INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Historically, it is accepted that patients with catastrophic, dismal, acute respiratory failure while breathing 100% O₂ develop reabsorption atelectasis, as a result of alveolar denitrogenation in units with low inspired ventilation-perfusion (A/) ratios (named "critical") (1, 2). This mechanism has been clearly demonstrated in healthy subjects submitted to elective surgery during general anesthesia (3). Yet, earlier data using the multiple inert gas elimination technique (MIGET) (4, 5) in acute lung injury (ALI), in which increased intrapulmonary shunt is the principal mechanism governing arterial hypoxemia, were against this concept (6, 7).

The aim of our study was to investigate, in patients with ALI, the pulmonary gas exchange response to 100% O₂ breathing, using MIGET. A subset of patients with chronic obstructive pulmonary disease (COPD), in whom A/ imbalance but not intrapulmonary shunt is the major determinant of gas exchange abnormalities, was also assessed. In COPD, the breathing of 100% O₂ deteriorates A/ heterogeneity without influencing intrapulmonary shunt, suggesting release of hypoxic pulmonary vasoconstriction (8, 9).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Twelve patients (age, 54 ± 17 yr [SD]; 9 males) with life-threatening acute respiratory failure needing mechanical ventilation were studied. Eight patients (48 ± 18 yr) had ALI, including four with acute respiratory distress syndrome (ALI-ARDS) and four without it (ALI-nonARDS) (10); the remaining four were older (66 ± 2 yr) (p < 0.01) and had an acute exacerbation of COPD (Table 1). The latter subset was intentionally small because basal findings were in keeping with previous findings (8, 9, 11). Other inclusion criteria were: clinical stable conditions; Pa_O2 > 60 mm Hg at a fraction of inspired oxygen (FI_O2) equal to or lower than 0.60; maintenance vasoactive therapy with no more than one drug; and absence of other comorbidities. All patients were intubated with a Portex cuffed endotracheal tube and mechanically ventilated with a Siemens 900C Servo Ventilator (Siemens- Elema BA, Solna, Sweden) using the volume-controlled mode with constant inspiratory flow. The study was approved by the ethics committee of Hospital Clínic, and written informed consent was obtained from each patient's next of kin.

Respiratory and Inert Gases

Inspired O₂ fraction was measured using mass spectrometry (Multigas Monitor MS-2; BOC-Medishield, London, UK). Minute ventilation was measured with a calibrated Wright spirometer and the measured valued was corrected to BTPS (Respirometer MK8; BOC-Medical, Essex, UK). Arterial and mixed venous blood samples were analyzed in duplicate for pH, PO₂, and PCO₂ (IL-1306; Instrumentation Laboratories, Milan, Italy). The venous admixture ratio [shunt fraction (S/T) = (Cc'_{O 2}  Ca_O2)/(Cc'_{O 2}  Cv_O2)], where Ca_O2, Cv_O2, and Cc'_{O 2} are arterial, mixed venous, and capillary O₂ content, respectively, was also calculated.

Distributions of A/ ratios were obtained using MIGET, whose local setup has been reported in full detail elsewhere (5, 12). The duplicate samples of each set of measurements were treated separately, the final data resulting in the average of parameters determined from each time point. Definitions of the principal inert gas variables are summarized in Table 2.

Hemodynamics

The electrocardiogram was continuously monitored. An 18-gauge plastic cannula was inserted into the radial artery for monitoring systemic arterial pressure (Psa) and for arterial blood gas sampling. A 7-French triple-lumen thermodilution balloon-tipped pulmonary artery catheter (Edwards Swan-Ganz; Baxter Healthcare Corporation, Irvine, CA) was inserted percutaneously and progressed into the pulmonary artery for measurement of pulmonary artery (Ppa), pulmonary artery occlusion, and right atrial pressures (HP Model 56S; Hewlett-Packard, Andover, MA). Thermodilution cardiac output (T) was measured by injecting 10 ml of cold 0.9% saline solution into the proximal end of the pulmonary artery catheter. The reported T values are the mean of four consecutive measurements.

Study Design

Patients were studied in the semirecumbent position under steady-state conditions, as demonstrated by stability (± 5%) of both ventilatory and hemodynamic variables, and by the close agreement between duplicate measurements of mixed expired and arterial O₂ and CO₂ (± 5%). All patients were sedated during the study with continuous intravenous infusion of midazolam (Dormicum; Roche SA, Madrid, Spain), according to patients' needs. Patients were also paralyzed with pancuronium bromide (Pavulón; Organón-Hermes SA, Sant Boi de Llobregat, Spain) to avoid muscular activity during the measurements. All other medications were maintained during the study. The ventilatory conditions were those established by the attending physicians and were kept unchanged during the protocol. We did not set out to assess detailed lung mechanics. Measurements at each time point were as follows: (1) at maintenance FI_O2 (FI_O2-m) (baseline); (2) at 30 and 60 min during 100% O₂ breathing; and (3) 30 min after resuming FI_O2-m; arterial respiratory gases were also measured every 15 min from baseline through 60 min after resuming FI_O2-m. All sets of measurements consisted of the following steps in sequence: simultaneous arterial and mixed venous blood, and mixed expired inert, and respiratory gas sampling; pulmonary and systemic hemodynamics; and ventilatory parameters.

Statistical Analysis

Results are expressed as mean ± SD unless otherwise stated. Comparisons between different inspired oxygen conditions within each subset of patients were done using an analysis of variance (ANOVA) for repeated measurements and Tukey test contrast analysis or Friedman test and Dunn's test for nonparametric data, respectively. Comparisons between baseline variables from each subset of patients (ALI versus COPD) and within ALI patients, with (ALI-ARDS) and without ARDS (ALI-nonARDS), were analyzed using unpaired t or Mann-Whitney tests for nonparametric data. A p < 0.05 was considered significant at all instances.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Ventilatory conditions (respiratory frequency, 14 ± 2 [ALI] and 14 ± 3 min¹ [COPD]; minute ventilation, 9.0 ± 1.7 [ALI] and 9.3 ± 2.4 L · min¹ [COPD]), including positive end-expiratory pressures (PEEP; 3.8 ± 1.8 [ALI] and 0 cm H₂O [COPD]), were kept constant throughout the whole study.

Baseline Conditions

Patients had normal Psa (ALI, 79 ± 11 mm Hg; COPD, 87 ± 18 mm Hg), mild to moderate increases in Ppa (ALI, 24 ± 12 mm Hg; COPD, 27 ± 4 mm Hg) and in pulmonary vascular resistance (ALI, 2.1 ± 1.2 mm Hg · L · min¹; COPD, 3.7 ± 1.5 mm Hg · L · min¹), and normal to reduced T (ALI, 5.5 ± 1.4 L · min¹; COPD, 3.5 ± 0.6 L · min¹) (p < 0.03). Although lung mechanic conditions of the ventilator, namely peak airway (25.4 ± 8.2 [ALI] and 31.0 ± 5.7 cm H₂O [COPD]) and plateau (21.0 ± 7.7 [ALI] and 23.8 ± 5.0 cm H₂O [COPD]) pressures (intrinsic PEEP was not assessed), were not significantly different between the two populations of patients, they were consistent with the underlying disease state of each type of acute respiratory failure. Except for mixed venous oxygen tension (Pv_O2) (p < 0.03), arterial pH and the other respiratory gas parameters were not different between each subset of patients (Table 2).

As expected, patients with ALI exhibited greater moderate to severe intrapulmonary shunt (p < 0.02) than those with COPD; by contrast, the former showed lower dispersions of blood flow (log SDQ; normal < 0.70 [12]) (p < 0.02) and of alveolar ventilation (ln SD; normal < 0.75 [12]) (p < 0.003 than the latter; similarly, mild to moderate high A/ ratios (p < 0.004) and the mean A/ ratio of the ventilation distribution (ALI, 1.64 ± 0.73; COPD, 3.37 ± 1.16) (p < 0.01) were greater in patients with COPD. Other inert gas indices, such as the low A/ mode (alveolar units with low A/ ratios) (ALI, 1.64 ± 0.73; COPD, 3.37 ± 1.16) (p < 0.01), the mean A/ ratio of the blood flow distribution (ALI, 0.80 ± 0.20; COPD, 0.69 ± 0.20), and dead space, mildly to moderately abnormal in both populations, were not significantly different.

Within the cohort of patients with ALI (Table 3), compared with patients with ALI-nonARDS those with ALI-ARDS exhibited lower Pa_O2 (p < 0.005) and higher Pa_CO2 (p < 0.05) and intrapulmonary shunt (p < 0.03); arterial pH was lower and dead space higher nonsignificantly in patients with ALI-ARDS, perhaps indicating a type 2 error. By contrast, no differences in hemodynamics and lung mechanics were observed between the two subgroups of patients with ALI.

Hyperoxic Conditions

There were no significant changes in systemic and pulmonary hemodynamics or in lung mechanics in each population of patients. In patients with ALI, arterial pH decreased at 30 and 60 min compared with baseline values (Table 2); in patients with COPD, there was also a trend to decrease at both time points, possibly because of a type 2 error. Figure 1 illustrates the time courses of the most relevant respiratory and inert gas markers. Both Pa_O2 and the ratio of Pa_O2/FI_O2 rapidly increased and reached a plateau, although less profoundly in patients with ALI; compared with baseline, Pa_CO2 increased in both subsets (Table 2). Similarly, Pv_O2 also increased moderately in both populations. Patients with ALI increased pulmonary shunt, whereas both S/T and log SDQ remained unaltered. By contrast, patients with COPD increased the log SDQ and showed a trend to increase the low A/ mode, without changes in intrapulmonary shunt. These findings were in keeping with those shown previously in patients with advanced COPD and acute exacerbation while breathing 100% O₂ during weaning (8) and recovery (9). Dead space showed a trend to increase in ALI and in COPD. The mean A/ ratios of the blood flow and ventilation distributions remained unchanged. There was a close correlation between intrapulmonary shunt (MIGET) and S/T (O₂ method) at both time points in patients with ALI, but not in those with COPD (Figure 2).

View larger version (14K): [in this window] [in a new window]   Figure 1.   Time courses (mean ± SEM) of PaO2/FIO2 ratios (in mm Hg), intrapulmonary shunt (in percentage of T), and dispersion of blood flow distribution (log SDQ, dimensionless) in both populations of patients (gray circles represent patients with ALI and closed squares those with COPD). Asterisks denote significant differences (p < 0.05) between each time point and baseline value within each subset of patients (see p values in Table 2) (for other abbreviations, see Table 2).

View larger version (20K): [in this window] [in a new window]   Figure 2.   Individual plots of intrapulmonary shunt and venous admixture ratio during 100% O2 breathing (at 30 and 60 min) in both subsets of patients (regression equation in patients with ALI is: shunt = [0.998 ×  S/ T]  2.357). Data in patients with COPD, depicted for comparison but not included in the regression equation, were below the identity line (for symbols, see legend to Figure 1).

Patients with ALI-ARDS increased Pv_O2 and Pa_CO2 and decreased pH significantly at 30 and 60 min, whereas Pa_O2, intrapulmonary shunt, and dead space showed a trend to increase (Table 3). In patients with ALI-nonARDS Pa_O2, Pv_O2, intrapulmonary shunt, and dead space increased significantly at 30 and/or 60 min.

Because Pa_CO2 is proportional to the CO₂ output (times a constant) and inversely proportional to alveolar ventilation (the product of minute ventilation times [1  dead space]), we estimated the changes of Pa_CO2 resulting from the increased dead space from baseline to 100% O₂ conditions. If we assume a CO₂ production of 200 ml/min that did not change and use the measured/calculated values of minute ventilation and inert gas dead space, we could estimate that mean Pa_CO2 increased at 30 min 5.5 mm Hg in patients with ALI, and 3.0 mm Hg in those with COPD.

Posthyperoxic Conditions

Compared with baseline values, arterial pH showed a trend to increase, whereas Pa_O2 and the ratio of Pa_O2/FI_O2 remained lower and Pa_CO2 higher (p < 0.05 each) in the two populations of patients (p values were not significant in COPD) (Table 2 and Figure 1). Whereas in patients with ALI intrapulmonary shunt remained elevated, close to hyperoxic levels (p < 0.05), both log SDQ and the low A/ mode were the two inert gas indices that showed a trend to be increased in patients with COPD. Dead space also showed a trend to remain increased in ALI and in COPD. As in the hyperoxic phase of the study, hemodynamic and lung mechanic parameters did not vary. Both respiratory and inert gas parameters exhibited, in each subgroup of patients with ALI (ALI-ARDS and ALI-nonARDS) (Table 3), a similar trend to that shown by the entire cohort.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The major finding of our study is that, in patients with ALI while breathing 100% O₂, there was a moderate increased of intrapulmonary shunt, that occurred at a fast rate and remained sustained thereafter, without altering the dispersion of pulmonary blood flow. By contrast, the latter A/ descriptor increased in patients with COPD without changes in intrapulmonary shunting. Likewise, there was a parallel increase of dead space in both populations. The two different abnormal A/ responses were accompanied by significant increases of arterial and mixed venous blood oxygenation along with increases of Pa_CO2, without hemodynamic or lung mechanic changes. The gas exchange response of patients with ALI-ARDS was similar to, although slightly more intense than, that shown by those with ALI-nonARDS.

The increment of intrapulmonary shunt in patients with ALI is likely to be related to the development of reabsorption atelectasis resulting from alveolar denitrogenation (1, 2). This has been already postulated in previous studies (1, 13) using, at different levels of FI_O2, measured/calculated respiratory gas parameters, not sufficiently robust to clearly differentiate true intrapulmonary shunt (zero A/ ratios) from areas with low A/ ratios (4). The mechanism of alveolar volume instability during increasing FI_O2 was investigated using lung modeling by the San Diego group in the mid-1970s (2). Under conditions of high FI_O2, it was demonstrated that low inspired A/ ratios, named "critical," could result in a condition of absent expired ventilation, hence inducing atelectasis. Such critical alveolar units become unstable and may ultimately collapse, thereby resulting in the development of atelectasis, a classic concept in pulmonary and critical care medicine (1). It is of note, however, that this situation is in part counterbalanced by the simultaneous increase in Pv_O2, as the number of critical inspired A/ ratios is decreased at each level of FI_O2 (2). Nevertheless, the parallel increase in mixed venous PO₂ also reduces the pulmonary vascular tone (16). Thus, the deleterious effects of pure oxygen inhalation on pulmonary gas exchange are amplified either by increasing intrapulmonary shunt, in ALI, or by mitigating hypoxic pulmonary vasoconstriction (HPV), in COPD.

Similarly, the injurious effects of reabsorption atelectasis on pulmonary gas exchange may be enhanced by the mechanical trigger imposed on terminal airways by mechanical ventilation (17). The repeated opening and closing of distal airways and/or the overexpansion of closed alveolar units may require sufficient tension to generate shear stresses to provoke epithelial disruption on the terminal bronchiolar wall ultimately leading to exposure of the surrounding parenchyma to more inflammatory changes. Moreover, hyperoxic conditions may be an aggravating factor that amplifies the initial injury of mechanical stress (17). In addition, all these events can be aggravated by the compensatory distension of remaining normal lung regions due to atelectasis. This will generate increased transpulmonary pressure, reduced pleural pressure, and increased dead space, even though lung mechanics remain unchanged. Presumably, the low levels of PEEP used in our study also facilitated increased intrapulmonary shunt during pure oxygen inhalation. It might be likely that the application of the "open-lung approach" (18, 19), to achieve optimal recruitment of lung regions, would make the effects of 100% O₂ breathing on pulmonary gas exchange less detrimental. Likewise, it is conceivable that the persistence of both increased intrapulmonary shunt and dead space after ceasing 100% O₂ breathing shown in our study could be related, at least in part, to these hyperoxia-induced mechanistic abnormalities.

Using MIGET and computed tomography (CT) scans of the lungs, hyperoxia-induced increased intrapulmonary shunt has been shown in healthy individuals during general anesthesia (3). To our knowledge, however, this is the first study in patients with ALI to demonstrate that 100% O₂ breathing increases moderately intrapulmonary shunt; furthermore, the increased intrapulmonary shunt observed in the present study did not reverse at once when resuming maintenance FI_O2. The finding that hyperoxia-induced increased intrapulmonary shunt in our study was not accomplished by release of HPV, as reflected by an increase of dispersion of blood flow (ln SD), is consistent with the concept that, at any given level of FI_O2, blood flow cannot be redistributed from areas with shunt or very low A/ ratios, as the resistance of the pulmonary vessels in these units remains unchanged (2). All in all, this gas exchange pattern is akin to the widespread vascular involvement described in patients with ALI-ARDS (20). The small changes of arterial pH cannot be invoked to explain a modulation of the pulmonary vascular tone, even though hypercapnia in experimental oleic acid pulmonary edema has been reported to cause a pH-mediated decrease in HPV (21). Using MIGET, there have been previously very few anecdotal cases showing the effects of 100% O₂ breathing in patients with ALI (22). In a series of patients not clearly defined as ARDS (6), many of them having pneumonia, intrapulmonary shunt remained unaltered during hyperoxic conditions. Efficiency of collateral ventilation, interdependence of the surrounding lung parenchyma, and/or the interaction of mechanical forces exerted during the application of mechanical ventilation, were postulated as potential factors (2). It may also be plausible, however, that those patients with "pulmonary ARDS" (6), is opposed to prevalent edema and alveolar collapse in those with "extrapulmonary ARDS," as suggested by Gattinoni and coworkers (23), are less likely to experience alveolar denitrogenation during hyperoxia (24).

Alternatively, patients with COPD further deteriorated A/ mismatch during 100% O₂ breathing , as shown by a significant increased dispersion of pulmonary blood flow (log SDQ), while intrapulmonary shunt remained unaltered. These findings extend and complement previous data (8, 9) and point to HPV abolition, an effect still shown at 30 min of resuming maintenance FI_O2. The role of HPV to prevent further deterioration of A/ inequality in patients with COPD, irrespective of the severity of airflow obstruction, has been extensively analyzed (25).

It is of note, however, that these two different gas exchange responses took place in the absence of Ppa or pulmonary vascular resistance changes. This may well indicate the greater sensitivity of the dispersion of blood flow assessed by MIGET to the presence of minor vascular changes in the pulmonary circulation, which would never be detected using conventional hemodynamic pressure-flow plots. Similar findings have been observed in patients with acute severe asthma (26) and life-threatening pneumonia (24) requiring mechanical support.

Arterial PCO₂ increased during 100% O₂ breathing in both populations. Conceivably, the increased dead space and the experimental evidence that increased A/ disturbances can worsen not only the O₂ transfer but also CO₂ exchange (27), are behind this increase. Notwithstanding, the Haldane effect (28), namely the changes in the CO₂ dissociation curve facilitating the release of CO₂ from bicarbonate and also from that directly bound as carbamate during 100% O₂, could also enhance the ratio of PCO₂ to blood CO₂ content (29, 30). We estimated, however, that the hyperoxia-induced increments of Pa_CO2 in the two subsets of patients could be attributed almost entirely to the simultaneous increased dead space, thereby indicating a marginal role of the Haldane effect. This was further supported by the persistence of hypercapnia when maintenance FI_O2 was restarted. This minor impact of breathing 100% O₂ on Pa_CO2 in our study is at variance with that shown in patients with acute exacerbations of COPD breathing spontaneously (31). The increased dead space suggests redistribution of pulmonary blood flow from high A/ ratios either to areas with collapsed (nonventilated) units (zero A/ ratios) in ALI, or to poorly but still ventilated units (low A/ ratios) in COPD. Alternative or complementary mechanisms could be, in patients with ALI, overexpansion of remaining normal lung regions caused by collapsed alveoli alluded to and, in patients with COPD, bronchodilatation secondary to the hypercapnia.

Taking all these findings at face value, the present study provides timely information of the deleterious effects of 100% O₂ breathing on intrapulmonary shunt in patients with ALI. This gas exchange response is consistent with the development of reabsorption atelectasis, a finding not rapidly reversed when maintenance FI_O2 is resumed.