The anesthetic gas xenon has distinctly different physical properties compared with air, nitrous oxide, or oxygen. This led us to predict that xenon would affect the performance of commercially available flowmeters.
Flow was generated by an anesthesia ventilator connected to a lung simulator via a semiclosed breathing circuit. With the system filled with air or with various concentrations of xenon or nitrous oxide in a balance of oxygen, the tidal volume was measured with two rotating vanes, a Pitot tube, a variable-orifice flowmeter, and two constant-temperature hot-wire flowmeters.
Although xenon minimally affected both rotating vane flowmeters, it caused the Pitot tube and the variable-orifice flowmeters to overread in proportion to the square root of the density of the gas mixture used (xenon is 4.6 times more dense than air). In contrast, the hot-wire anemometers underread with xenon; for example, their readings in the presence of 45% and 70% xenon were less than 10% of those displayed when air was used. Nitrous oxide minimally affected all the flowmeters except the variable-orifice device. The Pitot flowmeter was also affected, but only when its gas analyzer port was open to the ambient air so that it no longer corrected its readings for changes in gas composition. In these cases, nitrous oxide produced overreadings in the same manner as did xenon.
Among the four types of flowmeters studied, only the rotating-vane type is sufficiently accurate for use during anesthesia with xenon.
THE inert gas xenon is receiving renewed interest because it has many characteristics of an ideal anesthetic. For example, its blood-gas partition coefficient is extremely low (0.12 to 0.14), [1,2]enabling faster induction of and emergence from anesthesia than with other inhalational agents. [3,4]It is also inflammable, odorless, and extremely unreactive. Although it is equal to nitrous oxide (N2O) in analgesic potency, [6,7]it has a lower minimal alveolar concentration (71% in humans ) and can produce 1 minimal alveolar concentration anesthesia during normobaric conditions. Its high cost (approximately $12.40 in the United States or 1,800.00 yen/l in Japan) has made it unacceptable for anesthesia practice, but the cost can be minimized by using a closed rebreathing system. The xenon recycling system is also being investigated. sup #
Xenon has distinctly different physical properties compared with those of common gases, including oxygen, N2O, and air (Table 1). Consequently, its use in anesthetic concentrations (e.g., 50-70%) significantly alters the physical properties of a gas in the breathing system. This might affect the behavior of respiratory monitoring systems, particularly flowmeters, because many of their measurement principles depend closely on the physical properties of a gas. In fact, some commercially available flowmeters are designed to correct their readings for differences in gas composition (i.e., gas compensation). But xenon might still produce false readings, because these devices are not designed for xenon and therefore might not make proper compensations.
In this investigation, we evaluated the effects of xenon on the performance of flowmeters involving four different measurement principles; that is, the rotating vane or turbine, the Pitot tube, the variable orifice, and the constant-temperature hot-wire anemometry.
Materials and Methods
We evaluated two rotating vane flowmeters, Magtrak IV (Ferraris Medical, London, UK;Figure 1A), which is a conventional Wright respirometer, and Ohmeda 5400 (Ohmeda, Madison, WI;Figure 1B), which is incorporated in the Modulus CD anesthesia machine (Ohmeda). Their primary difference is that the axis of the vane is either perpendicular (Magtrak IV) or parallel (Ohmeda 5400) to the direction of the gas inflow.
As an example of a Pitot tube flowmeter, we evaluated Capnomac Ultima (Datex, Helsinki, Finland;Figure 2). The Pitot tube has two apertures on its central axis facing in opposite directions. The monitor measures the pressure difference between these two apertures Delta P, which is proportional to the density and the square of the flow velocity (V2) of a gas. Flow is calculated from V.
The variable-orifice flowmeter we evaluated was the one incorporated in the Ohmeda 7900 ventilator (Ohmeda;Figure 3). Its flow sensor has a plastic flap orifice that opens wider with increasing flows. The pressure decrease across the orifice Delta P is measured as an index of flow. 
Finally, as examples of constant-temperature hot-wire anemometers, we evaluated the Drager PM8050 anesthesia monitor (Dragerwerk, Lubeck, Germany) and the FC10 (Taema, Antony Cedex, France). These devices determine flow based on the amount of electric power needed to maintain the electrically heated wire inside the sensor at a constant temperature against the cooling effects of a moving gas.
Of these flowmeters, only the Capnomac Ultima and PM8050 are designed to perform gas compensation. The Capnomac Ultima does so based on the measurements of the incorporated gas analyzers, whereas the PM8050 measures the thermal conductivity of a gas with a second designated wire inside the sensor. Other flowmeters lack gas-compensation mechanisms.
These four types of flowmeters were evaluated individually. In addition, based on the results showing that Magtrak IV provided stable readings regardless of the gas compositions, this device was used in all other experiments as a reference. All the flowmeters except for the two rotating vanes were calibrated before use according to the manufacturer instructions. By design, neither of the rotating-vane flowmeters can be calibrated by users.
The Study Setting
Flow was generated by a standard anesthesia ventilator (AV-500; IMI, Saitama, Japan) that was set to deliver square-waveform breaths at a minute volume of 5 1, a respiratory frequency of 10 breaths/min, and an inspiratory: expiratory (I:E) ratio of 1:2. It was connected via a wire-guarded, low-compliance breathing circuit to a lung simulator (Ohmeda, Essex, UK) that was set at its highest compliance (50 ml/cm water) and lowest resistance (0 cm water [middle dot] 1-1[middle dot] s-1) to minimize the peak pressure within the breathing system.
All flow sensors except the variable-orifice resistors (Ohmeda 7900) were placed in the inspiratory limb so they would directly receive flow from the ventilator. The Ohmeda 7900 system has inspiratory and expiratory flow sensors, both of which needed to be placed as designated to have the tidal volume displayed. Furthermore, the Ohmeda 7900 displays only expiratory tidal volumes. In all cases, a distance of at least 30 cm was ensured between each sensor to avoid possible interactions.
The room temperature was maintained at 25 [degree sign]C. The circuit was first filled with air, then with 100% oxygen, and then with increasing concentrations of either xenon or N2O in a balance of oxygen. The tidal volume displayed by each flow monitor was recorded and averaged for 10 breaths for each gas composition. During the recording, the fresh gas supply from the anesthesia machine was maintained at 250 ml/min. Finally, the breathing circuit was filled with air again and the measurement was repeated to ensure the stability of the ventilator and the sensors.
The rotating-vane and Pitot tube flowmeters were tested at xenon and N2O concentrations of 30%, 50%, and 79%. Xenon, 10%, also was used for the Pitot flowmeter. The variable-orifice resistor and the hot-wire anemometers were tested at 10% or 5% increment increases, respectively, between 0 and 80%(xenon) or 0 and 100%(N2O). For the hot-wire anemometers, the concentration of xenon was increased slowly (1%/min) because, during the preliminary experiment, an acute exposure of the PM8050 sensor to 50 or 70% xenon caused its heated wire to burn out instantaneously.
The rotating-vane and variable-orifice flowmeters were further evaluated at the minute volume settings of 3, 7.5, and 10 1. Nitrous oxide was omitted in these experiments.
The concentration of xenon was determined using a xenon monitor (AZ-720; Anzai Medical, Tokyo, Japan) that measures absorption of a characteristic X-ray as an index of xenon concentration. This device sampled the circuit gas from and returned it to the expiratory limb at approximately 400 ml/min. The concentrations of oxygen and N2O were determined using the Capnomac Ultima in all the experiments. An additional 200 ml/min of the circuit gas was sampled from the expiratory limb but was not returned to the circuit, because Capnomac Ultima adds air to it after analysis. To evaluate the Pito flowmeter, we used two different arrangements. In one setting, the flow sensor was connected to the Capnomac Ultima monitor used for the previously noted determination of oxygen and N2O concentrations of the circuit gas (i.e., a normal arrangement). In another, two Capnomac Ultima monitors separately performed flow measurements with a Pitot sensor and gas analyses. The gas analyzer port of the monitor for flow measurements was left open to the ambient air so that it would assume the circuit gas as air, regardless of its actual composition. The purpose of the second arrangement was to eliminate the effect of automatic gas compensation from the tidal volume readings of Capnomac Ultima.
The tidal volumes measured in the air at the beginning and at the end of each experiment were identical (differences < 0.5%), which confirmed the stability of the experimental systems. The peak pressure in the breathing system was 14 to 16 cm water when the minute volume setting was 5 l.
Both rotating-vane flowmeters provided strikingly constant tidal volume readings for the wide concentration ranges of both xenon and N2O and also with the various ventilator settings examined (Figure 4). A closer look at the data revealed that the Ohmeda 5400 underread at all the concentrations of xenon tested, although the errors were less than 5% compared with the value measured using air. Nitrous oxide had no effects. In contrast, Magtrak IV overread progressively with increasing concentrations of xenon and N2O (Figure 4, Figure 5, and Figure 6, open circles). Thus, its readings were 9.1% and 5.7% higher in the presence of 79% xenon (480 +/− 0 ml, mean +/− SD) and N2O (465 +/− 5 ml) compared with air (440 +/− 0 ml), respectively, when the minute volume of the ventilator was set at 5 l. This xenon-induced overreading was attenuated at higher flows; at the minute volume settings of 3, 7.5, and 10 l, the readings with 79% xenon were 8.5 (294 +/− 5 vs. 270 +/− 0 ml, 79% xenon vs. air), 4.6 (680 +/− 0 vs. 650 +/− 5 ml), and 2.4%(860 +/− 5 vs. 840 +/− 5 ml) greater than those with air, respectively.
The Capnomac Ultima Pitot tube flowmeter exhibited far more pronounced overreadings with xenon, such that its displayed tidal volume was nearly doubled when 79% xenon was used (Table 2). The concentration of xenon never differed by more than 2% from that of the balance gas displayed by the Capnomac Ultima (which was calculated as 100% minus the sum of concentrations of all measured gases [oxygen, N2O, carbon dioxide, and a volatile anesthetic]), suggesting that the Capnomac Ultima handled xenon as a balance gas. Nitrous oxide was associated with constant readings when the Capnomac Ultima used for flow measurement simultaneously analyzed the circuit gas, indicating effective gas compensation. However, when the gas analyzer port of the Capnomac Ultima monitor-Pitot sensor system was open to the ambient air and gas compensation was thereby “disabled,” N2O also caused overreadings (Table 2).
The variable-orifice flowmeter (Ohmeda 7900) also registered nonlinear increases in its readings with increasing concentrations of xenon and N2O (Figure 5). These increases were in close agreement with those observed with the Pitot flowmeter when it gas compensation mechanisms were disabled by disconnecting its gas analyzer sampling port from the breathing circuit. The Ohmeda 7900 does not perform gas compensation. A linear regression analysis by the least-squares method revealed that, for xenon
log (displayed TV/TV using air)= 0.508 x log[mean MW of xenon-O (2) mixture/(mean MW of air (= 28.8))]- 0.008 [r2= 0.995, P < 0.001; the 95% confidence interval (CI) of slope 0.496 - 0.520] and for N2O log (dislayed TV/TV using air)= 0.510 x log [(mean MW of N2O-O2mixture)/28.8]- 0.012 (r2=0.990, P0.001; the 95% CI of slope 0.471-0.549)
where TV and MW denote the tidal volume and molecular weight, respectively, and the mean MW is the sum of the molecular weight of each component in the mixture times its fractional concentration. The slopes of approximately 0.5 in these equations for both xenon and N2O indicate that the ratios of displayed tidal volumes to that measured using air were nearly proportional to the square root of the mean molecular weight of the gas used compared with that of air.
In contrast, both of the constant-temperature hot-wire anemometers progressively underread as the concentration of xenon was increased (Figure 6A). When a xenon concentration exceeded 45% for the FC10 and 70% for the PM8050, the tidal volume reading was no longer displayed, presumably because it exceeded the lower limit of measurement. Again N2O had minimal effects on the PM8050 (with gas compensation), but the FC10 (without gas compensation) was associated with overreading as much as 16% with low concentrations (10-20%) of N2O compared with air (Figure 6B). Therefore, the maximum tidal volume reading of the FC10 was 510 +/− 0 ml at N (2) O 10% and 20% versus 440 +/− 0 ml with air.
Xenon differentially affected four types of respiratory flowmeters examined. The rotating vane or turbine flowmeters were minimally affected by xenon, which is consistent with the generally accepted concept that they are relatively insensitive to the effects of changes in gas composition. The slight overreadings of the conventional Wright respirometer (Magtrak IV) produced by xenon and N2O appear to be related to the greater densities of these gases compared with that of air, because xenon, being more dense than N2O (Table 1), exerted more effects. Viscosity does not appear to be important, because N2O caused overreading despite its lower viscosity than that of air.
In addition, this xenon-induced overreading was attenuated at greater minute volumes. Such flow-dependent effects of gas density on the Wright respirometer also have been shown for sulfur hexafluoride, with a density that is 5.1 times that of air and 3.3 times that of N2O, and may be explained as follows. When air is used, it is well known that the Wright respirometer tends to underread at flow rates less than 20 l/min, and particularly less than 10 l/min, because the momentum or kinetic energy (or both) of air at these relatively low flows is not sufficient to overcome the inertia and friction of the vane completely. The actual flows we used were approximately 8 and 14 l/min when the minute volume of the ventilator was set at 3 and 5 l, respectively. At these flows, xenon and N (2) O would cause more effective displacement of the vane compared with air because they would increase the momentum and kinetic energy of gas flow because of their greater densities. This would have resulted in the observed overreadings. At higher flows, however, air would carry a greater momentum and would rotate the vane more efficiently, rendering the effects of xenon less apparent.
The Pitot tube flowmeter registered a significant and concentration-dependent increase in its readings when xenon was used. This can be accounted for almost completely by a simple mathematic equation, as follows. When a mixture of xenon and oxygen flows at a velocity V through the Pitot sensor depicted in Figure 2, the pressure difference between the two apertures Delta PXe-O(2) is equal to the dynamic pressure of the gas:Equation 1where [small rho, Greek]Xe-O(2) denotes the density of a xenon and oxygen mixture. The Capnomac Ultima measures Delta PXe-O(2) and calculates V, which is integrated with respect to time and multiplied by the cross-sectional area of the sensor to yield the tidal volume. 
However, in these calculations, the Capnomac Ultima handles xenon as a balance gas and assumes that it is nitrogen because it is not equipped with a xenon analyzer. Consequently, when the Capnomac Ultima monitor measures flow and analyzes the composition of the circuit gas, [small rho, Greek]Xe-O(2) would be falsely replaced by the density of a nitrogen and oxygen mixture, with its nitrogen concentration equal to that of xenon ([small rho, Greek]N(2)-O(2)), leading to a miscalculation of the flow velocity as:Equation 2Because xenon is more dense than nitrogen (Table 1), a combination of Equation 1and Equation 2dictates that the flow velocity, and therefore the tidal volume, would be over-estimated by the factor of Equation 3
The density of a gas mixture is proportional to its mean molecular weight (MW) because equivalent moles of gases occupy the same volume assuming constant temperature and pressure. Therefore, Equation 4
Similarly, when the Capnomac Ultima Pitot sensor system sampled the ambient air for gas analysis instead of the circuit gas, the density of air ([small rho, Greek]air) would be used instead of [small rho, Greek]N(2)-O(2) in Equation 3, leading to an overestimation of the tidal volume by the factor of Equation 5
This reasoning is consistent with our results because the degrees of xenon-induced overreading relative to air (TV/TVairin Table 2) were nearly identical to V'/V in Equation 4and Equation 5[(MW/MWN(2)-O(2))1/2and (MW/MWair)1/2in Table 2] when the gas analyzer port of the monitor Pitot sensor system was connected to the circuit and was open to the ambient air, respectively.
In fact, this density effect was not unique to xenon but was also observed with N2O when the gas analyzer port of the Capnomac Ultima was open to air so that it would not perform gas compensation. In this case, the magnitude of overreading would be given by substituting PN(2)-O(2) for P (Xe-O)(2) in Equation 5:Equation 6
This is presented in Table 2in the column labeled “(MW/MWair)1/2” and is close to the observed readings relative to air in the presence of various N2O-oxygen combinations (TV/TVairwith the gas analyzer port open to air in Table 2).
The variable-orifice flowmeter (Ohmeda 7900) also overread with xenon and N2O, the magnitude of which was nearly proportional to the square root of the mean molecular weight (and thus density) of the gas mixture used. Again, this can be explained mathematically based on the fluid dynamics. When a xenon and oxygen mixture enters and leaves the variable-orifice sensor depicted in Figure 3at a velocity V, the pressure decrease across the orifice Delta P measured by the flowmeter is Equation 7where [small lambda, Greek] is the pressure-loss coefficient and depends on the area and shape of the orifice. Because these latter factors are determined by the bend of the plastic flap, and Delta P is the force causing this bend, [small lambda, Greek] is a function of Delta P and can be rewritten as [small lambda, Greek](Delta P) in Equation 7:Equation 8However, because Ohmeda 7900 does not make corrections for the use of xenon and assumes [small rho, Greek] as constant ([small rho, Greek]c), it would falsely calculate the flow velocity as Equation 9From Equation 8and Equation 9, the Ohmeda 7900 would over-read by the factor of Equation 10Because [small rho, Greek]cis constant, Equation 10dictates that the magnitude of overreading should be proportional to the square root of the density. Evidently, this reasoning is also valid for N2O because it involves no assumptions unique to xenon. Furthermore, it has been reported that this density effect adequately explains the influences of temperature, humidity, and altitude on other flowmeters of this type. 
In contrast to the devices discussed before, the hot-wire anemometers underread with xenon, such that the error exceeded 90% at concentrations likely to be encountered in the clinical setting (i.e., 45-70% xenon or approximately 0.65 to 1 minimal alveolar concentration ). Nitrous oxide, however, caused only modest overreading even in the absence of gas compensation (i.e., with the FC10), which is consistent with previous observations. [15,16]With this type of flowmeter, a measure of flow is the heat loss from the hot wire, which is influenced not only by flow but also by the thermal properties of a gas. [15-17]Consequently, our results are accounted for at least in part by the distinctly lower thermal conductivity and capacity of xenon compared with those of other commonly used gases (Table 1). However, to what extent these factors actually contribute to the observed errors cannot be demonstrated mathematically, because no theoretical Equation hasbeen established that accurately describes heat loss from the hot wire. [16,17]
This study has two implications. First, among the four types of flowmeters studied, only the rotating vanes are sufficiently accurate for use during clinical anesthesia with xenon. Other types require xenon-oriented calibration or gas compensation processes, although the readings of the Pitot tube and variable-orifice flowmeters can be corrected by simple mathematics. Second, because xenon produces concentration-dependent measurement errors in many flowmeters, one can conversely determine a xenon concentration based on these errors. More specifically, because the readings of each flowmeter depend on the actual flow and the concentration of xenon, a combination of two different types of flowmeters would provide both of these parameters. The exception would be the combination of a Pitot tube and a variable-orifice device because they are associated with identical measurement errors. Xenon analyzers involving such principles may deserve further investigation, because currently available ones are complex and expensive.
The authors thank Mieko Saito, M.S., for preparing the figures, and Kunio Suwa, M.D., for critically reading the manuscript.
# Marx T, Froba G, Bader A, Wagner D, Georgieff M: Investigation of a xenon-recycling technology for anaesthesia. 11th World Congress of Anaesthesiologists, Sydney, Australia, 1996, p. 1283.