INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Noninvasive ventilation (NIV) delivered via a face mask is being increasingly used in patients with acute exacerbations of chronic obstructive pulmonary disease (COPD). The physiologic and therapeutic effects of NIV used with a mixture of air and oxygen have been studied in various forms of acute respiratory failure (1). NIV acts primarily by reducing patient effort and work of breathing, at least in patients with acute exacerbations of chronic respiratory failure (1, 5). In selected series of acute COPD exacerbations, NIV has been shown to reduce morbidity, duration of hospital stay, and mortality (1, 3, 6). NIV has limitations, however, and is not successful in every patient. Clinical tolerance may be a key determinant in the success or failure of NIV (7). High levels of pressure are sometimes required to improve NIV efficiency but carry a risk of decreased face mask tolerance and increased leaking (5). Thus, pressure in the mask must be kept within a range consistent with both efficacy and tolerance. To extend the beneficial effects of NIV to a larger group of patients, including more patients with critical conditions, a potential solution may be to substitute a helium- oxygen mixture for the standard air-oxygen mixture during NIV. The low density of helium-oxygen as compared with air is associated with decreased resistance to gas flow (8). Helium-oxygen may enhance the unloading effects of NIV, thus facilitating both the patient's inspiratory muscle activity and the generation of NIV. Indeed, the internal load caused by airway resistance is high in patients with stable COPD and increases further during acute exacerbations. Previous studies have found some evidence that breathing helium-oxygen may reduce dyspnea and improve gas exchange in nonintubated patients with exacerbations of COPD (11) or with severe asthma (12), but none of these studies evaluated the effect of helium-oxygen versus air-oxygen on work of breathing. Combining NIV with a gas mixture lighter than the usual air-oxygen mixture, may thus improve the efficacy of NIV.

We report a study of the short-term physiologic effects of noninvasive ventilation used with a helium-oxygen mixture (NIV-HeO₂) in a group of 10 patients with acute exacerbations of COPD. Patient effort, work of breathing, and gas exchange were compared between NIV-HeO₂ and conventional NIV-AirO₂ at two predetermined levels of pressure support, (1) a low level intended to simulate unassisted spontaneous breathing and (2) a high level associated with effective pressure support.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

The experimental protocol was approved by an institutional review board for human subjects (Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale Créteil-Henri Mondor). Written informed consent was obtained from each study patient. We studied 10 consecutive patients who were either admitted to the medical intensive care unit of our institution for mild-to-moderate acute exacerbation of COPD (n = 7) or who developed mild-to-moderate respiratory distress after endotracheal tube removal following a period of mechanical ventilation (n = 3). The study was performed between July 1998 and February 1999. Patients enrolled in the study had definite or highly probable COPD based on medical history, physical examination, chest radiograph, and arterial blood gas findings (Table 1). Inclusion criteria were a recent exacerbation of dyspnea as the cause of admission or as a complication of extubation and at least one of the following: respiratory rate above 25 breaths/min, partial pressure of arterial oxygen below 60 mm Hg while breathing a low flow of oxygen, and arterial pH below 7.38. Patients with evidence of encephalopathy were not included. Exclusion criteria were as follows: need for immediate endotracheal intubation; respiratory rate below 12 breaths/ min; central nervous system disorders; hypercapnic encephalopathy; severe hypoxemia defined as a need for an FI_O2  0.5 or more to obtain Sa_O2  85% or  90% in patients with and without chronic hypoxemia before the episode, respectively; pneumothorax, hemodynamic instability, or enrollment in another investigative protocol.

Protocol

Patients were studied in the semirecumbent or sitting position. Two gas mixtures were tested, air-oxygen (AirO₂) and helium-oxygen (HeO₂). FI_O2 (25% to 40%) was adjusted in each patient to maintain Sa_O2 > 90% and was kept similar with the two gas mixtures. Two levels of pressure support ventilation (PSV) were used with each gas mixture. The low PSV level was intended to simulate conditions during unassisted spontaneous breathing (SB) and was consequently chosen as equal to or slightly greater than the pressure needed to compensate for the dead space and the resistive load imposed by the circuit and valve, i.e., 5 to 10 cm H₂O (15). The low PSV level was determined on a case-by-case basis as the level providing on-line monitored esophageal and/or transdiaphragmatic pressure swings of approximately the same magnitude as during SB, i.e., with no facial mask or equipment. However, to improve patient comfort and facilitate the experiment, this level was often increased by 2 to 3 cm H₂O. We used a low level of PSV instead of SB because this allowed us to record the breathing pattern, to keep the same apparatus in all conditions, and to be in a position to improve patient comfort when needed. The high PSV level was determined as the level providing a larger exhaled tidal volume (VT > 6 ml/kg), a respiratory rate below 35 breaths/min, and improved patient comfort. To ensure that any differences between the AirO₂ and HeO₂ conditions would be due only to the gas mixture difference, the same experimental apparatus was used in all four conditions (see Figure 1), and at each PSV level the two gas mixtures (AirO₂ and HeO₂) were delivered using the same ventilator, circuit, mode of ventilation, and patient equipment. In all four conditions, each period lasted 20 min. The two gas mixtures and the two PSV levels were tested in a random order. The four conditions studied were as follows: (1) SB-AirO₂ with a low PSV level (5 to 10 cm H₂O), (2) SB-HeO₂ with a low PSV level (5 to 10 cm H₂O), (3) NIV-AirO₂ with a high PSV level (15 to 25 cm H₂O), and (4) NIV-HeO₂ with a high PSV level (15 to 25 cm H₂O). Although the patients were not told which gas mixture they were inhaling, some detected the use of HeO₂ because they happened to talk to the investigator while breathing the mixture and found they had a "Donald Duck" voice.

View larger version (29K): [in this window] [in a new window]   Figure 1.   Experimental apparatus.

Experimental Apparatus

Gas mixture administration. As shown in Figure 1, the gas mixtures were administered through the ventilator. The helium-oxygen mixture, composed of 22% oxygen and 78% helium, was provided by a high-pressure source (cylinder of 200 bar-compressed gas; Air Liquide Medical, Bonnevil S/Marne, France). The gas was depressurized through a specific pressure regulator (HBS 315-8 no. 20 305; 4 bar) connected to the air inlet of the ventilator. The second gas inlet (oxygen) was connected to the wall oxygen supply to provide O₂-enrichment of the gas mixture and the desired FI_O2 as described below.

NIV was delivered through an insulated face mask (48988; Peters, Bobigny, France), connected to an antibacterial filter (PALL BB25, East Hills, NY) with a low internal dead space (< 30 ml).

Ventilator. The same ventilator, a prototype specially designed for this study, was used in all patients. This was a HORUS ventiltor (V2.110) adapted by the manufacturer (Taema, Antony, France) for helium-oxygen mixture use. Although the spirometric parameters measured by the ventilator were not used in this study and did not influence ventilator functioning, the breaths were flow-triggered, which required correction for any offset in the flow signal caused by the gas mixture. Because the zero of the hot-wire flowmeter was different with HeO₂ and AirO₂, requiring a specific electrical adjustment, such a procedure was carried out before each gas mixture switch. Preliminary measurements in a physical model have shown that the calibration coefficient of the hot-wire flowmeter was slightly modified in HeO₂ compared with AirO₂. Using the calibration coefficient defined for AirO₂ with an HeO₂ mixture resulted in systematic ventilator flow overestimation by 20% (16). With regard to the low flow range used for the triggering function, we have considered that the flow-based inspiratory and expiratory triggering functions used with this ventilator were similar with HeO₂ and AirO₂ once the zero was corrected. We verified in an in vitro study (data not shown) conducted using a standard bench test (17) that the main performance characteristics of this ventilator in terms of both flow-triggering and pressure generation were not substantially different between HeO₂ and AirO₂, provided the zero of the hot-wire flowmeter, needed for the triggering function, was electrically adjusted for each gas mixture prior to the experiment.

Ventilator settings. Both for HeO₂ and AirO₂, the ventilator was adjusted at its optimal configuration. The inspiratory flow trigger was set at its minimum level, i.e., to maximum sensitivity without auto-triggering (0.1 to 1.5 L/min). The expiratory flow trigger was adapted to the level of leaking based on patient comfort. The pressure rise slope was set close to the maximum value to ensure rapid achievement of the present pressure plateau. Because of the working principles of this prototype ventilator, no external positive end-expiratory pressure (PEEP) was used in any patient. FI_O2 was maintained constant throughout the study and was not allowed to exceed 40% because at higher concentrations the physical properties of the HeO₂ mixture become increasingly similar to those of AirO₂.

Measurements

In each condition, data were recorded during a 5-min period after a 15-min breathing pattern stabilization period. Flow was measured using a Fleish no. 1 pneumotachograph (Zürich, Switzerland) connected to a differential pressure transducer (MP45, ± 2 cm H₂O; Validyne, Northridge, CA) and located in the ventilator circuit, between the filter and the Y connector (see Figure 1). Preliminary static calibration in HeO₂ and in AirO₂ was performed routinely prior to experiments, using a standard volumetric method. For each gas mixture tested, the same volume of gas (1 L) was slowly pushed through the pneumotachograph using a tight syringe. The gain of the pressure transducer coupled to the flowmeter for each gas mixture was referenced to the integrated flow signal to obtain 1 Volt = 1 L. The gain was 1 for AirO₂ and 0.813 for HeO₂. The ratio between the gain with HeO₂ over the gain with AirO₂ (0.813) was similar to the ratio between dynamic viscosity in HeO₂ and in AirO₂. We first checked that the response of the pneumotachograph remained linear throughout the range of the flow rates used (0 to 1.5 L/s) both with AirO₂ and with HeO₂. It has been shown that using HeO₂ with a laminar flowmeter tends to extend the range of linearity by significantly decreasing Reynold's number in pneumotachograph capillaries (18). We found that the Fleish no. 1 pneumotachograph used in our study had a strictly linear response with HeO₂ up to at least 2.5 L/s. The best fit regression performed on the P-flow relationships for the Fleish no. 1 pneumotachograph from 0 to 2.5 L/s was described by a polynomial regression for AirO₂ (r² = 0.9963) and by a linear regression for HeO₂ (r² = 0.9988). The flow signal was integrated to yield VT.

Airway pressure (Paw) was measured using a differential pressure transducer (MP45, ± 100 cm H₂O; Validyne) located between the filter and the pneumotachograph. Esophageal (Pes) and gastric (Pga) pressures were recorded using a double-balloon catheter (Marquat, Boissy Saint Léger, France) inserted through the nose after topical anesthetic application, and advanced until the distal balloon was in the stomach and the proximal balloon in the middle of the esophagus. Each balloon, filled with 1 ml of air, was connected to a differential pressure transducer (MP45, ± 100 cm H₂O; Validyne). Esophageal balloon position was checked using the occlusion method (19). Placement of the gastric balloon was considered adequate if gentle manual pressure on the patient's abdomen produced Pga fluctuations and if swallowing by the patient produced a sharp increase in Pes caused by esophageal contraction with no increase in Pga. Transdiaphragmatic pressure (Pdi) was obtained by subtracting the Pes signal from the Pga signal, and transpulmonary (Ptp) pressure by subtracting the Pes signal from the Paw signal. Pressure and flow signals were digitized at 128 Hz and sampled using an analogic/numeric system (MP100; Biopac Systems, Santa Barbara, CA).

An arterial line was inserted in a radial artery to allow arterial blood gas analysis at the end of each test period. Blood gases were measured using an ABL 520 analyzer (Radiometer, Copenhagen, Denmark). Heart rate (HR) and rhythm were continuously monitored by standard three-lead monitoring electrodes, and systolic (SBP) and diastolic (DBP) arterial blood pressures by the radial arterial catheter. Oxygen saturation (Sa_O2) was also continuously monitored.

Data Analysis and Assessment of Patient Effort

Breathing pattern and minute ventilation (E) were determined from flow tracings. Inspiratory work of breathing (WOB) performed by the patient was computed from esophageal pressure and tidal volume loops as previously described (15). In brief, the inspiratory work per breath was calculated from a Campbell diagram by computing the area enclosed between the inspiratory esophageal pressure-tidal volume curve on the one hand and the static esophageal pressure-volume curve of the chest wall on the other hand, using a theoretical value for chest wall compliance (4% of the predicted value of the vital capacity per cm H₂O). Although we directed careful attention to minimizing leaks around the mask, this problem did occur and was taken into consideration. Because leaks around the mask are more likely to occur during inspiration, leading to overestimation of the inspired volume, we calculated inspiratory WOB by applying a correcting factor to the inspiratory flow based on the patient's expired E. For this, we measured the ratio of expired over inspired minute volume and applied a correction factor equal to this ratio to the flow signal used to measure inspiratory WOB. Inspiratory WOB was expressed as the work per breath (joules per breath, J/breath), as the work per volume unit (joules per liter, J/L), or as the work per time unit (joules per minute, J/min). Total transpulmonary work of breathing (tpWOBtot) was calculated from the transpulmonary pressure and expressed in J/L. The resistive (tpWOBres) and elastic (tpWOBel) components of tpWOBtot were also calculated, and expressed as the per cent of tpWOBtot. The Pes values at times with zero-flow were considered as indicating the beginning and end of inspiration. Use of a theoretical value for chest wall compliance (4% of the vital capacity per cm H₂O) probably results in some degree of error, which is, however, the same for the various periods compared and, consequently, does not invalidate comparisons. The relaxation curve of the chest wall was superimposed on the complete diagram, assuming that the end-expiratory elastic recoil pressure of the chest wall was equivalent to the Pes level at the beginning of the inspiratory effort. The beginning of the sharp negative deflection of the Pes curve was taken as the onset of the inspiratory effort. Any difference between this initial Pes level and the zero-flow point indicated the presence of intrinsic positive end-expiratory pressure (PEEPi) (20, 21). This PEEPi value was corrected for the presence of expiratory muscle activity during expiration as detected on Pga tracings, according to the method developed by Lessard and colleagues (22). To enhance the accuracy of inspiratory effort estimation, transdiaphragmatic pressure (Pdi) and the esophageal and transdiaphragmatic pressure-time products (PTP) were also measured. These parameters are independent from flow and volume signals. The esophageal PTP per breath (PTPeso/breath) was obtained by measuring the area under the Pes signal between the onset and the end of inspiration, and was referenced to the chest wall static recoil pressure- time relationship (23). The same principles were applied to PEEPi and expiratory muscle activity. The PTP per breath for the diaphragm (PTPdi/breath) was obtained by measuring the area under the Pdi signal from the onset of its positive inflection to its return to baseline. After elimination of artifacts caused by coughing, excessively small tidal volumes, and esophageal spasms, 10 to 30 consecutive breaths were used to compute average values.

Statistical Analysis

Data are reported as mean ± SD. The measurements obtained in the four conditions studied were compared, with each patient serving as his or her own control. Pairwise comparisons were done using Wilcoxon's nonparametric test for each PSV level; the Kruskal-Wallis test was used to compare the effects of HeO₂ alone (SB-HeO₂), NIV-AirO₂, and NIV-HeO₂ to the basal condition (SB-AirO₂); p values below 0.05 were considered statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Ten consecutive patients tolerated NIV well and were eligible for inclusion in the study. One additional patient declined to participate in the study because she was unwilling to undergo insertion of a balloon in the esophagus. Another patient (Patient 4) had a poor clinical tolerance that did not allow use of the low PSV level (SB), and, consequently, results with this level are reported for only nine patients. Characteristics of the 10 patients enrolled in this study are listed in Table 1. All but one had hypercapnia. Acidosis ( 7.38) was present in six; one had metabolic alcalosis related to diuretic treatment. Three patients (Patients 2, 7, and 9) were included immediately after tracheal extubation; their respiratory status was in a transitional phase between acute decompensation and the basal state of their disease. The following mean levels of PSV were used: 9 ± 2 cm H₂O for the low PSV level (SB) and 18 ± 3 cm H₂O for the high PSV level (NIV). Mean FI_O2 was 32 ± 4%.

Breathing Pattern, Gas Exchange, and Hemodynamics

The mean values of the main ventilatory parameters are presented in Table 2. The ratio of expired over inspired minute volumes, used in the WOB calculation to correct for leaks around the mask, did not differ significantly among the study conditions: mean values were 0.81 ± 0.17 and 0.77 ± 0.15 at the low PSV level with AirO₂ and HeO₂, respectively, and 0.78 ± 0.10 and 0.76 ± 0.13 at the high PSV level with AirO₂ and HeO₂, respectively. In other terms, using "uncorrected" WOB values would have resulted in similar differences between HeO₂ and AirO₂. Decreasing gas density by substituting HeO₂ for AirO₂ did not significantly modify breathing pattern parameters, whereas increasing the PSV level induced significant increases in VT, E, and VT/TI with both gas mixtures.

Blood gas and hemodynamic parameter values are reported in Table 3. Increasing the PSV level reduced Pa_CO2 and, at each PSV level, substituting HeO₂ for AirO₂ significantly reduced Pa_CO2 and increased arterial pH. Individual Pa_CO2 variations are shown in Figure 2. The reduction in Pa_CO2 was largest during NIV-HeO₂. Interestingly, the Pa_CO2 reduction induced by NIV-HeO₂ was well correlated with the basal Pa_CO2 value (SB-AirO₂) (Figure 3). Oxygenation and hemodynamic parameters were not significantly different across the four study conditions.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Individual differences in PaCO2 between AirO2 and HeO2. Nine patients were studied with the low pressure support ventilation (PSV) level and 10 were studied with the high PSV level. Hor-izontal bars indicate mean values *p < 0.05.

View larger version (9K): [in this window] [in a new window]   Figure 3.   Correlation between the PaCO2 decrease induced by NIV-HeO2 compared with baseline (spontaneous breathing with low pressure support with AirO2) and the PaCO2 at baseline.

Work and Effort to Breathe

The patients' inspiratory effort parameters are displayed in Table 4. All indices of respiratory effort were decreased with HeO₂ as compared with AirO₂ at the same PSV level (SB and NIV, Table 4 and Figures 4 and 7). Individual values for WOB (in J/min), PTPeso (in cm H₂O/s/min), and for Pdi (in cm H₂O) are shown in Figures 4, 5, and 6, respectively. The relative changes in WOB (in J/L) and Pdi (in cm H₂O) expressed as the percent of values measured during SB-AirO₂ (taken to replicate the basal state) are presented in Figure 7 for the other three conditions, namely, SB-HeO₂, which illustrated the effect of reducing gas density; NIV-AirO₂, which illustrated the effect of increasing PSV; and NIV-HeO₂, which illustrated the effect of both changes taken together. The largest reduction in patient effort to breathe was provided by the combination of both changes. The changes in Pes and Pdi swings with HeO₂ in Patient 5 are illustrated in Figure 8.

View larger version (15K): [in this window] [in a new window]   Figure 4.   Individual changes in WOB (J/min) in response to each condition. Nine patients were studied at the low pressure support ventilation (PSV) level and 10 were studied at the high PSV level. Horizontal bars indicate mean values. *p < 0.05; **p < 0.01.

View larger version (13K): [in this window] [in a new window]   Figure 7.   Relative changes in work of breathing (WOB) in J/L (left panel ) and in tidal transdiaphragmatic pressure swings (Pdi) in cm H2O (right panel ). SB-HeO2 = spontaneous breathing simulation with the helium-oxygen mixture. NIV-AirO2 = noninvasive ventilation with the air-oxygen mixture. NIV-HeO2 = noninvasive ventilation with the helium-oxygen mixture. Each of these conditions was compared with spontaneous breathing simulation with the air-oxygen mixture (used to simulate the basal state). *p < 0.05; **p < 0.01.

View larger version (15K): [in this window] [in a new window]   Figure 5.   Individual changes in the esophageal pressure-time product in response to each condition. Nine patients were studied at the low pressure support ventilation (PSV) level and 10 were studied at the high PSV level. Horizontal bars indicate mean values. *p < 0.05; **p < 0.01.

View larger version (15K): [in this window] [in a new window]   Figure 6.   Individual changes in transdiaphragmatic pressure in response to each condition. Nine patients were studied at the low pressure support ventilation (PSV) level and 10 were studied at the high PSV level. Horizontal bars indicate mean values. *p < 0.05; **p < 0.01.

View larger version (16K): [in this window] [in a new window]   Figure 8.   Tracings from Patient 5 showing that Pes and Pdi swings were smaller with HeO2 than with AirO2 during both simulated spontaneous breathing (low pressure support ventilation [PSV] level) and NIV (high PSV level).

Total transpulmonary work of breathing (tpWOBtot) and the partition into its resistive and elastic components performed by the "patient-ventilator" couple are presented in Table 5. HeO₂ induced a significant decrease in tpWOBtot expressed in J/L in both the SB and the NIV conditions. Interestingly, the partition of tpWOBtot into its resistive and elastic components was not influenced by the nature of the gas mixture (Table 5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study is, to our knowledge, the first evaluation of the effects on work of breathing and blood gases of a helium-oxygen mixture (HeO₂) used with noninvasive pressure support ventilation (NIV) in patients with acute COPD exacerbation. We found that substituting HeO₂ for AirO₂ enhanced the efficacy of NIV in this setting, providing a larger increase in pH, a larger decrease in Pa_CO2, and a larger reduction in patient inspiratory effort. These data suggest that the use of HeO₂ mixture may allow a larger population of patients to benefit from NIV.

Most of the recent clinical studies on the pathophysiology of severe COPD has shown that the respiratory muscle pump often functions close to its maximal capacity. This is because the respiratory muscles operate at a mechanical disadvantage because of the high resistive and elastic loads imposed on them as a result of airway obstruction and lung hyperinflation. An acute exacerbation of COPD imposes an additional load on the respiratory muscles, with the result that the patient can no longer maintain an effective alveolar ventilation and, therefore, experiences worsening hypoxia, hypercapnia, and acidosis, with the development of a vicious circle (24). The main clinical advantage afforded by NIV with a conventional air-oxygen mixture (NIV-AirO₂) is to stop or prevent the development of this vicious circle without requiring endotracheal intubation. In appropriately selected patients with COPD, NIV-AirO₂ has been found to provide better outcomes than intubation and mechanical ventilation (2, 3). A few studies have found that both asthmatics and patients with COPD benefited from HeO₂ administration, presumably because of the reduction in resistive load afforded by the low density property of helium (9, 10, 25). Using HeO₂ during mechanical ventilation may also be beneficial. This suggestion is based on preliminary data obtained in intubated asthmatics (26) and patients with COPD (27) receiving conventional mechanical ventilation. A decrease in the work of breathing with HeO₂ has recently been reported during T-piece weaning procedures in patients with COPD (28). To our knowledge, only one study has evaluated the short-term effect of NIV with HeO₂ (29) in patients with COPD, and found that NIV with HeO₂ was associated with greater reductions in both dyspnea and Pa_CO2 than NIV with AirO₂. However, no measurement of patient effort was reported, and neither were the respective beneficial effects of NIV and HeO₂ assessed. In the present study, we assessed not only the effects of NIV with HeO₂ but also those of NIV alone and HeO₂ alone.

Because the experimental setup was the same in all the study conditions, it is reasonable to assume that observed differences between AirO₂ and HeO₂ at a given PSV level were due to the difference of gas mixtures. To ensure the reliability of our measurements of work of breathing, gas flow, and volume, we directed special attention to NIV application and took into account during our measurements the potential problems raised by differences in the physical properties of HeO₂ and AirO₂. At each gas mixture change, the gas in the circuits was flushed out, and the pneumotachograph was calibrated; in addition, a correction procedure was used to allow the ventilator to work with HeO₂. We also estimated work of breathing by correcting the flow signal for inspiratory leaks, i.e., by forcing the inspiratory tidal volume value to be equal to the expiratory tidal volume value. The amount of leaking estimated based on the ratio of expired over inspired volume did not change between AirO₂ and HeO₂, indicating that "uncorrected" WOB measurements would have provided similar results.

The central message of this study is that, as compared with a basal state stimulated by a low PSV level (SB-AirO₂), the beneficial effects on pH, Pa_CO2, and inspiratory effort recorded with either NIV-AirO₂ or SB-HeO₂ were further increased by combining NIV-HeO₂. The inspiratory effort decrease was responsible for decreases in the values of all the indices used to evaluate respiratory effort (Tables 4 and 5 and Figures 4-7). Compared with the basal state (SB-AirO₂), respiratory effort indices fell by about 50% with NIV-HeO₂ versus only 35% with NIV-AirO₂ and 15% with SB-HeO₂ (n = 9). Our results with SB-HeO₂ are consistent with an early study by Swidwa and colleagues (11) in which indirect evidence indicated a decrease in mechanical work of breathing: in nonintubated patients with decompensated COPD, breathing HeO₂ was associated with changes in FRC, Pa_CO2, and CO₂ production but did not modify the breathing pattern. Swidwa and colleagues suggested that the 15% FRC decrease provided in their study by substituting HeO₂ for AirO₂ may lead to less hyperinflation, thereby improving respiratory muscle function and work of breathing. We quantitatively estimated the reduction in respiratory effort provided by HeO₂ in patients with COPD, both during simulated spontaneous breathing and during NIV. Furthermore, our data with AirO₂ are in full agreement with a report by Brochard and colleagues (1) of a 36% decrease in the pressure-time product index with NIV, at equivalent PSV levels. We found a significant reduction in the total transpulmonary work of breathing with HeO₂ because of similar reduction in the resistive and elastic components of this parameter. Several characteristics of fluid mechanics may explain a concomitant decrease in the resistive and elastic components of the work of breathing. These characteristics are all due to the fact that the density of the tested HeO₂ mixture used in our study (FI_O2 = 32%) was about three times smaller than the density of air. This reduced density (), together with the almost negligible increase in the dynamic viscosity (µ) of HeO₂ as compared with air ( 10%), results in an increase in kinematic viscosity  (= µ/) and in a decrease in the reynolds number (Re). This Re decrease with HeO₂ has the following effects: (1) flow becomes laminar over a larger number of airway generations, and (2) laminar entrance effects (also called developing flows) generated by the interactions between airway bifurcations, curvature, and gas flow develop over a longer distance in each airway than when flow is turbulent (9). Importantly, laminar flow remains density-dependent when fluid flows in a complex geometry such as that of the branching airway tree. Using the formula developed by Pedley (9) for predicting airway pressure drops in the laminar range, we calculated the pressure drop associated with use of 65% helium-35% oxygen and with 65% air-35% oxygen; results suggested that the airway resistance decrease with the HeO₂ mixture was about 38% for identical steady flows of the two gas mixtures. This suggests that (1) HeO₂ may decrease the resistance of airways in series, thereby reducing the resistive work of breathing, and that (2) the redistribution of gas throughout the lung by airway resistances in parallel may differ between HeO₂ and AirO₂. This latter dynamic phenomenon, which may occur in the respiratory rates seen in our patients (30), is probably ascribable to reductions (or modifications) in time constants (resistance times compliance) with the lighter gas and may have contributed to decrease the elastic work of breathing, together with other factors such as the nonsignificant trends toward PEEPi diminution and dynamic compliance augmentation. The lower Pa_CO2 value obtained with HeO₂ may be ascribable to a decrease in CO₂ production or to an increase in alveolar ventilation. In our study, minute ventilation did not change between AirO₂ and HeO₂, suggesting that alveolar ventilation may have increased as a result of reduced dead space. Indeed, the transport and/or mixing of carbon dioxide in small conducting airways is facilitated by diffusion. Because helium enhances the diffusion of carbon dioxide, particularly when convective gas velocities are small, use of HeO₂ is probably associated with a reduced dead space (9, 10). Moreover, the decrease in mechanical heterogeneity seen with HeO₂ may also reduce the pendelluft phenomenon, thereby improving CO₂ elimination in the intermediate regions of the lung. Using a multiple inert gas elimination technique, Christopherson and Hlastala (31) showed that the / balance improved in dogs breathing HeO₂. Accordingly, when patients were breathing this lighter gas, an improvement in / can be expected. These factors may play a role in patients with COPD. Also, respiratory muscle unloading may be associated with a decrease in CO₂ production, which may also have contributed to the reduction in Pa_CO2 at constant E in our study.

Another means of improving the efficacy of NIV in patients with COPD is to increase the level of PSV. However, high PSV levels can result in hyperinflation and barotrauma because of the delivery of high pressures at the end of inspiration (32). Also, high PSV levels can produce desynchronization between the patient's spontaneous breathing and the ventilator (33). In addition, increasing the PSV level may result in a very rapid flow acceleration potentially responsible for poor tolerance. High PSV levels also increase leaking around the face mask. Similar levels of unloading can be obtained at lower PSV levels by substituting HeO₂ for AirO₂ during NIV. The behavior of the respiratory control system differs between unloading achieved by use of HeO₂ and unloading achieved by use of higher PSV levels. The respiratory control system has two options for taking benefit of the unloading provided by HeO₂: it can either maintain the central respiratory output constant to produce more ventilation for the same effort or maintain ventilation constant to reduce respiratory muscle work. Both E and Pa_O2 remained unchanged in our study (see Tables 2 and 3) with HeO₂ (in both the SB and the NIV conditions), suggesting that the respiratory control system chose to reduce respiratory muscle work while maintaining ventilation constant (RR, VT, E). We investigated correlations between basal pH or Pa_CO2 levels and the response to NIV-HeO₂ in terms of patient effort. We found no significant correlations between the WOB decrease with HeO₂ (WOB) and the basal pH or Pa_CO2 value. However, this finding may be ascribable to the small number of patients (nine at both levels) and to the presence in one patient (Patient 5) of metabolic alcalosis. After exclusion of Patient 5, we found that three of the four patients with a pH about 7.35 had a WOB < 35%, whereas the four patients with a pH below 7.35 had a WOB > 35% ranging from 38 to 64%. Our data also suggest that patients who cannot benefit from NIV-AirO₂ therapy, i.e., 25 to 40% of patients with COPD according to previous studies (2, 24), may respond better to NIV-HeO₂. In particular, patients who fail to respond to or who require but cannot tolerate high PSV levels may be good candidates for NIV-HeO₂. One limitation of HeO₂ therapy, however, is that the physical properties of HeO₂ mixtures become increasingly similar to those of AirO₂ mixtures when the concentration of O₂ exceeds 40%.

In conclusion, we found that the improvements in pH, Pa_CO2, and work of breathing provided by noninvasive pressure support ventilation with AirO₂ or by HeO₂ breathing alone can be increased by coupling the two methods, i.e., using HeO₂ ventilation during noninvasive pressure support ventilation. Our data indicate a need for further studies aimed at defining the role for NIV-HeO₂ as compared with NIV-AirO₂. Use of NIV-HeO₂ may provide greater reductions in the need for intubation as compared with NIV-AirO₂. In the subset of patients who fail to improve with NIV alone, adding HeO₂ may improve clinical tolerance and potentially further reduce the need for invasive ventilation in selected patients.