injury (14, 15). It is unknown whether such protection would extend to VILI.

With these considerations in mind, we designed the present study to investigate the possible impact of hypercapnic acidosis on the development of VILI in an isolated, perfused rabbit lung preparation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Care, techniques, and procedures used in the study were approved by the animal care committee of our institution.

Isolated Perfused Lung Preparation

We used a previously described, isolated, perfused lung preparation (New Zealand rabbits) (16, 17). No inhibitor of cyclooxygenase was added to the perfusate. Perfusion was set at a constant flow of 300 ml/ min and left atrial pressure (Pla) at 6 mm Hg (referenced to the top of the lung). Initial ventilation (Veolar ventilator; Hamilton Medical, Rhazüns, Switzerland) was set in the pressure controlled mode (pressure controlled ventilation [PCV]), with a positive end-expiratory pressure (PEEP) of 3 cm H₂O, a respiratory rate (RR) of 20 breaths/min, and an inspiratory time fraction of 0.33. The ventilator was connected to two gas tanks: the first tank (O₂: 40%; CO₂: 5%; N₂: 55%) was connected to the air inlet and the second (O₂: 40%; CO₂: 25%; N₂: 35%) to the oxygen inlet. Thus, adjustment of the ventilator FI_O2 knob allowed titration of the inspired CO₂ fraction (FI_O2) to the desired perfusate PCO₂.

Protocol

The experimental protocol consisted of four successive 20-min periods. During the first period (Baseline), lungs were ventilated at 15 cm H₂O peak airway pressure (Paw) with normocapnia (perfusate PCO₂ = 40 ± 5 [mean ± SEM] mm Hg). Subsequently, the preparations were randomized into three groups and studied for three additional periods (T1, T2, and T3). In the control group (C, n = 7), peak Paw and perfusate PCO₂ were kept at the Baseline values. In the high-pressure normocapnic group (HPNC, n = 7), peak Paw was raised in increments of 5 cm H₂O at the beginning of each new period to a maximum value of 30 cm H₂O during T3. The targeted arterial CO₂ tension (Pa_CO2 and pH were the same as in the C group. In the high-pressure hypercapnic group (HPHC), n = 7], peak Paw was increased as in the HPNC group, but the targeted Pa_CO2 was between 70 and 100 mm Hg.

Measurements

Paw, flow, VT, pulmonary artery pressure (Ppa), Pla, and pulmonary capillary pressure (Pcp) were measured as previously described (16). Three blood gas analyses were performed during each period. At Baseline and T3, plasma nitrite + nitrate (NOx) concentrations were assayed (18, 19). Edema formation (weight gain [WG] from the beginning to the end of each period) was used to calculate changes in WG from Baseline (WG at T1,T2,T3 as WG[Tx]  WG[Baseline]). The ultrafiltration coefficient (K_f) was determined at the end of each period (16), and changes in K_f from Baseline (K_f at T1,T2,T3) were calculated as K_f (Tx)  K_f (Baseline) (17). When these other measurements were completed, recovered fluid from a bronchoalveolar lavage (BAL) of the left lung was divided into two aliquots to measure the BAL hemoglobin level (Hb_BAL)(20), and, after centrifugation, the supernatant protein (bicinchonic acid assay, [BCA]; Pierce, Rockford, IL) and NOx concentrations (18, 19). NOx was used to track lung NO production. The lung wet weight (WW_L)/dry weight (DW_L) ratio was measured with lung tissue pieces sampled from each of the right lung lobes.

Statistics

The mean values of variables measured only once were compared among groups with the Kruskal-Wallis analysis of variance (ANOVA) on ranks; the Student-Newman-Keuls' adjustment was performed for multiple comparisons. Variables for which one measurement was made in each experimental period were analyzed with a repeated measures ANOVA. Correlations were examined through linear regression analysis. The level of  was set at 0.05. In the RESULTS section, all data are summarized as mean ± SEM.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The three study groups had similar characteristics at Baseline, as summarized in Table 1. In particular, the initial WG did not differ among the groups. The Baseline K_f was higher in the C group than in the HPHC group (Table 1), but this difference was small, and values in all three groups were well within the normal range for rabbit lungs (21). The ischemic time (21 ± 1 [mean ± SEM] min), the initial hematocrit (6.2 ± 0.1%), the initial perfusate pH (7.30 ± 0.03), and the perfusate temperature (36.5 ± 0.1° C) did not differ among the groups. After randomization, the controlled variables for the three groups differed at each stage of the study, as intended (Table 2).

Indices of Lung Injury

Neither edema formation (WG: Figure 1a) nor permeability (K_f: Figure 1b) increased significantly over time in the C group. In the HPHC group, WG but not K_f increased from T1 to T3 (p = 0.04 and p = 0.25, respectively). In the HPNC group, both WG (p = 0.01) and K_f (p = 0.005) increased significantly over time. At the end of the protocol (T3), WG and K_f were significantly greater in the HPNC than in the HPHC or C groups (Figures 1a and 1b; WG: p < 0.001; K_f p = 0.01). The HPHC group differed significantly from the C group for WG (p = 0.03) but not for K_f (p = 0.27).

View larger version (22K): [in this window] [in a new window]   Figure 1.   Indices of lung injury. *p < 0.05 within-group comparison (T3 versus T1), p < 0.05 for between-group comparison at the same time;   versus C group.  versus C group and HPHC group.

At the end of the experimental period, the WWL/DWL ratio was higher in the HPNC than in the HPHC or C groups (p < 0.005; HPNC group = 8.33 ± 0.25; HPHC group = 7.19 ± 0.18; C group = 6.85 ± 0.34). The BALF protein concentration was higher in the HPNC group than in the C group (p < 0.05) (Figure 1c), whereas BALF hemoglobin levels were higher in the two active treatment groups than in the C group (p < 0.05) (Figure 1d). The BALF protein concentration was higher in the HPNC than in the HPHC group (p < 0.05), with a similar but statistically insignificant trend noted for BALF hemoglobin (p = 0.24).

NOx in BALF and Perfusate

The overall mean circulating NOx was 41.2 ± 3.1 µml/L at Baseline and 58.8 ± 3.3 µml/L at T3 (p < 0.01). The level of NOx in the perfusate increased over time in all groups. This change was less marked in the HPHC group than in the other two groups, a difference that did not reach statistical significance (Figure 2a). The level of NOx in the BALF, measured at the end of the experimental period, was highest in the HPNC group, intermediate in the HPHC group, and lowest in the C group (Figure 2b, p < 0.05 for all comparisons). BALF NOx correlated significantly with all indices of injury: K_f (r = 0.65, p < 0.001); BALF hemoglobin (r = 0.62, p < 0.01); and BALF protein (r = 0.7, p < 0.001).

View larger version (16K): [in this window] [in a new window]   Figure 2.   NOx concentration (a) change over time in perfusate and (b) at the end in BALF. *p < 0.05 versus C and HPHC groups,  p < 0.05 versus C group.

Hemodynamic Data

At Baseline, there were no significant differences in vascular pressures among the study groups (Table 1). Throughout the experiment, Pla was adequately controlled at values around the targeted value (Table 2). After randomization, the mean pulmonary artery pressure () measured during tidal ventilation remained close to the Baseline value and similar in all groups until T2. At the beginning of T3, it became higher in the HPNC and HPHC groups than in the C group (Table 2). Mean differed among all groups only at the end of T3 and was highest in the HPNC, intermediate in the HPHC, and lowest in the C group (data not shown, p < 0.05 for all comparisons). Pcp measured under apneic conditions with a continuous positive airway pressure (CPAP) of 5 cm H₂O was similar among groups from Baseline (Table 1) to the end of the study [9.2 ± 0.2 cm H₂O (C group); 9.8 ± 0.4 cm H₂O (HPNC group); 8.5 ± 0.5 cm H₂O (HPHC group); p = 0.1]. Pulmonary vascular resistance (as reflected by Ppa minus Pla under conditions of constant lung perfusion, interrupted ventilation, and a CPAP of 5 cm H₂O), remained practically stable and comparable among groups until the end of T2 (Table 2). At the end of T3, by contrast, this variable was significantly higher in the HPNC group than in the other two groups (Table 2) (p < 0.05). At this time, pulmonary vascular resistance correlated with WG (r = 0.58, p = 0.005) and K_f (r = 0.5, p = 0.02).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our main finding was that respiratory acidosis decreased the severity of VILI in this model. Hemodynamic effects of the acid-base status were unlikely to account for this observation, because they were modest and only apparent after significant lung injury developed. Our data raise the hypothesis that respiratory acidosis could reduce VILI by attenuating the increased NO production arising from high-pressure mechanical ventilation.

Adequacy of the Experimental Model

The validity of the present experiments depends crucially on the actual induction of VILI in the group of isolated lungs ventilated with high pressure in normocapnia (HPNC). Injury was assessed by measurement of edema formation (WG) and altered lung permeability to markers covering a wide range of sizes (i.e., water [K_f], protein [BALF protein concentration], and red blood cells [BALF hemoglobin concentration]). Each of these measurements taken in isolation has limitations for assessing injury. However, all of them significantly differed in the HPNC and the C groups at the end of the experiment (Figure 1). In our previous study of the same model, injurious ventilatory patterns induced these same multiple abnormalities, which correlated closely with histologic scores of both alveolar and extraalveolar hemorrhage (17). To the extent that extravasation of red blood cells is indicative of either anatomic or severe functional disruption of lung structures, we are confident that the mechanical stress imposed on the lungs ventilated with high pressures was sufficient to produce injury, at least in normocapnia.

Possible Mechanisms of Protection Against VILI Afforded by Respiratory Acidosis

The study data clearly indicate a protective effect of hypercapnic acidosis against VILI in this model. Indeed, with the possible exception of BALF hemoglobin, all markers of lung injury that were noted to be increased in the HPNC group were blunted in the HPHC group (Figure 1). The study was primarily conducted to establish this fact. It was not designed to sort out the relative contributions of PCO₂ and pH. Nor did it specifically address the possible mechanisms of protection, for which, nevertheless, a few clues exist in the data.

Hemodynamic factors. In the course of experimentation, the and vascular resistance increased less in the HPHC than in the HPNC group (Table 2). This difference might have been due to hypercapnia per se, which is a known vasodilator in the pulmonary circulation (22). In the presence of lung injury, differences in Ppa may have a significant impact on edema formation and possibly on injury. Nevertheless, several arguments run against a fundamental role of hemodynamic factors in the lung-protective effects of respiratory acidosis. First, although hypercapnia largely blunted the formation of edema (Figure 1a), it had no influence on capillary pressure throughout the experiment. Second, the hemodynamic difference between the HPNC and HPHC groups was present only after significant injury developed (end of T3) and was limited to a modestly higher in the HPNC group (19.6 ± 1.1 mm Hg compared with 16.5 ± 0.8 mm Hg; p < 0.05). Respiratory acidosis per se had no discernible effect on the vascular tone prior to injury, probably because the CO₂-mediated vasodilation in this condition is offset by acidosis-mediated vasoconstriction (22) (i.e., at T1 and T2; Table 2). These considerations converge to indicate that pulmonary hypertension was most likely a marker rather than a cause of VILI in this model, as also documented by other investigators (23).

Possible role of NO. We found that circulating NOx tended to increase less from T1 to T3 in the HPHC group than in the other two study groups (Figure 2a). In addition, BALF NOx at the end of the experimental period was lower in the HPHC group than in the HPNC group (Figure 2b), and correlated with all indices of injury. Since BAL and circulating NOx track the formation of NO within the lung (24), our results suggest that the latter may have been blunted by respiratory acidosis. Indeed, depressed NO production by high airway concentrations of CO₂ has been reported by other investigators, both in vivo and in isolated perfused rabbit lungs (12). That BALF NOx was higher in the HPNC than in the C group would be consistent with recent data indicating that lung stretch may stimulate the production of NO by the airway mucosa (25, 26). At the cellular level, NO may be both protective and cytotoxic, the net result being highly dependent on the particular pathophysiologic state (13). Reflecting this complexity, inhalation of NO was protective in a model of oxidant-induced lung injury (27), whereas the endogenous production of NO was deleterious to the lung in experimental endotoxemia (28).

If the cytotoxic potential of NO predominated under the conditions of the present study, our results might fit into the following scenario: (1) stretch-induced overproduction of NO contributed to VILI; and (2) hypercapnic acidosis was protective by interfering with this mechanism. Further experiments will be required to test this hypothesis.

Limitations of the Study

The model used in the study has important limitations that are discussed elsewhere (16), and characteristics that may prevent direct extrapolation of our results to the intact animal, and certainly to patients. For instance, we kept blood flow constant. In the intact animal, respiratory acidosis enhances both sympathetic activity and catecholamine secretion, which tend to increase cardiac output and pulmonary vascular resistance and pressure (3). This could partly offset the beneficial effect of hypercapnia on VILI. In addition, assuming that reduced NO production is an important mechanism by which respiratory acidosis ameliorates VILI, the net result may be either beneficial or detrimental according to the underlying type of lung insult requiring mechanical ventilation. Although hypercapnia has been reported to be associated with an improved outcome (1, 29, 30) and to cause no serious complications, the improved outcome could be entirely due to other aspects of the ventilatory strategy used in the studies in which this was found, and not to hypercapnia per se. Moreover, whether hypercapnia resulting from hypoventilation and hypercapnia resulting from inhaled CO₂ may provide equivalent lung protection is still unknown.

Conclusions

Two recent studies by Kavanagh and colleagues demonstrated the lung-protective potential of hypercapnic acidosis in an ischemia-reperfusion model of lung injury (14, 31). Our results show that the protection afforded by hypercapnic acidosis is not limited to ischemia-reperfusion injury, but also applies to VILI. This is of great interest, since lung-protective ventilatory strategies tend to induce respiratory acidosis. Respiratory acidosis may worsen gas exchange in patients with acute respiratory distress syndrome (32) and may have undesirable effects on organs other than the lungs, as reviewed elsewhere (3). If confirmed in intact animals, our results would, however, indicate that hypercapnic acidosis should not necessarily be viewed as an undesirable side effect to be tolerated, but rather as a potentially useful adjunct to current strategies for lung protection during mechanical ventilation. Our results also reinforce the concept that medical interventions designed to restore normal physiologic parameters may be counterproductive in critically ill patients, at least under certain clinical circumstances.