INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion Appendix References

INDIRECT CALORIMETRY by the gas-exchange method in the spontaneously breathing or mechanically ventilated subject is an exacting process that challenges the accuracy and error sensitivity of the measuring instrumentation. Several in vitro techniques for the validation of these devices have been described in the literature (1-8, 10-15, 19). They vary in complexity and difficulty of construction, and they often require substantial calibration before being useful as validation tools. Solvent-burning lung models have been an attractive validation method because of their apparent simplicity and the availability of high-purity fuels that have predictable respiratory quotient (RQ) values. When completely and properly burned, the fuel should consume only O₂, and combustion by-products should contain only water vapor and carbon dioxide (CO₂). However, predictable and adjustable rates of O₂ consumption (O₂) and CO₂ production (CO₂) have not generally been possible with this type of model, and additional test methods are required to validate these measurements (1, 4, 6, 12, 14, 18).

The methanol-burning lung model proposed by Damask et al. (3) has been used as a method for validating the accuracy of gas-exchange measurement instruments that measure O₂ and CO₂. Methanol is a simple alcohol that can be completely burned in an O₂-nitrogen environment. This follows from the equation

(1)

Methanol is burned in an alcohol lamp with a noncombustible wick, such as asbestos, within a closed chamber that is ventilated with a constant or intermittent flow of O₂-bearing gas. The stoichiometric equation describes the quantitative O₂, CO₂, and production of water as products of combustion (1). The ratio of CO₂ to O₂ in the burning of methanol yields an RQ of 0.667.

The reaction rate, and hence the fuel-consumption rate, is determined by the O₂ available to the flame, the fuel flow to the burning region, temperature, and combustion product clearance. In conventional methanol-burning lung models, these parameters are difficult to control and equally difficult to sustain over the course of a gas-exchange measurement run. Although the RQ should not vary in the conventional models, the burn rate invariably does. At elevated O₂ concentrations, it is increasingly difficult to sustain a pure methanol flame without other unwanted by-products (17).

To validate the accuracy of a new gas-exchange measuring instrument, we considered the conventional methanol-burning model and were confronted with some of its undesirable limitations, primarily the inability to adjust the fuel burn rate and hence achieve a stable and predictable O₂ and CO₂. Because our gas-exchange instrument is capable of continuous, long-term (>4-5 h) automated operation over a wide range of O₂ concentrations, minute ventilation, and O₂ and CO₂ values, we sought to enhance the conventional methanol-burning validation method to satisfy the following requirements: 1) stable and adjustable O₂ and CO₂ rates, 2) wickless burning of fuel, 3) combustion in atmospheres with O₂ concentration >80%, 4) ability to replenish fuel supply while a test run is in progress without breaching the vessel or interrupting gas-collection activities, and 5) ability to ignite and extinguish the flame without breaching the vessel.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion Appendix References

A cylindrical acrylic chamber 20 cm in diameter and 36 cm in height was fabricated. It had a removable, gasketed cover and floor of the same material. Acrylic was used for the chamber because of its availability and ease of machining. Alternatively, other materials may be used, such as heat-resistant glass or polycarbonate plastic. A single 15-mm hole was drilled into the cylinder wall 9 cm from the floor. This port accommodated an electrical igniter. The cover was fitted with two ports: one for the entry of fresh gas and the other for the removal of combustion by-products. The size of the holes was sufficient to accommodate ventilator or pulmonary function fittings. The inspiration port was fitted with an 11-mm ID, 25-cm silicone tube that directed fresh gas toward the lower one-third of the cylinder. Two gas-sampling leur-lock fittings penetrated the top cover. The floor of the cylinder had two supporting blocks that raised the entire assembly 10 cm above the laboratory table.

A stainless steel shallow cone funnel burner (Fig. 1) was designed with an integral gas-feed system that supplied the funnel mouth directly with an adjustable fresh gas flow (FGF). At its mouth, the funnel diameter was 31.75 mm; it tapered linearly to 1 mm, with a cone angle of 120°. The funnel assembly was 50.8 mm in diameter and was machined from a single piece of cylindrical stainless steel. Six symmetrically spaced gas-feed lines surrounded the funnel cone and emerged into a circumferential groove just above the funnel mouth. These gas lines were fed from a common plenum at the base of the assembly. The plenum had a gasketed cover and a gas-feed fitting. The FGF was diverted from the inspired-gas limb that ventilated the lung model; thus no additional minute ventilation was added. The flow was controlled by a small, electrically powered, diaphragm pump. The pump, via the gas-feed lines and groove, was able to direct fresh gas toward the center of the burner at a rate of 200-4,000 ml/min. This flow swept away CO₂ and water vapor while providing fresh gas of a known O₂ concentration for combustion. The FGF passing through the wall of the assembly provided cooling as well.

View larger version (25K): [in this window] [in a new window]   Fig. 1.   Cross section of funnel burner assembly. Free gas flow (FGF) enters plenum at base, rises through 6 symmetrically spaced vertical channels in wall of assembly, and emerges into circumferential groove at funnel mouth.

The burner was supported by a 3-mm-diameter, 10-cm-long stainless steel tube that penetrated a neoprene stopper which was placed into a 15-mm hole drilled into the cylinder floor. This tube also acted as the fuel source line for the burner. A small electrically powered fan, which enhanced convective mixing in the chamber, was suspended from the cover and hung within the upper quarter of the cylinder. The wires powering the fan penetrated the cover via a gas-tight fitting. The cylinder was filled with water to within 5 mm of the rim of the funnel. This water jacket completely surrounded the burner and helped to maintain the burner assembly at temperatures below the methanol boiling point. The top of the cylinder was then secured. The configuration of the assembly is shown in Fig. 2.

View larger version (31K): [in this window] [in a new window]   Fig. 2.   Assembly of lung model.

The stainless steel tube that supported the burner was connected to a length of noncompliant nylon pressure tubing. This tubing was used to avoid level changes within the funnel burner when the model was ventilated with a pressure-varying flow source. Materials that are affected by or are soluble in methanol should be avoided. The tubing was connected with appropriate fittings to a three-way stopcock at a convenient distance from the cylinder. A 20-ml glass syringe was attached to one of the two remaining ports of the stopcock and was used to prime the funnel with fuel. The remaining port was connected by another length of pressure tubing and fittings to a syringe-infusion pump (model 55-2275, Harvard Apparatus, South Natick, MA). Infusion devices with peristaltic or interrupted-flow delivery patterns may cause varying levels of methanol within the funnel while the pump runs through its mechanical cycle and can thus yield varying burn rates. Depending on the measurement interval of the gas-exchange device to be validated, this effect may or may not be a problem. The infusion pump used should be calibrated gravimetrically, and its flow pattern should be carefully assessed. High-purity methanol was used as fuel (Optima grade, Fisher Scientific, Pittsburgh, PA). The connecting tubing between the infusion pump and funnel was primed, and the funnel was arbitrarily filled to approximately two-thirds of its capacity. The infusion pump was set at a rate that would generate the desired O₂ and CO₂ (see APPENDIX). If a change in O₂ and CO₂ was desired, all that was required was a proportional change in the fuel pump rate.

The input and output ports were located at the top of cylinder and were connected to the device measuring gas exchange. Alternatively, if a mechanical ventilator was used, the inspiratory and expiratory limbs of the patient circuit were separated and connected to the corresponding locations on the chamber. The exhaust stack of the ventilator was then connected to the unit to be validated. A compliant rubber reservoir bag was connected with a T fitting into the expiratory limb to allow for volume expansion when tidal volumes were delivered through a ventilator. The diaphragm pump was connected to a variable-voltage power supply and was adjusted to deliver FGF appropriate to the desired O₂. The input port of this pump was connected to the input limb of the fresh gas source. The output port of the pump was connected to a fitting in the base of the burner assembly. We found empirically that a FGF that was three to five times the O₂ was adequate and did not extinguish the flame. This flow was sufficient to sweep away the CO₂ of the burning fuel pool and provided sufficient quantities of oxidizer to the flame to ensure complete combustion. The absolute value of FGF to the burner assembly was not critical. The only requirement was that it be stable and greater by a factor of at least two than the expected O₂. Inadequate FGF can result in the production of CO as well as possible extinction of the flame. An electrical igniter was positioned over the mouth of the funnel via the port in the side of the cylinder, and the fuel was ignited.

The internal volume of the apparatus described was ~11 liters. This volume had to be washed out with gas of the desired fraction of inspired O₂ (FI_O2) before accurate measurements were made. It was possible to ventilate the model, as described, with minute ventilation of between 5 and 30 l/min without adversely affecting the burn rate. Either pulsatile or continuous gas flow could have been used.

Under conditions of high fuel-infusion rates and high FI_O2, substantial amounts of heat are generated. Excessive heat must be removed to prevent the acrylic cylinder and the pulmonary function tube from melting. High rates of ventilation (>20 l/min) as well as a cooling jacket over the outer wall of the chamber are recommended. The flame can be extinguished by stopping the infusion pump and allowing the remaining fuel within the funnel to be consumed. Alternatively, after the pump is stopped, the stopcock may be directed toward the 20-ml priming syringe, and the remaining fuel in the funnel can be withdrawn. Fire precautions were necessary.¹ A mass spectrometer (Random Access Mass Spectrometer; Marquette Medical Systems, Milwaukee, WI) analyzing from 4 to 250 atomic mass units (AMU) was used to evaluate the contents of the effluent gas leaving the chamber. A CO analyzer (Draeger 190; National Draeger, Pittsburgh, PA) and a series of sample collection tubes (Draeger alcohol 100/A; Draegerwerk, Aktiengesellschaft, Germany) were used to detect CO and unburnt methanol gas respectively.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion Appendix References

The lung model was tested at varying rates of fuel infusion and ventilated with gases having an O₂ content between 21 and 100%. A clean burn was evidenced by the absence of soot and the presence of a clear blue flame. The intensity of the flame varied with the O₂ concentration, from being almost invisible in ambient light at room air concentrations to having an intense blue color at FI_O2 = 1.00. At elevated FI_O2 levels, the lowest infusion rates could not sustain a flame, because the fuel-consumption rate increased in relation to the FI_O2. At the lower fuel-infusion rates, the funnel burner emptied, and the flame was extinguished. Conversely, at FI_O2 = 0.21, the highest infusion rates breached the confines of the funnel (Table 1). Given the dimensional constraints of the burner used, the O₂ rates can be varied from 69.3 ml/min at an FI_O2 of 0.21 to 1,524.6 ml/min at an FI_O2 of 1.00 (Table 1). As the FI_O2 was varied from 0.21 to 1.00, only the expected gas species (i.e., H₂O, N₂, O₂, CO₂, and Ar) were observed. The exhaust-gas concentrations of CO and unburnt methanol are reported in Table 2. The magnitudes of these values indicate that, in this model, the error caused by CO and unburnt methanol is <0.005% of the predicted values of O₂ and CO₂, which is insignificant.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion Appendix References

Our implementation of the methanol-burning lung model addresses a number of constraints that limit the effectiveness of this apparently simple validation technique. There are problems with implementing wick-based solvent burners that prevent their use in a consistent and quantitative manner over the full range of FI_O2. In a sense, wick-based lung models are, in fact, quantitative. The measured volume of methanol consumed by the model after a period of time can be calculated (APPENDIX), and the unit being validated should account for the O₂ consumed and the CO₂ produced. For those gas-exchange instruments for which results are reported over more extended periods of time, this could be a sufficient method of validation (10). Devices that report their results on a minute-by-minute or breath-by-breath basis require a more stable and adjustable validation tool.

A constant flow of fuel into the burning region of the wick should yield a constant burn rate and, therefore, a stable O₂ and CO₂. A burning wick that maintains a constant geometry presents a constant capillary pressure to the fuel supply below it. Unfortunately, capillary flow rate is difficult to predict and to maintain. Changes in wick geometry that result from wick handling make repeatable fuel-consumption rates difficult to achieve. In addition, the fuel level in the alcohol-lamp reservoir drops as fuel is consumed, thus changing capillary feed of fuel to the wick. Furthermore, attempting to replenish the fuel in these burners while the wick is lit requires the fuel-infusion rate to be matched with the fuel-consumption rate. This is not a trivial task, because the absolute fuel-consumption rate of the lamp may not be known. These conditions are magnified as fuel consumption increases with elevated FI_O2 and ventilation.

As O₂ concentration is increased, the effects of increased CO₂ must be considered as well. CO₂ in the immediate proximity of the flame will displace and reduce O₂ available to the flame, causing further changes in the burn rate. Initially, the flame appears to burn well, but it may be extinguished after a short period of time. Although the vessels used in conventional methanol-burning models are ventilated to remove CO₂ and provide fresh gas for combustion, the dynamics of this clearance process have not been well considered.

In the wick-based models, the alcohol lamp is placed on the floor in the center of a 10- to 15-liter chamber (3, 13, 15). The fresh gas and exhaust are provided by a conventional ventilator hose set. Although the O₂ and CO₂ rates are modest, on the order of 200-300 ml/min at O₂ concentrations of 21%, the vessel is often ventilated at minute ventilation rates of 5-15 l/min. The ventilating gas cannot be directed across the alcohol lamp, because large convective flows interfere with the burning process and can possibly extinguish the flame. Additionally, in these models, both the entry of fresh gas and exhaust of combustion products are from the same orifice. This results in a large component of the fresh gas administered during the inspiratory phase of the respiratory cycle being withdrawn on expiration.

Efficient clearance of CO₂ at the level of the alcohol lamp by these otherwise large convective flows is not assured. Consequently, a significant amount of diffusion is required before the CO₂ can reach those areas of the vessel where convective flow will carry it away. The diffusive component of gas mixing is relatively slow at atmospheric pressures and cannot be relied on to clear sufficient CO₂ in the geometries of conventional methanol-burning lung models. Consequently, at higher O₂ concentrations, the increased CO₂ rate of the flame displaces O₂ required to sustain a desired combustion rate. Changing ventilation at any level of FI_O2 will also change the CO₂ clearance rate. We have measured the CO₂ concentration in the immediate proximity of a burning lamp at high levels of FI_O2 and have observed fluctuating CO₂ values that can be >70% and comparable changes in O₂ concentration. Because CO₂ is heavier than either component of air, stratification of the internal atmospheric environment of the model can change the FI_O2 presented to the flame and alter its fuel-consumption rate.

Under certain conditions, this process may starve the flame for O₂ and begin producing unwanted CO before extinguishing the flame. Wick-based methanol-burning lung models must control many factors to achieve a stable and repeatable O₂ and CO₂. Changes in ventilation, FI_O2, and fuel level will all have an effect on the fuel-consumption rate. Reproducing a given wick geometry from one session to the next is difficult as well. The approach taken in our model addresses these issues.

To avoid the unpredictability and lack of adjustment of using a wick as the site of the combustion process, a burner was conceived with a variable surface area in which an open pool of methanol was ignited and burned directly. Combustion models developed for burning methanol and other fuels suggest that a single layer of molecules, at a small distance from the bulk liquid methanol, are actually ignited and participate in combustion (9). Some of the thermal energy released is used to volatilize the liquid methanol and provide a source of fuel to sustain the flame. The reaction rate of methanol combustion is described by a series of complex differential equations that account for the availability of fuel and O₂, the rate of heat released, ambient temperature, and the effect of produced CO₂ and water vapor. In simplified terms, the flame size that arises, and hence the consumption rate of the fuel, increases monotonically, although not linearly, in relation to the fuel surface area and/or the availability of oxidizer.

This concept was tested with a heat-resistant glass funnel that was fed, via its stem, by methanol fuel provided by a syringe pump. The funnel burner is a geometric structure that presents a liquid surface cross-sectional area proportional to the volume of fuel in the cone. The initial level of the fuel within the funnel was established arbitrarily. After the methanol was ignited, it burned with a clear, pale blue flame. While methanol was consumed, the pump constantly replenished the fuel supply in the funnel cone. The fuel burn rate adjusted itself whether the funnel was initially over- or underfilled for a reasonably wide range of infusion rates. The upper portion of the funnel provided larger cross-sectional areas to the fuel, causing a burning rate that consumed excess methanol, while the smaller cross sections toward the funnel stem consumed fuel at lower rates and allowed the fuel level to rise. Over a period of 1-2 min, the infusion rate of the fuel-infusion pump and the consumption of methanol by the flame reached a steady state. This system was self-adjusting and acted as a passive servoloop, the set point of which is determined, within limits, only by the fuel-infusion rate.

With  FI_O2 held constant, the fuel level within the funnel, and hence the cross-sectional area of the pool, was constant as well. When this equilibrium was achieved, the infusion rate of methanol provided a basis for calculating a quantitative O₂ and CO₂. These consumption and production values can thus be easily translated into rates that cover the physiological range of humans and animals (see APPENDIX). However, the glass funnel was unable to sustain a flame at FI_O2 >0.4. While holding the fuel-infusion rate constant and increasing the FI_O2, the fuel level within the funnel cone and the flame stabilized in a region closer to the funnel stem, where a smaller cross section for the fuel was available. The flame's color changed from an almost invisible blue to an incandescent yellow before being extinguished. After the flame was extinguished, the methanol pool was also observed to have a milky appearance compared with the colorless appearance of pure methanol. This milky appearance is characteristic of a methanol-water mixture. We hypothesize that increasing levels of CO₂ and water vapor in the proximity of the flame were inadequately cleared in the lower regions of the funnel burner, despite the thermal convective currents that arose from combustion. This caused the flame to be starved for O₂ and the fuel to be polluted by water. A combination of these conditions eventually extinguished the flame.

The stainless steel funnel burner described in MATERIALS AND METHODS was designed to address these issues. A shallow cone angle was used to bring the fuel pool closer to the level of the funnel mouth over a wide range of fill levels. The clearance of both CO₂ and water vapor was further assured by an integral FGF provided radially at the edge of the funnel cone. The FGF increased convective currents around the flame without extinguishing combustion. After these concepts were implemented, we were able to achieve complete combustion with FI_O2 ranging from 0.21 to 1.00.

The simplicity of the funnel burner belies the underlying complexity of the combustion process that is being regulated. The reaction rate of methanol and O₂ ultimately determines the fuel-consumption rate. A burning pool of fuel is constantly engaged in a chaotic change in the availability of fuel and oxidizer because of the thermal convection currents that arise from the flame and the displacement of oxidizer and fuel when combustion products are generated (16). In the short term, this results in a varying fuel-consumption rate. Over a range of fuel-infusion rates and varying oxidizer availability and temperature, the funnel geometry permits the fuel level to rise or fall rapidly, thus continually adjusting the cross-sectional area of the fuel to accommodate these varying conditions. This automatically servoregulates or adjusts the fuel-consumption rate to equal the fuel-infusion rate. Consequently, there is no need to precisely set the FGF flow or the initial level of the fuel within the funnel. For the system to achieve a steady state, a stable fuel-infusion rate and relatively stable FI_O2 and FGF are required. However, all that is required to predict the O₂ and CO₂ of this model is knowledge of the fuel-infusion rate. Absolute values of FGF or FI_O2 are not required to determine fuel consumption. As long as the funnel does not empty or overfill, the flame will consume precisely the fuel infused. Doubling or halving the fuel-infusion rate (within the limits indicated by Table 1) will, after equilibration, double or halve the O₂ and CO₂ while maintaining a RQ = 0.667. Changing the FI_O2 will also not change the O₂, CO₂, and RQ after a short period of equilibration (again, within the limits indicated in Table 1). Within these constraints, there is no necessity to monitor the funnel level or make adjustments to achieve a specific fuel level within the cone. The geometry of the funnel is also not particularly critical. A steep cone angle should be avoided, because burning in the smaller cross-sectional areas positions the flame away from adequate ventilation. A simple conical funnel was chosen for the ease of machining it. Other shallow funnel geometries that continuously change cross-sectional area with fill level (hyperbolic, parabolic, etc.) should work as well.

A wide range of production and consumption rates is easily established by adjusting the fuel-infusion rate (Table 1). The reaction rate increases either in the presence of larger fuel cross-sectional areas or at higher FI_O2. Consequently, if the fuel-infusion rate is held constant and the FI_O2 is increased, the fuel level in the funnel drops until it reaches a smaller cross-sectional area of the funnel that is consistent with the steady-state fuel consumption at the elevated FI_O2. Conversely, if FI_O2 is reduced, the level of the fuel in the funnel will rise to a larger cross-sectional area. Thus, within the geometrical constraints of the funnel burner, O₂, CO₂, and RQ are independent of the absolute values of FI_O2, minute ventilation of the chamber, and FGF. At equilibrium, they are dependent only on the fuel-infusion rate.

Mass spectrometer analysis of the effluent gas confirmed the completeness of the burn. The mass spectrum was examined from 4 to 250 AMUs; at AMUs >50 and <10, there were no significant signals above the noise floor. The fuel-infusion rate was gradually increased from 20 to 60 ml/h as FI_O2 was increased from 0.21 to 0.99 to maintain a relatively uniform fill level within the funnel. The expected results of the burning process includes H₂O (AMU = 18), N₂ (AMU = 28), O₂ (AMU = 32), Ar (AMU = 40), and CO₂ (AMU = 44). Additionally, the method that the mass spectrometer uses to generate its ionized-particle beam also produces monoatomic species of C (AMU =12), N (AMU = 14), and O (AMU = 16). Small spectral peaks were apparent for these species as well. The mass spectrometer is not always able to distinguish between different molecules with identical molecular weights. Thus CO (molecular weight of 28) is indistinguishable from N₂, and methanol (molecular weight of 32) cannot be separated from O₂. Therefore, we used alternative instruments to detect these possible components of the effluent gas. The concentrations of CO and volatilized methanol in the effluent gas were insignificant (Table 2).

Measurement of the performance of a validation technique is both technically and philosophically challenging. Although we confirmed the nominal characteristics of this model with a Douglas collection bag method, this is by no means a technique sufficiently accurate to confirm the performance of a validation method. The consumption of all of the fuel infused was verified over the range of infusion rates and FI_O2. The purity of the fuel used, the gravimetrically determined accuracy of the fuel-infusion rate, the presence of the expected components identified in the effluent gas, as well as the absence of combustion products that would suggest an alternative chemistry for methanol combustion, ensure the validity of this approach.

Although the system described was evaluated for its ability to predictably consume methanol, the process described suggests that other low-molecular-weight solvents could achieve comparable success as completely burned fuels.

Gas-exchange and indirect calorimetry measurements are complex and sensitive to a wide range of errors. Thus, validation of these measurements must be carefully considered lest they introduce a new set of comparably complex and error-prone techniques. We have developed an extended validation method for gas-exchange measuring equipment with the use of burning methanol. The apparatus described is simple to reproduce, use, and calibrate. It derives its accuracy from the chemistry of methanol combustion, high-purity methanol fuel, and readily available, highly accurate gravimetric measurement of the fuel-infusion rate. O₂, CO₂, and RQ, in this model, are largely independent of many of those factors that affect other validation methods. This will facilitate the validation of instruments measuring gas exchange over a wide range of quantitative and repeatable test conditions.