INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Reducing the density of inspired gas by using a mixture of helium and O₂ (HeO₂) instead of air and O₂ (airO₂) can be beneficial in spontaneously breathing patients with upper or lower airway obstructive disease (1, 2). In acute severe asthma, breathing HeO₂ increases peak expiratory flow and Pa_O2, and decreases pulsus paradoxus, Pa_CO2, and dyspnea (3). In patients with COPD, breathing HeO₂ increases expiratory flow and decreases airway resistance (6, 7). Moreover, evidence suggests that these favorable effects can also be obtained during mechanical ventilation in status asthmaticus (8) and in patients with COPD (9, 10), even though the data are still preliminary in the latter patient population, and await confirmation by prospective studies. However, the physical properties of helium could interfere with several key ventilator functions such as gas mixing, inspiratory and expiratory valve accuracy, flow measurement, triggering, positive end-expiratory pressure (PEEP), and automatic leakage compensation, thus raising issues regarding the accuracy of ventilator performance (11) and patient safety.

The present study was designed to test the performances of seven standard intensive care unit (ICU) ventilators available in Europe during HeO₂ utilization, compared with theoretical predictions based on the physical properties of helium, and to develop correction factors (12) to ensure the safe use of HeO₂ during mechanical ventilation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Ventilators and HeO₂ Administration

The ventilators studied were the Veolar FT, Galileo (Hamilton Medical, Rhäzuns, Switzerland), Evita 2, Evita 4 (Drägerwerk, Lübeck, Germany), Servo 900C, Servo 300 (Siemens-Elema, Solna, Sweden), and 7200 Series (Nellcor Puritan Bennett, Pleasanton, CA). Helium was contained as a fixed mixture of 78:22 helium and O₂, respectively (heretofore referred to as "heliox," to differentiate it from the final heliox and O₂ mixture delivered by the ventilator, termed HeO₂), in a 50-L canister pressurized at 150 bar (Pangas Swiss Calibration, Luzern, Switzerland), and delivered through a pressure regulator at 2 to 8 bar, according to manufacturer specifications, into the ventilator inlet normally used for air.

Fraction of Inspired O₂ (FI_O2)

The FI_O2 set on the ventilator (FI_O2set) was compared with the FI_O2 actually delivered by the ventilator (FI_O2del), determined by a rapid paramagnetic gas analyzer (Normocap Oxy; Datex, Helsinki, Finland), as well as with the FI_O2 indicated by the ventilator's own O₂ sensor. The range of FI_O2set was 0.21 to 1.0, with 0.05 increments.

Delivered Tidal Volume and Inspiratory Flow Measurements

Ventilators were connected to a test lung model (PneuView AI 2601I TTL; Michigan Instruments, Grand Rapids, MI, cross-calibrated independently from the manufacturer, Metron AS, Trondheim, Norway). Briefly, the test lung is built around a chamber whose compliance (Crs) and "airway" resistance (Raw) are determined by precision spring-loading and variable cross-section resistors. A transducer measures pressure (P) inside the chamber. Tidal volume (VT) is determined as VT = Crs × P (measured at zero flow rate). Calculated VT is verified against a calibrated volume scale attached to the device. Inspiratory flow rate (I) is computed by derivation of VT. For all tests, Crs and Raw were set at 0.05 L/mbar and 5 mbar · L · s¹, respectively, and conditions were those of room air temperature, barometric pressure, and humidity (ATPS). Using this model, flow and volume measurements are independent of the physical properties of the gas mixure, as opposed to measurements performed with a pneumotachograph or other flow measuring devices in which a resistive component is present (13). VT indicated by the lung model VT was further checked against VT measured with a precision density-independent spirometer (5420 Volume Monitor; Ohmeda, Louisville, KY) placed on the inspiratory limb. Pressure, volume, and flow signals were stored in a laptop computer equipped with the PneuView software package.

The following parameters were compared: VT set on the ventilator (VTset); inspired (VTi) and expired (VTe) VT determined by the ventilator's measuring device (if present); effective VT delivered by the ventilator, as measured by the lung model and spirometer (VTdel).

After a complete test to verify the absence of leaks, the protocol detailed subsequently was conducted at an FI_O2set of 0.21, 0.3, 0.35, 0.4, 0.45, 0.5, 0.6, and 1.0. More measurements were made at the lower end of the FI_O2set range because there is probably little benefit to be obtained clinically through a reduction in density above an FI_O2 of 0.6. The tests were conducted in both volume-controlled and pressure-controlled modes.

In volume-controlled mode, VTset was increased by 50-ml increments from 100 to 1,300 ml. For each VTset, 10 successive measurements were performed and averaged. Other ventilator settings were: ventilatory rate 10/min, constant flow pattern, inspiratory:expiratory time ratio 1:2, no PEEP; these settings were maintained constant throughout the volume-controlled protocol. In the pressure-controlled mode, the pressure limit was varied by 5-cm H₂O increments, between 5 and 60 cm H₂O, at each FI_O2set. For each level of pressure limit and FI_O2, VTdel was compared with VTdel obtained with an FI_O2 of 1.0 (i.e., without heliox), used as control. Ventilatory rate and PEEP were the same as in volume-controlled tests.

PEEP

PEEP was determined as the end-expiratory pressure measured in the lung model chamber by the pressure transducer after 10 mechanical breaths, for PEEP values set on the ventilator ranging from 5 to 25 cm H₂O, at each various FI_O2set.

Theoretical Predictions

This approach pursued the following goals: first, to determine if the consequences of using HeO₂ on FI_O2del and VTdel could be predicted from its physical properties, and thus if the latter alone could account for the changes observed; second, to validate the observed modifications in VT and FI_O2; third, to validate the correction factors to be applied to VT. For this purpose, standard equations on the physics of gas flow were used, integrating the main operating conditions and design of the inspiratory valves, gas mixing devices (technical information obtained from the manufacturers), and the physical properties of O₂, helium, nitrogen, and air (Table 1).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Theoretical Predictions

Briefly, the following approach and principles were used:

Dynamics of gas flow. The density of HeO₂ (He) is markedly inferior to that of airO₂ (Figure 1) (1), which should alter the dynamics of flow through the ventilator's tubing and inspiratory valves. Indeed, laminar or turbulent conditions may prevail, the nature of flow being described by Reynold's number (Re) according to the following equation (14, 15):

(1)

View larger version (22K): [in this window] [in a new window]   Figure 1.   Ratio of densities of airO2 and HeO2 to density of air, at various values of FIO2.

where = flow (ml/s),  = density of the gas (g/ml), r = radius of the tube (cm), µ = gas viscosity (g/[cm · s]). Re  4,000 predicts turbulent flow, Re  2,000 laminar flow (14, 15).

In laminar conditions, the pressure difference (P) necessary to obtain a given flow is determined by Equation 2:

(2)

where k1 (coefficient of linear resistance) = 8l/r⁴ (from the Hagen-Poiseuille equation),  = gas viscosity (g/[cm · s]), l = length of tube, r = radius of tube, lam = laminar flow (ml/s).

In turbulent conditions:

(3)

where k (coefficient of nonlinear resistance) = fl/4²r⁵, where f = a friction factor dependent on the Reynolds number and the roughness of the tubing wall. In smooth tubing f = 0.316/ Re^1/4, r = radius of the tube (cm), and turb = turbulent flow (ml/s).

As can be seen, under laminar conditions, P is uninfluenced by the density of the inhaled gas mixture, whereas in turbulent conditions P depends on density (the lower the density, the lower Re and thus the lower k2). Hence, for a given P under turbulent conditions, which are those prevailing in ventilatory tubing and valves, the lower the density of the inhaled gas, the higher the flow.

Inspiratory valve design and theoretical consequences of using HeO₂ on delivered VT and FI_O2 in volume-controlled mode. With the exception of the Servo 900C, all the ventilators tested in this study are equipped with electromagnetic proportional solenoid valves (16), consisting of two compartments connected by an opening whose surface section is controlled by a ball valve driven by a microprocessor-controlled electromagnetic motor magnetic motor (Figure 2). To minimize inertia and valve response time (average 4 to 10 ms), the maximal opening is small and varies within a tight range (1.2 to 14.4 mm², depending on the ventilator). Thus, to obtain the high inspiratory flow rates needed during mechanical ventilation in the face of elevated resistance, these valves operate under turbulent conditions. Two types of electrodynamic valves equip most modern ventilators, categorized according to the magnitude of P across the valve opening:

View larger version (31K): [in this window] [in a new window]   Figure 2.   Schematic representation of an electromagnetic proportional inspiratory valve. Arrows indicate path through the valve followed by the gas (heliox in this example).

Low-pressure valves. The Veolar FT and Galileo are equipped with one such valve, downstream from the airO₂ mixing chamber (Figure 3). P varies between 200 and 340 mbar, resulting in a flow of 250 to 3,000 ml/s depending on the surface section of the opening (1.2 to 14.4 mm²). Flow across the valve's opening is described by the following equation (14):

(4)

View larger version (26K): [in this window] [in a new window]   Figure 3.   Schematic representation of the inspiratory valve and gas mixing setup in the Veolar FT and Galileo (right panel ), and the Evita 2, 4, and Servo 300 (left panel ). Arrows indicate path followed by the gases.

where = flow (ml/s), P = pressure difference across the opening (dyn/cm²),  = density of the gas (g/ml), S₁ = cross-section of cylindrical compartments (cm²), S₂ = cross section of opening (cm²), and C = restriction coefficient (0.6 to 0.7, depending on shape of opening and Reynolds number). Equation 4 predicts that for a given degree of valve opening and P, flow depends on gas density. Furthermore, as can be seen from Figure 1, when air and O₂ are used, there is very little variation in overall gas density when FI_O2 varies. However, replacing air with heliox results in density changing by a factor of 8 as FI_O2 increases between 0.22 and 1.0 (Figure 1). Hence, if the terms of Equation 4, C, S₁, S₂, and P remain constant (K), Equation 4 can be simplified as follows:

(5)

and

(6)

where (HeO₂) = flow of the HeO₂ mixture leaving the ventilator (ml/s), K(HeO₂) = constant as defined above, and (HeO₂) = density of the HeO₂ mixture leaving the ventilator (g/ml).

By definition, K(air) = K(HeO₂), hence from Equations 5 and 6:

(7)

finally,

(8)

Thus, flow through the valve should increase when density is lower. From these equations, a theoretical volume correction factor for a given FI_O2 can be determined that can be applied to VTset:

(9)

The results are summarized in Table 2.

The 7200 Series contains two such inspiratory valves (maximal P is 687 mbar). Hence, total flow delivered (del) can be computed as:

(10)

where O₂ = flow through O₂ valve (ml/s), air = flow through air valve (ml/s), set = flow set on the ventilator (ml/s), and FI_O2 = inspired O₂ fraction set on the ventilator. Applying the volume correction factor for an FI_O2 of 0.22 obtained from Equation 9, Equation 10 becomes

(11)

Thus:

(12)

Volume correction factors at different FI_O2set are shown in Table 2.

Regarding FI_O2, since in the Veolar and Galileo the mixing chamber is located upstream from the single inspiratory valve (Figure 3), density should not affect FI_O2. Indeed, each gas is admitted to the mixing chamber from its mural or canister source through mechanical (Veolar FT) or electrodynamic (Galileo) valves until a total chamber pressure is reached. The latter is equal to the sum of partial pressures of both gases at the desired combination of O₂ and air (or heliox in this study) determined by the FI_O2 setting on the ventilator control panel. Thus, because this system is pressure- and not volume-controlled, ideal gas laws apply (17). To simplify, pressure-control consists of a transfer of a volume of gas from a first compartment (mural or canister) of a given volume (V₁) and pressure (P₁) into a second compartment (mixing chamber) of a given volume (V₂) at a given pressure (the target pressure P₂). According to Boyle's law, at equilibrium P₁ × V₁ = P₂ × V₂. Hence, in the absence of flow and resistance, density exerts no influence, and the lower density of heliox should not influence FI_O2. With the 7200 Series, FI_O2del should be determined by the proportion of flow for each gas, and computed as follows:

(13)

where:

(14)

and

(15)

Finally:

(16)

Conversely, to obtain the desired FI_O2, FI_O2set should be corrected as follows:

(17)

Table 3 shows the predicted FI_O2del obtained for a given FI_O2set, according to Equation 16, as well as corrected FI_O2set obtained from Equation 17.

In the case of the Servo 900C, the inspiratory valve is of a different type. A scissors-type mechanism, controlled by a stepper motor, regulates the cross-section by pinching a silicon tube. The peak inspiratory flow is determined by the pressure entering the valve, while the scissors-type mechanism allows an increase or decrease in flow by approximately 10% increments (16). Information from a pressure/flow monitoring device is fed back into the stepper motor controlling the scissors-type valve, thereby allowing constant regulation of inspiratory valve opening during inspiration. Due to the difficulty of modeling flow across this type of valve, no mathematical predictions were made for the Servo 900C.

The gas mixing chamber is also upstream from the inspiratory valve, and thus FI_O2 should not be influenced by the presence of heliox.

High-pressure valves. The Evita 2, 4, and Servo 300 are fitted with two high-pressure valves, one for O₂, the other for air (or heliox in this study), upstream from the mixing chamber (Figure 3). A P of 1 to 3 bar results in particular thermodynamic conditions across the valve, known as "supercritical," the description of its complex nature being beyond the scope of this study (18). Suffice it to say that under such conditions, flow is described by Equation 18:

(18)

where a = constant independent of the nature of the gas; P_abs = absolute admission pressure (dyn/cm²); c = speed of sound in the particular gas (m/s), where c = /M, and where  = adiabatic coefficient (N/m²) and M = molecular mass (kg/m³); x = position of the ball valve drive shaft (determines the surface of the opening) in centimeters.

Total flow (del) delivered by both (O₂ and air or heliox) high-pressure valves can be determined by Equation 10. However, as seen from Equation 18, for a given P and valve opening, flow for each valve depends on the speed of sound c in this gas mixture, which in turn depends on /M. In turn, these two variables depend on FI_O2, ranging from 1.556 to 1.40 for  and 9.94 to 31 for M, for O₂:HeO₂ mixtures of 78:22 to 0:100, respectively. If a 78:22 mixture of helium and O₂ is administered through a valve calibrated for air, as in our study, flow will be increased by a factor of 1.81, which can be used to correct Equation 10 to determine total flow (del) delivered by both high-pressure valves:

(19)

which can be further simplified to:

(20)

Finally:

(21)

It is thus also possible to determine a "corrective volume factor," the true delivered volume being computed by the following equation:

(22)

The volume factors for the various FI_O2 used in this study are shown in Table 2.

Regarding FI_O2, the same equations (Equations 16 and 17) as for the 7200 Series apply, but with different constants accounting for supercritical conditions. Hence, Equations 16 and 17 become:

(23)

Conversely, to obtain the desired FI_O2, FI_O2set should be corrected as follows:

(24)

Results are shown in Table 3.

FI_O2

The results are outlined in Table 4. There was some degree of discrepancy between FI_O2set and FI_O2del. The magnitude of this discrepancy varied both between the ventilators tested and as a function of FI_O2set, being the greatest when FI_O2set was between 0.3 to 0.6. As can be seen, the Veolar FT, Galileo, Servo 300, and Servo 900C were the most accurate, whereas with the Evita 2 and Evita 4 FI_O2del was lower than FI_O2set, by an average of 10%, reaching 18% at an FI_O2set of 0.5. With the 7200 Series, FI_O2del was considerably higher than FI_O2set. There was excellent agreement between these findings and the theoretical predictions of FI_O2del, except for the 7200 Series (Table 3). The FI_O2 reported by the ventilator sensor and that measured with the rapid paramagnetic gas analyzer were identical on all machines (data not shown).

Delivered VT in Volume-controlled Mode

There was variable congruency between VTset and VTdel. With the Veolar FT, Galileo, and Evita 2, VTdel was consistently higher than VTset, the magnitude of this discrepancy being inversely related to FI_O2, i.e., the lower the FI_O2 (the higher the heliox fraction), the more VTdel exceeded VTset, as shown in Figure 4. Thus, VTdel exceeded VTset by 25% at an FI_O2 of 0.5 and by 60% at an FI_O2 of 0.22. As can be seen in Figure 4, for a given FI_O2, there was a linear relationship between VTdel and VTset (the slopes of the regression lines for each of the three ventilators being nearly identical, the set of lines for only one ventilator is shown). From the regression equations, a volume correction factor to be applied to VTset to obtain a given VTdel for each FI_O2 can be determined, as shown in Table 5. The Evita 4 followed the same pattern, but for a VTset higher than 500 ml, the relationship between VTset and VTdel became nonlinear, VTdel exceeding VTset by as much as 100% when low FI_O2 and high VTset were combined (Figure 5). This problem could be solved by inactivation of the leakage compensation mechanism, in which case the machine's performance was identical to the Evita 2. Furthermore, a high-priority alarm of inoperative flow measurement was activated, which could not be silenced by recalibration of the flow sensor, but required inactivating inspiratory flow monitoring. With the Servo 900C VTdel also exceeded VTset, but by more than 40% at an FI_O2 of 0.22. With the 7200 Series, VTdel was consistently lower than VTset, the magnitude of this discrepancy being inversely related to FI_O2, and being considerable when FI_O2 < 0.9 (Figure 6). Finally, the Servo 300 was unaffected by the use of HeO₂, whatever the FI_O2, as shown by the line of identity between VTset and VTdel (Figure 7).

View larger version (18K): [in this window] [in a new window]   Figure 4.   Relationship between VT set on the ventilator (VTset) and VT actually delivered by the ventilator (VTdel), at various values of FIO2, with corresponding regression equations (panel ), for the Veolar FT, Galileo, and Evita 2.

View larger version (21K): [in this window] [in a new window]   Figure 5.   Relationship between VT set on the ventilator (VTset) and VT actually delivered by the ventilator (VTdel), at various values of FIO2, with corresponding regression equations (panel ), for the Evita 4.

View larger version (17K): [in this window] [in a new window]   Figure 6.   Relationship between VT set on the ventilator (VTset) and VT actually delivered by the ventilator (VTdel), at various values of FIO2, with corresponding regression equations (panel ), for the 7200 Series.

View larger version (11K): [in this window] [in a new window]   Figure 7.   Relationship between VT set on the ventilator (VTset) and VT actually delivered by the ventilator (VTdel), according to FIO2 with corresponding regression equations (panel ), for the Servo 300. Only the FIO2 of 0.22 and 1.0 are shown to avoid excessive overlap, but the relationship was identical at all FIO2.

Delivered VT in Pressure-controlled Mode

With the exception of the 7200 Series, in pressure-controlled mode, VTdel was identical to control measurements at FI_O2 of 1.0, at all pressure limits and FI_O2 tested, with all ventilators. However, VTi reported by the ventilator underestimated true VTdel on all ventilators, by a magnitude equal to that between VTset and VTdel in volume-controlled mode. With the 7200 Series, at all FI_O2 < 0.5, repeated opening-closing of the heliox inspiratory valve caused major vibrations within the ventilator, prompting cessation of the test.

Expired VT

The findings apply to both volume- and pressure-controlled modes.

There was considerable variation among the values of VTe reported by the ventilators. With the Veolar FT, Galileo, Servo 900C, and Servo 300, indicated VTe was lower than VTdel. As with VTset, this underestimation was linear and was a function of FI_O2, i.e., the lower the FI_O2, the larger the disparity between VTe and VTdel (Figure 8). Hence, even though there were quantitative differences between the ventilators, correction factors could be determined. These factors are indicated in Table 5. As can be seen, for the Veolar FT and Galileo the factors were the same as for VTset. Correction factors for the Servo 900C and Servo 300 were smaller than those of the Veolar FT and Galileo.

View larger version (17K): [in this window] [in a new window]   Figure 8.   Relationship between expired VT reported by the ventilator (VTe) and actually expired VT (VTmeas), at various values of FIO2, with corresponding regression equations (panel ), for the Veolar FT and Galileo.

With the Evita 2, Evita 4, and 7200 Series, reported VTe was considerably higher than VTdel. This overestimation increased nonlinearly as FI_O2 decreased, VTe determination becoming inoperative below an FI_O2 of 0.8. Concomitantly, a high-priority alarm of expiratory flow monitoring malfunction was activated. Recalibration of the flow sensor did not correct this problem.

PEEP

Measured PEEP was in good agreement with set PEEP, the maximal discrepancy being ± 0.5 cm H₂O, independently of FI_O2.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the seven nonmodified ICU ventilators tested, the use of HeO₂ led to alterations in ventilator function, whose nature varied from one machine to another, as does the ease with which these modifications can be corrected. Before discussing these results, let us briefly summarize them:

1. The Veolar FT, Galileo, Servo 900C, and Servo 300 delivered an accurate FI_O2 at all tested FI_O2, whereas the Evita 2 and Evita 4 administered a lower than set FI_O2, particularly in the 0.35 to 0.6 range. The 7200 Series delivered a considerably higher than set FI_O2.
2. In volume-controlled mode, delivered VT was higher than set VT on the Veolar FT, Galileo, Evita 2, and Servo 900C, the magnitude of this discrepancy being inversely related to FI_O2. There was a linear correlation between VTset and VTdel. The Evita 4 followed the same pattern up to a VTset of 500 ml, but further VTset increases entailed considerable discrepancy with VTdel, with a nonlinear relationship. The 7200 Series delivered a considerably lower than set VT when FI_O2 < 0.9. With the Servo 300, VTset and VTdel were identical. Reported VTe was underestimated by all ventilators, also as an inverse linear function of FI_O2, except for the Evita 2, Evita 4, and 7200 Series, which were unable to monitor expired VT.
3. The observed results for FI_O2 and VTdel were in agreement with theoretical predictions based on the changes in density and principles of ventilator design, except for the Servo 300 and 7200 Series.
4. In pressure-controlled mode, VTdel was identical to control VT obtained at an FI_O2 of 1.0 with all ventilators tested except for the 7200 Series, which malfunctioned.
5. PEEP was adequately obtained with all ventilators at all FI_O2 tested.

Consequences on Delivered VT and FI_O2 in Volume-controlled Mode

The findings can be best understood by turning to the principles of valve design and gas flow discussed previously. The effects on FI_O2 and VTdel are closely linked and will be discussed together. As we have seen, the Veolar FT, Galileo, and Servo 900C are equipped with a single inspiratory valve, located downstream from the gas mixer (Figure 3). Calibration of the valve is performed for a single given FI_O2. There is no built-in system designed to adjust for changes in gas density. Hence, if density decreases, flow should increase, and vice versa. This is not a problem when airO₂ mixtures are used, because the densities of these two gases are quite close (Table 1 and Figure 1). Hence, whatever the FI_O2, the overall density change of the gas mixture is very small, and of no consequence on ventilator function. However, the considerably lower density of heliox (Table 1 and Figure 1) entails a marked increase in flow, which is proportional to the fraction of heliox in the gas mixture, and thus inversely proportional to FI_O2. FI_O2 itself is not influenced by the reduction in density, however, because mixing occurs upstream from the valve (Figure 3). This explains why the Veolar FT, Galileo, and Servo 900C correctly administered the FI_O2set (Table 4), whereas VTdel was higher than VTset (Table 5), as predicted from theoretical calculations (Tables 2 and 3).

The Evita 2 and Evita 4 are fitted with two proportional inspiratory valves, one for air, the other for O₂, gas mixing occurring downstream from these valves (Figure 3). Consequently, replacing air with heliox should, for the reasons outlined previously, increase flow through the valve normally regulating air admission, whereas there will be no effect on the O₂ inspiratory valve. Because VTdel results from the outflow from the mixing chamber, which in turn results from the combined flows of both gases, total outflow will be increased, and VTdel will be > VTset, proportionally to the fraction of heliox present in the mixture, as predicted (Table 2) from Equations 21 and 22 and as confirmed by our tests (Figures 4 and 5). However, contrarily to what occurs with the Veolar FT, Galileo, and Servo 900C, FI_O2 also changes. Indeed, the higher the fraction of heliox, the more heliox flow through the valve will exceed its set value, and as a result, FI_O2del by the ventilator is lower than FI_O2set, as predicted (Table 3) by Equations 23 and 24 and as documented in our tests (Table 4).

Finally, an additional problem occurs with the Evita 4, with which VTdel exceeds VTset in a nonlinear manner for any FI_O2 < 0.5 and VTset > 500 ml (Figure 5). This is due to the automatic leakage compensation mechanism. Indeed, the expiratory flow sensor malfunctions with HeO₂, and the mechanism erroneously interprets this as being caused by gas leaks in the patient or ventilator circuit. Automatic compensation triggers an increase in the degree of inspiratory valve opening, resulting in a major increase in VTdel. Volume correction factors can be computed for all FI_O2 < 0.5, according to nonlinear regression equations. Their general form would be y = ae^bx, where y = volume factor, x = VTset, and b = correction factor dependent on FI_O2. However, using such factors might prove impractical in the clinical setting. Another approach is to inactivate the flow monitoring system through the control panel menu, which corrects this problem but requires an independent monitoring of VT.

The major discrepancies in both FI_O2 and VT observed with the 7200 Series, which differed from theoretical predictions, can be explained by the design of the machine. Indeed, as previously stated, gas mixing occurs downstream from two inspiratory valves, one for O₂, the other for air. Flow of gas coming out of each valve is measured by a hot-wire pneumotachograph, and this information is fed back to the valve to continuously adjust the size of its opening, which thus regulates flow. Using heliox instead of air has major consequences on the hot-wire pneumotachograph of that particular valve for the following reason. The device comprises two platinum wires located in the mainstream flow, heated electrically and maintained at 400° C. Cooling of the wires by convective heat loss is proportional to gas flow. The intensity of electric current required to maintain the wire at 400° C can thus be translated into flow units. Heat loss can be quantified by the following equation:

(25)

where = heat loss, Ac = surface of wire exposed to gas (cm²), Ts = wire temperature (°K), h'c = coefficient of flow geometry around wire, V = gas velocity (ml/s), Mv = volume mass of gas (kg/m³), Cm = specific heat conductivity of gas (µcal · cm · s · °K), and Ta = gas temperature (°K). The specific heat conductivity of helium is six times that of air (Table 1). Consequently, heat loss will increase markedly, and the device interprets this as being due to an increase in flow. However, because of the sixfold larger loss of heat the magnitude of flow overestimation is also very large. Hence, the feedback mechanism will markedly reduce valve opening, which in turn will markedly reduce heliox flow. This explains the fact that with this machine VTdel was lower than VTset, and also that the magnitude of this discrepancy was substantial. Concomitantly, the O₂ valve function is unaltered by the use of heliox. Thus, because gas mixing occurs downstream from these two valves, and because, as we have seen, flow through the heliox valve is considerably reduced, FI_O2del will increase, as observed (Table 4). At an FI_O2set of 0.21, only the air (or in this case heliox) valve is operative, which explains that FI_O2del = FI_O2set, since the FI_O2 in the heliox tank is 0.22. Because of these problems, the machine cannot be used safely with heliox.

Although resting on the same basic inspiratory gas delivery design principles as the Evita 2 and 4 (Figure 3), the Servo 300 did not behave according to the theoretical predictions with regard to FI_O2 (Tables 3 and 4) and VTdel (Tables 2 and 5). Indeed, the machine is also fitted with two high-pressure proportional inspiratory valves, one for O₂, the other for air (or heliox in this study), each valve being encased in a gas delivery unit. A mixing chamber is located downstream from these two units. Our understanding is that the reason for this lies in the monitoring devices contained within the gas delivery systems, because each of the latter is fitted with a temperature probe and a differential pressure transducer. These devices are intended to compensate for changes in density and temperature caused by gas expansion within the ventilator. These changes are negligible when standard wall inlet pressures (3 to 4 bar) are used for O₂ and air. However, the Servo 300 is designed to operate with high-pressure sources, in which case the effects of gas cooling on density during expansion are much greater. Hence, in case of changes in the physical properties of the gas, the temperature and pressure/flow sensors detect a discrepancy between the set and actual flows, and correct the valve's opening through a feedback mechanism. The same mechanisms are operative when density changes for reasons other than temperature, e.g., when heliox is used instead of air, and this probably explains why VTset and VTdel are identical with this machine.

As for the Servo 900C, the actually observed difference between VTdel and VTset was smaller than that predicted (Table 2) from Equations 8 and 9. The reason for this is not quite clear, but probably stems from the effect of HeO₂ on the flow derivation device. Briefly, this device consists of a main tube with a wire mesh membrane across its lumen. The resistance entailed by the latter forces a portion of the gas flow into a small parallel tube, where pressure is measured by a differential transducer. The information is then fed back into the mechanism controlling the inspiratory valve opening, any drop in pressure entailing an increase in its opening and vice versa. If air is replaced by heliox, resistance across the main tube's membrane will decrease, flow will pass through the main tube, and thus decrease in the side tube. Hence, pressure in the side tube will decrease, and the feedback mechanism will compensate by further opening the inspiratory valve, interpreting this as a true decrease in inspiratory flow. This should result in VTdel being higher than VTset, but the magnitude of the difference cannot be precisely predicted by Equations 8 and 9, because more than one mechanism is involved in the process.

A final series of factors should be considered in interpreting our results. First, during inspiration, not all the VTset is delivered to the patient or the lung model, as a result of a volume of gas being trapped in the ventilator's inspiratory circuit. Two factors account for this occurrence: gas compressibility and circuit compliance.

Gas compressibility. A portion of the VTset is trapped in the inspiratory circuit, because of the compressibility of the gas. In the clinical operating conditions of ICU ventilators, it is estimated that this volume is equal to 1 ml/cm H₂O/L (19). In the conditions of our tests, mean inspiratory pressures were  10 and  20 cm H₂O for VTset of 500 ml and 1,000 ml, respectively, which represents a volume of gas trapped within the circuit due to compressibility alone of 5 to 10 ml.

Compliance of the circuit. Total compliance of the circuits (including tubings, water traps, humidifiers, and filters) ranges from 2.5 to 3 ml/cm H₂O. During our tests, water traps, humidifiers, and filters were removed. Thus, only the compliance of the tubings (and pneumotachograph with the Hamilton machines) should be taken into consideration. Results are the following: Hamilton Veolar FT and Galileo, 1.2 ml/cm H₂O; Nellcor Puritan Bennett 7200 Series, 1.6 ml/cm H₂O; 0.8 ml/cm H₂O for the other machines tested. Hence, at the mean pressure regimen documented in our tests, the trapped volume would at most amount to 32 ml. Taking both these mechanisms into account, the volume of gas trapped within the inspiratory circuit should be  42 ml in the conditions of our tests.

Considering the above, VTdel should be lower than measured expired tidal volume (VTmeas) in our study, and the magnitude of this difference can be expressed by a factor that includes the compressibility of the gas and the compliance of the circuit. This factor can be determined at FI_O2 1.0, because then the difference between VTdel and VTmeas is due to compressibility/compliance solely, and not to any confounding factor such as the effect of density on inspiratory valve function. It should, however, be noted that, provided temperature is unchanged, the compressibility of a volume of gas is a function of neither its density nor its viscosity, and hence no difference in the volume of trapped gas between airO₂ or HeO₂ should occur.

Second, the ventilators have various built-in corrections designed to compensate for the aforementioned and other mechanisms, to ease the clinician's task.

Physical conditions. Ventilators deliver gas in ambient temperature and pressure conditions (ATP), whereas expired gas is in body temperature and pressure saturated (BTPS) conditions. Most ventilators correct automatically the expiratory flow/volume signal to account for this difference. In the tests we performed, both inspiration and expiration occur under ATP conditions. Thus, the automatic correction should be disabled, or a correction equation applied, to obtain true expiratory volumes. The Evita 2 and 4, Servo 300 and 7200 Series feature an automatic correction of expiratory flow, which takes into account the BTPS, whereas the Hamilton machines do not have this function. However, during startup calibration of the proximal pneumotachograph is performed under ATP conditions if no air warmer is used, and under BTPS conditions if such a device is used.

Precision of the ventilators. Electromagnetic valves have a degree of precision in the order of 5% compared with measured values (data from Hamilton Medical). Flow measurements have a precision of < 10%. The volume values reported by the ventilators can thus vary from 50 to 100 ml for a VTset of 1,000 ml. Depending on the direction of this variation, this error can mask or overestimate the value of trapped gas volume.

Special functions of some of the ventilators tested. The Evita 2 and 4, Servo 300 and 7200 Series all have a built-in function for automatic compensation of circuit compliance. The latter is measured during the startup procedure when the machine is turned on. This procedure was performed before all our tests, and hence VTdel already takes into account circuit compliance correction on these ventilators.

To summarize, several factors can influence the exact value of the various correction factors we determined: (1) during inspiration: gas compressibility (compression), circuit compliance, precision of the inspiratory valve, presence or absence of automatic compensation of circuit compliance; (2) during expiration: gas compressibility (expansion), circuit compliance, precision of flow sensors, presence or absence of automatic BTPS correction. It must once again be stressed, however, that none of these interferences with VT measurement will be influenced by density. Hence, these factors should not interfere with the comparison of HeO₂ and airO₂.

Consequences on Delivered VT in Pressure-controlled Mode

In pressure-controlled modes, VTdel depends on the mechanical properties of the lung model and is mostly independent of inspiratory valve function, the latter generating flow until a preset target pressure level is attained. At equilibrium, i.e., at the preset pressure level, the volume of gas that has been transferred into the lung model is the same regardless of density, as predicted by Boyle's law (17), and as outlined in METHODS. Therefore, the density reduction associated with the use of HeO₂ should decrease the time needed to reach the target pressure, but should not influence VTdel compared with VTdel obtained in control conditions with the same target pressure level but at an FI_O2 of 1.0. However, the VTi reported by the ventilator should underestimate VTdel. Our results confirm this hypothesis.

The malfunction observed on the 7200 Series in this mode is explained by the effect of heliox on the hot-wire pneumotachograph flow sensor and its feedback mechanism on the inspiratory valve discussed previously. Indeed, in pressure-controlled mode, the circuit is pressurized by a user-adjustable peak inspiratory flow. However, because of the large overestimation of inspiratory flow entailed by heliox, the valve is constantly cycling between open/closed positions, leading to marked mechanical oscillations and dysfunction.

Expired VT Determination

VTe is determined by different systems in the ventilators tested, with variable consequences on the accuracy of indicated VTe.

The Veolar FT and Galileo are equipped with a variable orifice flowmeter (20) located proximally to the Y-piece of the patient-ventilator circuit. In brief, the device consists of a membrane with a flexible V-shaped membrane center portion, placed across the mainstream flow. Pressure is measured on both sides of this resistance, by a differential pressure transducer, whose output is proportional to flow. Because of the center membrane's flexibility, the cross-section of the opening increases (and thus resistance decreases) as flow increases. Thus, resistance remains constant over a flow range of 0.1 to 2.5 L/s (20), allowing a linear response of the device, which is calibrated for a given gas density. Flow across the device's orifice being turbulent, if air is replaced by heliox, resistance will decrease, and the pressure difference across the membrane will drop. Thus, the device will underestimate true flow, as confirmed by our experiments. Furthermore, because the same device measures flow during both inspiration and expiration, the same correction factor can be applied to inspired and expired VT (Table 5).

The Servo 900C and Servo 300 are fitted with a flow derivation device, such as detailed in the previous paragraph, the same discussion applying to VTe determination. The Evita 2, Evita 4, and 7200 Series rely on a hot-wire technique flowmeter. Hence, for the reasons previously outlined, the magnitude of expiratory flow and VTe will be considerably overestimated, the difference being of such magnitude that the expiratory flow monitoring malfunction alarm is activated, precluding any monitoring of expiratory flow and VTe when FI_O2 is < 0.8.

PEEP

The Veolar FT, Galileo, Evita 2, and Evita 4 are equipped with diaphragm-type expiratory valves (16). An electromagnetic (Veolar FT and Galileo) or pneumatic (Evita 2 and Evita 4) motor applies a pressure on the diaphragm equal to the set level of PEEP. During expiration, the diaphragm is either completely open (expiratory circuit pressure > PEEP) or closed (expiratory circuit pressure < PEEP). Thus, there is virtually no resistive component to this type of valve, and its performance should be independent of gas density. Our results are in accordance with this, showing no discrepancy between set and measured PEEP.

The Servo 900C and Servo 300 use a scissors-type valve, of the type used for inspiration. As seen previously, this valve is based on variable compression of a segment of tubing. There is thus a resistive component to maintaining a given PEEP. Hence, reducing density should lead to an inability to keep the desired PEEP level. However, as our results showed, this was not the case. The reason for this lies in the feedback mechanism incorporated into the expiratory valve, by which any drop in pressure induces further compression of the tubing segment to maintain the preset PEEP.

In conclusion, HeO₂ interferes with several key ventilator functions, especially in volume-controlled mode. The nature and magnitude of these alterations vary between ventilators, but all could lead to potentially hazardous management decisions on the part of the ICU clinician, owing to the erroneous data provided by the machine. The ease with which these problems can be corrected varies from one ventilator to another, and the 7200 Series cannot be used with heliox. Except for the latter machine, most alterations stem from the known physical properties of helium rather than from ventilator dysfunction, as demonstrated by our study in which there was very good concordance between theoretical predictions and measurements. Consequently, correction factors and tables can be determined (12), allowing for the safe use of HeO₂ in the clinical setting.