INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Mechanical ventilation in patients with respiratory failure caused by the acute respiratory distress syndrome (ARDS) is obviously lifesaving. In recent years, however, it has been realized that the use of large inflation pressures and volumes, which are often needed to normalize arterial blood gases, may aggravate lung injury (1). Accordingly, current recommendations for mechanical ventilation in ARDS include use of the smallest possible tidal volumes (V T) compatible with adequate arterial oxygen tension (Pa_O2), without necessarily attempting to normalize Pa_O2 (2). This strategy is named "permissive hypercapnia" (3).

Although permissive hypercapnia has been reported to increase Pa_O2, the importance of this effect has been variable (4). This is explained by the multiplicity of possible actions of hypercapnia and reduced VT, which include, in addition to the desired prevention of ventilator-induced lung injury, an increased cardiac output () altered regulation of pulmonary perfusion, decreased alveolar ventilation, derecruitment, and the Bohr effect (3). The addition of an increase in positive end-expiratory pressure (PEEP) (6) may further improve Pa_O2. We therefore thought it of interest to investigate the effects of permissive hypercapnia without added PEEP on pulmonary gas exchange, using the multiple inert gas elimination technique (MIGET), an approach that allows quantification of all of the pulmonary and extrapulmonary determinants of arterial oxygenation. We found a consistent pattern of increased shunt (S/T) explained by the combined effects of increased and decreased alveolar ventilation. This finding may have important implications for the practical implementation of permissive hypercapnia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Eight ventilated patients who fulfilled the criteria for ARDS of the American-European Consensus Conference on ARDS (8) were studied within 7 d of ARDS onset. All had a pulmonary artery catheter in place for clinical reasons. Formal contraindications to permissive hypercapnia (3), major hemodynamic instability, or suspected myocardial ischemia were exclusion criteria. The patients consisted of five males and three females with a mean age of 49 yr (range: 18 to 79 yr). All were sedated with midazolam 2 mg/h and morphine 2 mg/h intravenously, and were paralyzed with pancuronium 2 to 4 mg/h intravenously. The origin of ARDS was intrapulmonary in three of the patients, extrapulmonary in three, and possibly mixed in two cases. Five patients were enrolled at Erasmus Hospital in Brussels and three at Lausanne University Hospital. The uniform protocol used in the study was approved by the ethical committees of both institutions. Informed consent was obtained in writing from a next of kin of each patient.

Protocol

The complete study protocol was conducted with the patient in the supine position and under deep sedation and muscle relaxation, using a volume-controlled mode of ventilation with constant inspiratory flow, a fractional inspired oxygen (FI_O2) setting of 0.8, and a constant level of PEEP identical to that ordered by the patient's clinician. The study proceeded in three sequential phases. In Phase 1 (high VT), VT was set to achieve a plateau end-inspiratory pressure (Pplat) of approximately 35 cm H₂O, with an adjustment of respiratory rate (RR) if required to maintain the arterial carbon dioxide tension (Pa_CO2) at the prestudy level. In Phase 2 (low VT), VT was progressively reduced over a period of 1 h without changing RR, until one of the following endpoints was reached: a 50% reduction of VT, an increase in Pa_CO2 > 20 mm Hg, or a Pplat  25 cm H₂O. Because it was anticipated that with these conditions, would be higher than in Phase 1, the protocol was completed with Phase 3 (high VT + dobutamine). For this purpose, the ventilator settings of Phase 1 were progressively reestablished over a period of 1 h until the patient was in the same stable state, as assessed through continuously monitored heart rate and vascular pressures and an arterial blood gas analysis. Thereafter, an intravenous infusion of dobutamine was titrated as needed up to a maximal dose of 10 µg/kg/min in order to obtain the same as in Phase 2.

In each phase of the study the infusion of inert gases was started when the desired steady state was reached, and measurements of expired gases, blood gases, vascular pressures, and were made from 30 to 60 min later. Pressure-volume curves were recorded only in Phases 1 and 3, and not on Phase 2.

Assessment of Ventilation-Perfusion Relationships

The MIGET had previously been used at the bedside by our group (9, 10), and only a brief description of it will be given here. A mixture of six inert gases (sulfur hexafluoride [SF₆], ethane, cyclopropane, halothane, diethylether, and acetone), dissolved in 5% dextrose-in-water, was infused intravenously at a rate of 5 ml/min. Samples of mixed expired air and mixed venous and arterial blood were collected simultaneously 45 min after the infusion was begun. Blood samples (10 ml) were drawn into heparinized supertight syringes (Hamilton). To avoid loss of soluble gases by condensation, mixed expired air was collected with a specially designed heated circuit (tubing and an 18-L mixing box) inserted between the patient and the ventilator (Servo 900C; Siemens, Erlangen, Germany). To prevent the gas-collection circuit from acting as a compressible volume, its inlet was fitted with a low-resistance solenoid valve activated by the ventilator to be closed during inspiration. The blood samples were incubated for 45 min in a heated agitator in the presence of highly pure nitrogen for the extraction of inert gases. Mixed expired air, as well as the nitrogen used to extract inert gases, were transferred to dry, supertight syringes (Hamilton) for the later determination of partial pressures by gas chromatography. In addition, a blood sample from each patient was used to measure the solubility of each gas. From the measured solubilities of the six gases and their concentrations in arterial and mixed venous blood and mixed expired gas, two relationships were developed: the ratio of arterial to mixed venous blood gas concentrations (retention), and the ratio of mixed expired gas to mixed venous blood gas concentrations (excretion), each of which was plotted against the solubility for each gas to derive retention-solubility curves. The representative distributions for blood flow and ventilation were then derived from these curves, using the 50-compartment model of Wagner and coworkers (11). From the recovered distributions, the inert gas S/T, the inert gas dead space-to-V_T ratio (inert gas VDVT), the dispersion of the distribution of ventilation (log SD/A), and the dispersion of the distribution of perfusion (log SD ) were calculated. Log SD A and log SD have an upper limit of normal of 0.6 (12).

The alveolar-arterial (A-a) differences for the different inert gases (retention minus alveolar excretion [R  E] were calculated and plotted as a function of solubility, and were analyzed qualitatively (13), as previously done by our group (10, 12). When the solubility of any gas is expressed in ml gas/100 ml of blood/mm Hg barometric pressure, a given gas permeates a lung area having a A to Q· to ratio corresponding to the numerical value of the gas's solubility. Therefore, an increase (or a decrease) in the A-a difference (i.e., R  E) for a given gas is interpreted as reflecting a deterioration (or an improvement) in A/ matching in a lung area having a A/ corresponding to the solubility of that gas. In normal lung, A-a differences for an inert gas have an upper limit of normal of 0.1 (12).

Hemodynamic and Respiratory Data

Blood pressure (BP) was measured invasively. The pressure traces recorded from the pulmonary artery catheter provided the end-expiratory values of pulmonary artery pressure (Ppa), pulmonary artery occlusion pressure (Ppao), and right atrial pressure (Pra). was measured with the thermodilution method and was divided by body surface area to obtain cardiac index (CI). Pulmonary vascular resistances indexed to body surface area (PVRI) were computed with the standard formula.

Arterial pH and the partial pressures of oxygen and CO₂ in arterial (Pa_O2, Pa_CO2) and mixed venous blood (Pv_O2, Pv_CO2) were measured with an automated blood gas analyzer. The oxygen saturations of arterial (Sa_O2) and mixed venous blood (Sv_O2) were obtained with a Cooximeter (OSM3 Hemoximeter; Radiometer, Copenhagen, Denmark). The samples were drawn anaerobically into minimally heparinized syringes, and the results were read immediately. The concentration of hemoglobin was measured spectrophotometrically after hemolysis and transformation into cyanmethemoglobin. From these data, venous admixture (VA/T) and the arteriovenous difference in oxygen content [C(a-v)O₂] were computed with standard formulas. Expired air was led from the ventilator to a metabolic cart (Medical Graphics Corporation, St. Paul, MN) in order to measure total expired ventilation (E) and CO₂ production (CO₂), and to calculate the Bohr dead space-to-VT ratio (VD/VT Bohr). Because of the high FI_O2 used in the protocol (see the subsequent discussion), the value of oxygen consumption (O₂) provided by the metabolic cart was unreliable. The values of O₂ reported in this study were therefore computed as the product of CI × C(a-v)O₂.

VT, PEEP, and Pplat were measured by means of the pressure and flow sensors built into the Servo 900C ventilator. This machine provides an end-expiratory occlusion mechanism for the measurement of intrinsic PEEP (PEEPi). Airway occlusion can also be timed to end-inspiration, which allows the construction of quasistatic pressure-volume curves of the respiratory system with the method described by Fernandez and coworkers (14). Quasistatic respiratory system compliance (CRS) was calculated as (VT/(Pplat  PEEP  PEEPi), ml/cm H₂O.

Normalization Procedure

To dissociate the effects of increased and decreased VT from those of hypercapnic acidosis per se on pulmonary gas exchange during permissive hypercapnia, we manipulated the mathematical model used in the MIGET as previously reported (12). We recalculated the R  E differences from the data collected during hypercapnia first with constrained at its baseline value (normalization of ), and then with both and VT constrained at their baseline values (normalization of both and VT). The normalization procedure essentially consisted of estimating the mixed venous inert gas tension (Pv) from the arterial (Pa) and the mixed expired (PE) values. Since the MIGET is a steady-state approach, the mass balance can be expressed in the equation:

(1)

and thus:

(2)

where  is the blood-gas partition coefficient and RR is the respiratory rate. From the estimated Pv, retention and excretion could be estimated when and VT were returned to their baseline values in the distribution recorded during Phase 2. It is important to note that: (1) inert gases with low blood-gas partition coefficients will be more influential on Pv; and (2) the normalization procedure does not include any assumption about possible associated changes in regional distributions of and VT.

Data Analysis

The mean ± SEM of all recorded variables was computed. Statistical analysis was done through analysis of variance, with protocol phases as a repeated factor. When the overall effect of phase was significant, pairwise comparisons were made with modified t tests (Fisher's protected t test) (15). The alpha level of all tests was set at 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 shows the ventilatory and hemodynamic data collected in the three phases of the study from Phase 1 to Phase 2, VT and Pplat were reduced, as imposed by the study protocol, and these changes were effectively reversed in Phase 3. PEEPi was minimal and was not significantly altered throughout the study. As expected, CI increased from Phase 1 to Phase 2. This change was substantial (40 ± 9%) and highly significant (p = 0.0003), and could be effectively reproduced by the administration of dobutamine in Phase 3. Mean Ppa () was slightly higher in Phase 2 than in Phases 1 or 3. No significant changes in PVRI could be detected. In the range of tidal excursions prevalent in Phases 1 and 3, the quasistatic pressure-volume curves were linear (i.e., there was no lower inflection point above PEEP and no upper inflection point below Pplat [data not shown]). Quasistatic CRS remained unchanged.

Data relevant to gas exchange are presented in Table 2. According to protocol, Pa_CO2 increased by 12 to 27 mm Hg. On average, the measured FI_CO2 was slightly higher than 0.8. In each patient, it remained absolutely stable throughout the study. The individual responses of Pa_O2 to permissive hypercapnia and to dobutamine were variable (Figure 1). In the group as a whole, Pa_O2 decreased significantly from Phase 1 to Phase 2 (mean change: 17 ± 8 mm Hg, p = 0.04) and remained at that level in Phase 3. The reduction in VT was associated with a significant increase in Pv_O2, which mainly reflected a Bohr effect, in view of the constancy of Sv_O2 and O₂. VA/T was significantly higher in both Phase 2 and Phase 3 than in Phase 1. The Bohr VD/VT was abnormally large throughout the study, and was not affected by changes in VT, in accordance with classical observations (16).

View larger version (15K): [in this window] [in a new window]   Figure 1.   Effects of permissive hypercapnia or dobutamine infusion on PaO2 and inert gas S/ T. Open symbols: mean; dashed line: SE. On average, PaO2 decreased (p < 0.05) from Phase 1 (ventilation with high VT) to Phase 2 (low VT, [i.e., permissive hypercapnia]), and remained at the Phase 2 level in Phase 3 (restoration of high VT plus intravenous infusion of dobutamine). However, the time course of the change in PaO2 was quite variable from patient to patient. In contrast, the behavior of S/ T was consistent: in all subjects it increased substantially from Phase 1 to Phase 2; in all but one subject it decreased from Phase 2 to Phase 3, but remained higher than in Phase 1.

The inert gas VD/VT was in agreement with values reported by others in mechanically ventilated ARDS patients (10, 17), and was substantially below the value obtained from CO₂ excretion (Table 2). The inert gas VD/VT was no more affected by changes in VT than was its Bohr counterpart.

As is usual in ARDS (9, 10, 17), the true shunt measured with inert gases (S/T) was substantial, ranging from 9% to 59% in Phase 1 (lower part of Table 2). There was also an increased inhomogeneity in the distributions of A and , as attested by larger than normal log SD and log SD A values. The presence of shunt and inhomogeneity of A/ accounted for the large discrepancy between Bohr and inert gas VD/VT values, as explained in detail elsewhere (13, 18). With permissive hypercapnia, S/T increased markedly, from 32 ± 6% to 48 ± 5% (p < 0.0001). On return to the high VT with maintained at the permissive hypercapnic level with dobutamine, S/T decreased significantly, to 38 ± 6% (p = 0.001), although it remained higher than in Phase 1 (p = 0.015). Observations in individual patients were remarkably consistent with this mean time course (Figure 1). It is to be noted that the fractional perfusion of lung units with a A/ ratio of between 0.005 and 0.1 did not change throughout the study, being 5.0 ± 2.3% in Phase 1, 3.5 ± 2.0% in Phase 2, and 6.1 ± 2.3% in Phase 3 (p = NS). The indices of A/ dispersion (log SD and log SD A) also did not change significantly throughout the study.

Consideration of the R  E-versus-solubility plot (Figure 2) highlights a further difference between the effects on gas exchange of permissive hypercapnia and dobutamine. The reduction of VT increased R  E only for gases with the lowest solubility in blood (SF6, ethane, and cyclopropane), indicating an increase in S/T and possibly in the perfusion of lung units with very low A/ ( < 0.1). With the restoration of high VT and the infusion of dobutamine (Phase 3), R  E decreased for the gases of very low solubility (SF6, ethane) and increased for those with an intermediate solubility range (halothane). This pattern indicates less shunt and more perfusion to units of intermediate A/ (0.1 to 1.0) in Phase 3 than in Phase 2, although was the same in both conditions. The R  E for all gases except the most soluble ones (ether and acetone) was higher in Phase 3 than in Phase 1. Thus, dobutamine increased shunt to a lesser extent than did permissive hypercapnia, but increased to a greater extent the perfusion to regions of low-to-intermediate A/.

View larger version (19K): [in this window] [in a new window]   Figure 2.   Effects of permissive hypercapnia or dobutamine infusion R  E of inert gases. As compared with baseline with ventilation at a high VT, permissive hypercapnia with a low VT increased R  E only for SF6, ethane, and cyclopropane, indicating an augmentation of S/ T and of A/ mismatch ing in lung units with low A/ (< 0.1). With the restoration of high VT and the infusion of dobutamine, R  E decreased for SF6 and ethane, and increased for halothane, indicating less shunt and a reduction in A/ mismatching in lung units with low A/ (< 0.1), with a slight deterioration in A/ matching in units with intermediate A/ (0.1 to 1.0). R  E was higher with dobutamine infusion than at baseline except for ether. Thus, dobutamine increased shunt to a lesser extent than did permissive hypercapnia, but enhanced to a greater extent the perfusion of regions of low-to-intermediate A/. Data are means ± SE.

The effects of the normalization procedure applied to dissociate the effects of increased and decreased VT from those of hypercapnic acidosis per se on gas exchange are illustrated in Figure 3. Normalization of restored R  E for SF₆ and ethane halfway back to its baseline values. Normalization of both and VT returned the differences in R  E to their baseline values.

View larger version (21K): [in this window] [in a new window]   Figure 3.   Effects of reduction of VT and of increased on R  E of inert gases in permissive hypercapnia. Normalization of in the mathematical lung model of the MIGET fed with the data collected in permissive hypercapnia reduced R  E differences for SF6 and ethane, but these differences remained higher than at baseline. With normalization of both and VT, R  E differences returned to baseline values. These results indicate that an increase in and decrease in VT each contribute to half to the deterioration in A/ matching in permissive hypercapnia. Data are means ± SE.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The recent American-European Consensus Conference on ARDS listed the effects of permissive hypercapnia on pulmonary gas exchange among the questions to be addressed by future research (2). In this respect, the major contributions of the present MIGET study are threefold. First, we found that permissive hypercapnia in ventilated ARDS patients, as induced by reductions in VT and inflation pressures at constant PEEP and reductions in cycling frequency, alters A/ matching, with a large increase in shunt as a major effect. Second, this deleterious consequence of permissive hypercapnia is fully explained by a concomitant increase in and decrease in alveolar ventilation. Third, the magnitude of the deterioration in gas exchange induced by permissive hypercapnia may be largely underestimated or even unsuspected on the basis of arterial Pa_O2 alone.

With institution of permissive hypercapnia in ARDS patients, previous studies have documented large increases in VA/T as measured through blood oximetry with an FI_O2 substantially below 1.0 (venous admixture) (7, 19) or equal to 1.0 ("true" shunt) (20), but the results of such studies have not been uniform (21). Venous admixture cannot discriminate between shunt and units with low A/, and the measurement of true shunt with an F I_O2 of 1.0 has been criticized because the breathing of pure oxygen may in itself influence the A/ distribution (22). The MIGET has none of these limitations. Under the conditions of our study, permissive hypercapnia consistently altered A/ matching with a large increase in S/T (Figure 1).

The mechanical heterogeneity of the lung in ARDS (23) favors regional hyperinflation, leading to the formation of underperfused, well-ventilated lung units, especially with a high VT and high PEEP (17, 24). This could be counteracted with a reduction of VT, potentially decreasing alveolar dead space and the dispersion of A/. Since the inert gas V D/VT in our study was constant (Table 2), the inert gas dead space volume decreased with VT. However, this observation does not necessarily imply a change in the ventilation of unperfused alveoli, because the value of VT may influence other determinants of physiologic dead space. For instance, the efficiency of mixing in the conducting airways improves at low VT (25). In the present study, the dispersion of A/ ratios was not notably modified with permissive hypercapnia, offering little support for the hypothesis of an improved matching of A and in response to reduction of VT. The explanation for this finding could be either that regional hyperinflation did not occur with a high VT or that regional hyperinflation was not reversed by permissive hypercapnia. The general absence of an upper inflection point on the pressure-volume curves recorded in Phases 1 and 3 favors the first possibility (26).

It is well known that intrapulmonary shunt S/T varies directly with , and this has been tentatively explained by an inhibition of hypoxic vasoconstriction due to a combination of increases in PV_O2, Ppa, and pulmonary blood flow (27). Our results suggest that the direct relationship between S/T and is not necessarily explained by changes in distributions of caused by changes in pulmonary vascular tone, but offer no alternative explanation. Permissive hypercapnia usually boosts (7, 20), owing to the combined effects of increased sympathetic tone related to respiratory acidosis and enhancement of venous return by the lower mean intrathoracic pressure (3). The change in intrathoracic pressure appears to be of minor importance, since permissive hypercapnia with small VT but high levels of PEEP still increases (28). Permissive hypercapnia markedly increased in our patients (Table 1), and could therefore, have been responsible for the increase in S/ T noted from Phase 1 to Phase 2 in our study. We incorporated Phase 3 in the study protocol precisely to test this hypothesis. The intravenous infusion of dobutamine under conditions of a high VT to reproduce the recorded in permissive hypercapnia increased S/T halfway from its baseline value. It might be argued that the interpretation of this result would be confounded by the potential effects of dobutamine on pulmonary vascular tone, which could influence the distribution of perfusion independently of . In humans, dobutamine at the doses used in the present study has minimal effects on Ppa, producing either no change or a slight decrease PVRI (29), suggesting mild pulmonary vasodilatation by this agent. Our data are consistent with this pattern (Table 1), which could also account for enhanced perfusion to areas of low-to-intermediate A/ (Figure 2). However, the normalization procedure for reproduced the effects of a real increase in on S/T, suggesting that dobutamine had no effect on pulmonary vascular tone that could have affected the distribution of perfusion. This result is in keeping with the previous observation in intact dogs that dobutamine at doses below 10 µg/kg/ min does not affect pulmonary vascular tone as defined by pulmonary vascular pressure measurements at a controlled in hyperoxia or hypoxia (30). Thus, an increased resulting from permissive hypercapnia accounted for about half of the associated increase in true pulmonary shunt.

What other factors could have caused the increased pulmonary S/T induced by permissive hypercapnia? Concordant information from animal models of acute lung injury suggests that respiratory acidosis either does not affect or improves A/ matching (31, 32). We tested the hypothesis that alveolar hypoventilation resulting from permissive hypercapnia might contribute to an increase in shunt. The normalization procedure for V T showed that this was indeed the case (Figure 3). Alveolar hypoventilation is not a classically recognized cause of true pulmonary shunt. However, a decrease in ventilation in units with very low A/ decreases the value of the ratio proportionally much more than would an identical increase in perfusion. In other words, it is conceivable that decreased ventilation in units with very low A/ would convert them into units that, although still aerated, could not, with the MIGET, be distinguished from units having true shunt. Derecruitment of units receiving a disproportionately small fraction of V T, and favored by ventilation with a high FI_O2 (22, 33- 35), could have occurred despite the absence of a detectable change in CRS (Table 1). However, the results of the normalization procedure for alveolar ventilation argue against this explanation in the present study.

Whatever its mechanism, the massive increase in S/T caused by permissive hypercapnia in our study will be of interest to the clinician in that it was not paralleled by a decrease in Pa_O2 of the same relative magnitude (Table 2). In isolated instances, Pa_O2 even increased (Figure 1). This apparent paradox is easily explained with classical physiologic concepts. In part, it is related to the inverse relationship that exists in the steady state between Pa_O2 and C(a-v)O₂ at any constant level of shunt (36). With permissive hypercapnia, O₂ did not change (Table 2) and increased (Table 1), implying that C(a-v)O₂ decreased, thus partly or fully offsetting the effects of the higher shunt on arterial oxygenation. In addition, acute respiratory acidosis shifts the oxyhemoglobin dissociation curve to the right (Bohr effect), which increases Pa_O2 at a given level of O₂ and S/T.