A Hand-Held Device to Measure Oxygen Uptake: Performance Characteristics, Patient Selection and the Propagation of its Measureme

The classic oxygen-based Fick Principle has long been considered the gold standard for measuring cardiac output, but its rigorous implementation is being used less and less frequently during cardiac catheterization and in intensive care because, in our opinion, the instrumentation necessary to measure oxygen consumption from a patient’s expired gases is cumbersome, expensive, time-consuming and difficult for unskilled operators to use effectively. Thus, instead of actually measuring the patient’s rate of oxygen consumption, many cardiac catheterization laboratories simply estimate it from regression equations based on the patient’s height, weight, age, sex, heart rate or other anthropomorphic data.^1–4 Unfortunately, studies^5–11 have shown consistently and convincingly that using an assumed or estimated rate of oxygen consumption, instead of a measured value, “frequently results in large errors in the determination of cardiac output”.⁵ Even worse, inaccurate cardiac output determinations subsequently affect clinical decisions regarding ventricular function, valve areas and systemic and pulmonary vascular resistances.
An inexpensive, fist-size device called MedGem^®(Microlife USA, Inc., Dunedin, Florida) recently became available for measuring oxygen consumption. A series of studies has compared the new device’s measurements of oxygen uptake with those made by conventional metabolic carts, and the results appear promising.^12–20 However, little is known about the characteristics of the device that would affect its utility during cardiac catheterization and in intensive care. Therefore, the purposes of this report are: (1) to delineate the advantages and limitations that would affect its use in these settings; (2) to define the patient population in which its use would be safe and effective; and (3) to estimate the accuracy of cardiac output determinations based on its measurements of oxygen consumption.
All of the measured variables in the Fick equation, the arterial and mixed venous oxygen concentrations and the oxygen consumption rate, contain random measurement errors that are propagated into the calculated cardiac output.^21,22 Therefore, the accuracy of cardiac output determinations based on measurements of oxygen consumption by the MedGem or conventional metabolic carts cannot be assessed from the inaccuracy of those instruments alone, but must take into account the measurement errors of the oximeters and co-oximeters used to analyze blood oxygen. Thus, an additional purpose of this report is to present a statistical model that calculates cardiac output error as a function of the random errors in the measurements of arterial and venous oxygen concentrations and the oxygen consumption rate. Using the methods presented in the Appendix, interested readers can compute the cumulative measurement error in cardiac output determinations made with the particular set of instruments in their institutions. Our results support the conclusions that the MedGem is suitable for a wide range of adults and children at rest (but not during exercise), and that cardiac output determinations made with the MedGem’s measurements of oxygen uptake are clinically acceptable if the oximeter used to measure oxygen in arterial and mixed venous blood is also sufficiently accurate.

Materials and Methods

Description of the test instrument. The MedGem can be connected to the patient with a disposable mouthpiece that is used with a nose clip. A hard-shell facemask is also available that is used with a soft, disposable insert. Both the mouthpieces and facemask inserts are equipped with a microbial filter. A 12 VDC wall transformer powers the MedGem. The small power jack also serves as a serial data port (RS-232). The device weighs only 110 grams, and has dimensions of 5.5 x 5.5 x 11.5 cm.
The MedGem starts recording data as soon as the patient’s first breath is detected. At the completion of the measurement, the device beeps, the indicator light changes to amber and the VO₂ reading is displayed on the LCD.
Data collection. In order to obtain information regarding the new device’s performance characteristics, we reviewed the previous publications that compared it with reference methods, reviewed the manufacturer’s in-house studies and interviewed the manufacturer’s engineering staff to obtain information that was not in the published specifications. To assess the precision (repeatability) of the new device, we made paired measurements on healthy volunteers. To explore the effects of a range of tidal volumes and breathing frequencies, we performed experiments on anesthetized dogs, pigs and goats. We also timed each measurement to determine the average measurement time. These experiments were conducted under protocols approved, respectively, by our university’s Institutional Review Board and the Institutional Animal Care and Use Committee.
To measure the dead space that the device adds to a patient’s airways, we used a self-sealing film (Parafilm, Sargent Welch, Buffalo, New York) to block the openings of the air tubes, facemask and mouthpiece, and weighed them before and after filling them with water.
In the experiments on healthy volunteers, we attempted to make all measurements in a basal metabolic state so that the repeated measurements would be as consistent as possible. Each volunteer had had no food or caffeine-containing beverages for 12 hours, had not exercised in 24 hours, had not smoked for 12 hours and had rested for 10 minutes before the first measurement. After the first measurement, there was a 10-minute rest period followed by the second measurement. All measurements were made in the supine position with the subject using one hand to hold the MedGem in place. On one volunteer, one measurement per day was made on 10 different days. To check the maximal oxygen uptake rate the device can measure, the same subject walked on a treadmill at gradually increasing speeds until an error code was displayed.
To assess the repeatability of cardiac output determinations made with the MedGem, we performed an experiment on a mongrel dog that weighed 31.8 kg. The animal was anesthetized with sodium pentobarbital (30 mg/kg), and an endotracheal tube was inserted. Cutdowns were performed, and the right jugular vein and the femoral arteries and veins were isolated. After the surgical procedures were complete, sodium heparin (200 U/kg) was administered intravenously. Arterial pressure was measured with a transducer connected to a catheter inserted into one of the femoral arteries and advanced into the abdominal aorta; the same catheter was used intermittently to take samples of arterial blood. A catheter in a femoral vein was used to administer additional anesthetic and return sampled blood to the animal. For access to mixed venous blood, a catheter was inserted into the right external jugular vein and was advanced until the tip of the catheter was in the pulmonary artery. The catheter was then connected to a pressure transducer so that the position of the tip of the catheter could be verified by the morphology of the pressure waveform. A Grass multichannel chart recorder was used to record the aortic and pulmonary arterial pressures. In order to connect the dog to the MedGem, we cemented a standard 15 mm endotracheal tube connector into a MedGem mouthpiece. The MedGem was interfaced to a personal computer with a serial cable and software available from Microlife.
During each cardiac output determination, we started the MedGem and then began to withdraw 10-mL samples of blood from the aorta and the pulmonary artery. We withdrew these samples slowly so that each syringe contained blood from multiple breathing cycles. Then two 50-µL cuvettes were filled from each syringe and analyzed on an AVOXimeter 1000E (Avox Systems, Inc., San Antonio, Texas). The rest of the blood in the syringe was returned to the animal. After the MedGem completed its measurement, a 5-minute rest period was observed before the next cardiac output determination was made. We also attempted to make noninvasive measurements of oxygen uptake only in pigs and goats.

Results and Discussion

Bias, precision and accuracy. We assessed the MedGem’s precision three different ways. One subject measured her rate of oxygen uptake under the previously-defined basal conditions on 10 different days, we made paired, same-day measurements on 12 volunteers, and we made 6 consecutive measurements on an anesthetized dog. The results from 10 repetitions on the same individual were a mean of 195.60 ± 9.20 mL O₂/min, and a CV of 4.71%. Similarly, the paired measurements on 12 volunteers (Table 1) yielded a mean difference between trials of 4.67 ± 9.52 mL O₂/min, a relative error of 4.6% and a within-run precision of 7.2 mL O₂/min. The 6 consecutive canine measurements had a mean of 141.3 ± 4.0 mL O₂/min and a CV of 2.85%. Combining consecutive measurements into 3 pairs yielded a within-run precision of 3.87 mL O₂/min. Our estimates of precision are consistent with those of Melanson et al¹² who reported a standard deviation (SD) of “trial-to-trial” differences of 567 kj/day (19.5 mL O₂/min); dividing this by the mean of 6,697 yields a relative error of 8.47%. Similarly, Alam et al¹⁷ reported a “within session” coefficient of variation of 8.2–9.6% for the MedGem versus 2.5–4.5% for the DeltaTrac. These estimates of imprecision contain contributions both from instrument noise and biological inconsistencies. Therefore, it is interesting to compare them with the manufacturer’s in-house study in which a mechanical simulator was used to simulate ventilation and metabolism.³¹ Mechanical simulators have often been used to eliminate biological variability in assessments of the intrinsic repeatability and accuracy of metabolic carts.^32,33 Twenty-two BodyGems were each tested 6 or 7 times on the mechanical simulator so that a coefficient of variation could be computed for each instrument. The CVs ranged from 0.33% to 2.86% and averaged 1.45%. In units of oxygen uptake, the average standard deviation of repeated measurements would be 2.67 mL O₂/min.
Table 2 contains results from studies in which the MedGem’s measurements were compared with conventional metabolic carts. The statistics we computed from the published data provide estimates of the MedGem’s bias and accuracy. To be consistent and make straightforward comparisons, data originally reported in kJ/day have been converted to mL O₂/min. Similarly, we have also computed the percent error for each of the studies listed. As the mean differences show, there was little if any bias between the MedGem’s measurements and those made by the reference methods. Except for the second study by Nieman et al.¹⁹ that included young children (lowest row), the bias was on the order of 1 mL O₂/min or less. In that case, the mean difference of 3.72 mL O₂/min is still only 1.78% of the mean; however, Reeves et al¹⁶ reported a bias of 10% between their reference instrument and the particular MedGem they tested. By contrast, St. Onge et al,¹⁵ like the studies shown in Table 2, found “no difference in resting metabolic rate between the two methods”. The standard deviations of the differences between the MedGem’s measurements and the reference methods indicate that, on average, the MedGem measures oxygen uptake to an accuracy of approximately ± 16.6 mL O₂/min or a percentage error ranging from 6.6 to 8.3%. Because inconsistencies arise from subtle differences in the design or execution of experiments, it is usually difficult to compare a new instrument with a reference method in a way that is rigorous and meaningful. Two of the cited studies illustrate this point. First, Melanson et al¹² initially found that the BodyGem’s measurements exceeded those of their reference instrument by a small but statistically significant amount; they later showed that the energy cost of using one hand to hold the BodyGem to the mouth was responsible for this apparent bias. By contrast, Stewart et al,¹⁴ who reported the smallest bias (0.58 mL O₂/min), employed a unique experimental design that avoided this problem. These investigators used a clamp to hold the MedGem in place so that no energy expenditure was necessary, and they placed the MedGem inside the flow-through hood of the DeltaTrac system. This approach also eliminated trial-to-trial inconsistencies and allowed the two instruments to analyze the same expired air. This experimental design in our opinion seems the soundest approach to assess bias and accuracy in vivo while avoiding irrelevant, extraneous influences. Therefore, when we computed the cardiac output error that results from random measurement errors in oxygen uptake measurements, we considered bias to be negligible and used the estimate of inaccuracy shown in Table 2 from Stewart et al.¹⁴

Patient selection, operating ranges and limitations. When we attempted to define the patient population and the operating conditions in which the MedGem could be used safely and effectively, we found discrepancies between the device’s technical specifications and published reports. For example, the specifications in the operator’s manual gave a range of tidal volumes from 500–1,500 mL/breath, yet Nieman et al¹⁹ and Fields et al³⁴ had successfully used the device on children as young as 7 years old. The 5-to-7 years age group would have resting tidal volumes in the 270–480 mL/breath range.³⁵ In fact, the smallest tidal volume reported by Nieman et al¹⁹ was 232 mL/breath. By interviewing the manufacturer’s staff, we found that the operating range is not limited by tidal volume per se. Instead, two criteria in the device’s software set the upper limit of permissible airflow. If the instantaneous airflow exceeds 2.1 L/min for a single breath or exceeds 1.5 L/min for 5 consecutive breaths, the device reports an error message rather than a measurement.
Our experiment on a goat sedated with ketamine and xylazine revealed two other limits set in software. When we used an endotracheal tube cemented into a mouthpiece to connect the device to the goat that was breathing shallowly and rapidly, the device transmitted the airflow and oxygen flux waveforms to a computer, but it failed to count each oscillation as a breath. After 30 seconds, the device reported an inappropriate error message to the effect that no breath had occurred in the last 30 seconds. Repeated attempts gave the same result. In an experiment on an anesthetized, 23 kg pig, we made the same observation. The manufacturer confirmed that two other limits set in software are a minimum peak airflow of 100 mL/sec and a minimum breathing frequency of 2 breaths/min. Table 3 lists the characteristics of the device that are pertinent to patient selection and operating conditions.
Because the hemodynamic profile of patients with cardiac disease can be essentially normal at rest, maneuvers such as supine bicycle exercise are sometimes used for diagnostic purposes during cardiac catheterization. As Table 3 shows, the range of oxygen consumption rates that the device reports is limited to 72–721 mL O₂/min. One of our subjects walked on a treadmill at gradually increasing speeds until it gave an error message and thus confirmed the maximum reportable oxygen uptake. Baim and Grossman’s cardiac catheterization handbook³⁶ shows oxygen consumption rates for two adult patients during supine bicycle exercise, and those rates are approximately in the 800–900 mL O₂/min range. Therefore, the device would not be useful during even mild exercise in the average-size adult. Although the device would not be useful in exercise, its span of reportable rates of oxygen uptake is sufficient to accommodate a wide spectrum of patients under resting conditions. For example, the Lindahl equation³⁷ predicts, as a rough approximation, that an 11-kg child consumes 75 mL O₂/min. Similarly, according to Bergstra’s formula,² a 6-foot, 20-year-old male weighing 110 kg consumes about 348 mL O₂/min at rest.
As Table 3 shows, we found that using the MedGem with the mouthpiece adds approximately 80 mL of extra dead space, whereas using it with the facemask adds between 61 and 152 mL. The amount of dead space added by the facemask could not be ascertained exactly because it depends partly on the volume of the mask not occupied by a particular patient’s face. If external dead space is imposed that equals or exceeds the patient’s resting tidal volume, the only ventilatory response that can effectively maintain alveolar ventilation is to increase tidal volume by the same amount. Adults can easily compensate for much greater amounts of external dead space than those in Table 3. For example, Khayat et al³⁸ found that adding 600 mL of external dead space evoked a compensatory increase in tidal volume, no change in breathing frequency, cardiac index, or stroke volume, and only a 4–5 mmHg increase in end-tidal PCO₂. An even smaller increase in end-tidal PCO₂ occurred with 200 mL of external dead space.
In pediatric applications, the dead space that the device adds to the patient’s physiological dead space determines whether it can be used safely. At this writing the device has been used in two studies of conscious children as young as 7 years old who had the mental ability to follow instructions and cooperate during the measurement.^19,34 It could also be used in younger children if they were sedated or anesthetized. However, it would be prudent not use the device on a child whose sustained ventilatory response to 80 or 160 mL of added dead space would be inadequate to maintain alveolar ventilation. Although the device could possibly be used on children younger than 7 years old, it is not clear whether the cut-off criterion at a minimum peak airflow of 100 mL/sec would preclude such use. Thus, more investigation and possibly modifications to the device will be necessary in order to further define the smallest patient on whom the device can be used safely and effectively.
Cardiac output error. Our calculations of cardiac output error are based on random instrument errors and do not include the well-known physiological sources of error in Fick cardiac output determinations such as the incomplete mixing of venous blood, changes in mean lung volume during the measurement of oxygen uptake, and oscillations in cardiac output and venous oxygen levels associated with the breathing cycle. The literature contains ample precautions for minimizing these sources of error,^27,44 but how much random instrument error is acceptable? In the older literature, there is an often-expressed conclusion that Fick cardiac output determinations are repeatable to within ± 10–15% or less^39–43 and that direct Fick and thermodilution determinations generally agree with each other to within ± 10–15% or less.⁴³ These reports are based on laborious but highly accurate methods such as the Douglas bag⁴⁵ and the apparatus of Van Slyke and Neill.²³ The review article by Warburton et al46 indicates that more recent reports also support this assertion. Therefore, we considered it highly desirable to achieve cumulative measurement errors of 10% or less since our calculations do not include the well-known physiological sources of cardiac output error.
As mentioned previously, the accuracy of cardiac output determinations based on the MedGem’s measurements of oxygen consumption cannot be assessed from the variability of its measurements alone, but must take into account the random measurement errors of the instruments used to measure blood oxygen. To use a representative example, the five-wavelength AVOXimeter 1000E is widely used in cardiac catheterization laboratories in the U.S. Without hemolyzing the blood samples, it measures %FO₂Hb to an accuracy of ± 1%,^47–49 and tHb to ± 0. 45 g/dL.^47,48 The conventional hemolyzing multi-wavelength co-oximeters are found less frequently in cardiac catheterization laboratories; their accuracy specifications are similar and have been summarized elsewhere.⁵²
Figure 1 shows the error in cardiac output that results from measuring oxygen consumption with the MedGem and using the AVOXimeter or one of the other multi-wavelength co-oximeters with the same accuracy. In this figure and those that follow, cardiac output error, expressed as a percentage, is plotted as a function of the arteriovenous difference between the oxyhemoglobin fractions, also expressed as a percentage. The vertical line intersects the curves at typical resting values: cardiac output of 5 L/min and a – v %FO₂Hb of 25%. The upper curve in Figure 1 was generated from Equation 15 (Appendix), and it shows the cardiac output error that results from implementing the Fick Principle in the form of Equation 6, i.e., using the oxygen concentrations that the oximeter reports for each blood sample. At a resting cardiac output, this practice results in an error of 17.2%. By contrast, the lower curve was generated from Equation 18, and it shows a cardiac output error of 9.5% results from using the same instruments but substituting the oximeter’s tHb and FO₂Hb readings into Equation 7. Figure 1 shows clearly that implementing the Fick Principle in the form of Equation 7 is the preferable approach when oxygen content is deduced from oximetric measurements.
The canine measurements are shown in Figure 2, and they reinforce the conclusion that implementing the Fick Principle in the form of Equation 7 is preferable to using the oxygen content that the oximeter computes for each sample, as illustrated by Equation 6. Our statistical model can be used to calculate either the inaccuracy or imprecision of cardiac output determinations. Therefore, we compared the observed imprecision of 6 cardiac output determinations with the imprecision predicted by the statistical model. Using the oxygen content that the oximeter computes on a sample-by-sample basis gave a mean cardiac output of 4,303 ± 1,195 mL/min and hence an imprecision of 27.8%. By contrast, using the fractional oxyhemoglobin readings and an average tHb in Equation 7 gave a mean cardiac output of 4,771 ± 611 mL/min and an imprecision of 12.8%. These data points fall just above the curves that were generated by inserting the observed imprecision of the individual measurements into the statistical model.
Figure 3 shows the effect of oximeter inaccuracy on the error in cardiac output. At one time over 1,400 of the two-wavelength Oxicom oximeters (Waters Instruments, Rochester, Minnesota) were used in cardiac catheterization laboratories, and many are still in use. Based on the assumption that the concentrations of carboxy- and methemoglobin are negligible, it measures % saturation. Its accuracy is reported to be ± 2.70%^50,52 or ± 2.65%.⁵¹ As Figure 3 shows, at a resting cardiac output, a 17.1% error occurs if the Oxicom’s %Sat readings in arterial and mixed venous blood are used in combination with MedGem’s measurement of oxygen uptake. The lower curve is the same as Figure 1; it shows a cardiac output error of 9.5% when the %FO₂Hb accuracy is ± 1% rather than ± 2.7%. For this comparison the same tHb inaccuracy (± 0.45 g Hb/dL) was used in both cases.
The curves in the previous figures were generated for the instance in which each variable in Equation 7 is measured only once. Figures 4 and 5 illustrate the benefits of using average values instead of single measurements. The lower curve in Figure 4 shows the effect of using the average of 2 oxygen uptake measurements and measuring the other variables once. At a resting cardiac output of 5 L/min, measuring oxygen uptake twice and using the average lowers the cardiac output error from 9.5% to 8.1%, and, as the graph shows, the benefit of averaging 2 oxygen uptake measurements increases at higher a–v oxygen differences. Although this approach to minimizing the error in cardiac output is effective, we did not think that busy clinicians were likely to consider it practical because each oxygen uptake measurement takes nearly 10 minutes.
Experimenting with our statistical model makes it possible to devise other, more practical, strategies to minimize the cumulative error in cardiac output. For example, the multi-wavelength oximeters and co-oximeters measure both %FO₂Hb and tHb in each blood sample, and tHb should be the same in arterial and venous blood. Therefore, measuring each blood sample twice yields 2 measurements of FO₂Hba, 2 of FO₂Hbv, and 4 of tHb. In the case of the AVOXimeter, each analysis takes less than 10 seconds and requires only 50 µL of blood. Thus, without contributing to iatrogenic blood loss and adding only 20 seconds to the procedure, this strategy yields a total error in cardiac output that is nearly the same as that obtained by averaging 2 of the MedGem’s 10-minute measurements of oxygen uptake. As Figure 5 shows, the two error rates are 8.25% and 8.1% at a typical resting cardiac output. Furthermore, averaging the blood measurements rather than repeating the oxygen uptake measurement yields smaller errors at lower arteriovenous oxygen differences.

Summary

The MedGem has some definite limitations when compared with conventional metabolic carts. There is no means for attaching it to the patient, it cannot be used with inspired oxygen concentrations different from room air, it cannot be used with ventilators, its use with endotracheal tubes requires prior improvisation, its deadspace precludes its use in infants, and there is no easy way to verify its calibration. Furthermore, adults performing even the mildest exercise are likely to exceed its measurement range. However, as a means for performing Fick cardiac output determinations at rest, the MedGem has significant advantages over conventional instruments. Metabolic carts typically cost tens of thousands of dollars, take up valuable floor space, and require a highly trained operator, whereas the MedGem’s price in 2006 was $2,500, and we found it extremely simple, reliable, and easy to operate. Furthermore, if the error-reduction strategies are employed that we presented here, the cumulative effect of random measurement errors on the calculated cardiac output can be kept under 10%. Therefore, we hope that the hazardous practice of using an assumed or estimated rate of oxygen consumption to compute cardiac output will be abandoned in favor of measured values.

Acknowledgments. We would like to thank Heather A. Haugen, Ed Pearce and Michael Clouthier of HealtheTech for the loan of two MedGems, a generous supply of disposables and an even more generous amount of advice and encouragement. We are grateful to James Elliott, DVM, for supervising and assisting in the animal experiments.

References

1. LaFarge CG, Miettinen OS. The estimation of oxygen consumption. Cardiovasc Res 1970;4:23–30.
2. Bergstra A, Van Dijk RB, Hillege HL, et al. Assumed oxygen consumption based on calculation from dye dilution cardiac output: An improved formula. Eur Heart J 1995;16:698–703.
3. Crocker RH Jr, Ockene IS, Alpert JS, et al. Determinants of total body oxygen consumption in adults undergoing cardiac catheterization. Cathet Cardiovasc Diagn 1982;8:363–372.
4. Saksena FB. Nomogram to calculate oxygen consumption index based on age, sex and heart rate. Pediatr Cardiol 1983;4:55–56.
5. Fakler U, Pauli C, Henning M, et al. Assumed oxygen consumption frequently results in large errors in the determination of cardiac output. J Thorac Cardiovasc Surg 2005;130:272–276.
6. Kendrick AH, West J, Papouchado M, Rozkovec A. Direct Fick cardiac output: Are assumed values of oxygen consumption acceptable? Eur Heart J 1988;9:337–342.
7. Wolf A, Pollman MJ, Trindade PT, et al. Use of assumed versus measured oxygen consumption for the determination of cardiac output using the Fick principle. Cathet Cardiovasc Diagn 1998;43:372–380.
8. Laitinen PO, Räsänen J. Measured versus predicted oxygen consumption in children with congenital heart disease. Heart 1998;80:601–605.
9. Li J, Bush A, Schulze-Neick I, et al. Measured versus estimated oxygen consumption in ventilated patients with congenital heart disease: The validity of predictive equations. Crit Care Med 2003;31:1235–1240.
10. Shanahan CL, Wilson NJ, Gentles TL, Skinner JR. The infuence of measured versus assumed uptake of oxygen in assessing pulmonary vascular resistance in patients with a bidirectional Glenn anastomosis. Cardiol Young 2003;13:137–142.
11. Dehmer GJ, Firth BG, Hillis LD, Oxygen consumption in adult patients during cardiac catheterization. Clin Cardiol 1982;5:436–440.
12. Melanson EL, Coelho LB, Tran ZV, et al. Validation of the BodyGem hand-held calorimeter. Int J Obes Relat Metab Disord 2004;28:1479–1484.
13. Storer TW, Kamps K, Dabari B, et al. Validation of the MedGem device for measurement of resting metabolic rate. Med Sci Sports Exerc 2004;36(Suppl May):S247.
14. Stewart CL, Goody CM, Branson R. Comparison of two systems of measuring energy expenditure. J Parenter Enteral Nutr 2005;29:212–217.
15. St.Onge MP, Rubiano F, Jones A Jr, Heymsfield SB. A new hand-held indirect calorimeter to measure postprandial energy expenditure. Obes Res 2004;12:704–709.
16. Reeves MM, Capra S, Bauer J, et al. Clinical accuracy of the MedGem indirect calorimeter for measuring resting energy expenditure in cancer patients. Eur J Clin Nutr 2005;59:603–610.
17. Alam DS, Hulshof PJM, Roordink D, et al. Validity and reproducibility of resting metabolic rate measurements in rural Bangladeshi women: Comparison of measurements obtained by MedGem and by DeltaTrac device. Eur J Clin Nutr 2005;59:651–657.
18. Nieman DC, Trone GA, Austin MD. A new handheld device for measuring resting metabolic rate and oxygen consumption. J Am Diet Assoc 2003;103:588–593.
19. Nieman DC, Austin MD, Chilcote SM, Benezra L. Validaton of a new handheld device for measuring resting metabolic rate and oxygen consumption in children. Int J Sport Nutr 2005;14:208–216.
20. Compher C, Hise M, Sternberg A, Kinosian BP. Comparison between MedGem and DeltaTrac resting metabolic rate measurements. Eur J Clin Nutr 2005;59:1136–1141.
21. Mahutte CK, Jaffe MB. Effect of measurement errors on cardiac output calculated with O2 and modified CO2 Fick methods. J Clin Monitor 1995;11:99–108.
22. Critchley LAH, Critchley AJH. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurements techniques. J Clin Monitor Comp 1999;15:85–91.
23. Van Slyke DD, Neill JM. The determination of gases in blood and other solutions by vacuum extraction and manometric measurements. J Biol Chem 1924;61:523–557.
24. Gregory IC. Assessment of Van Slyke manometric measurements of oxygen content. J Appl Physiol 1973;34:715–717.
25. Kusumi F, Butts WC, Ruff WL. Superior analytical performance by electrolytic cell analysis of blood oxygen content. J Appl Physiol 1973;35:299–300.
26. Selman BJ, White YS, Tait AR. An evaluation of the Lex-O2-Con oxygen content analyser. Anaesthesia 1975;30:206–211.
27. 27. Guyton AC, Jones CE, Coleman TG. Circulatory Physiology: Cardiac Output and Its Regulation. 2nd Edition. Philadelphia: W. B. Saunders Company, 1973.
28. National Committee for Clinical Laboratory Standards (US). Fractional oxyhemoglobin, oxygen content and saturation, and related quantities in blood: Terminology, measurement, and reporting. Villanova (PA): Document C25-T, 1992, Vol. 12 (No.11).
29. Johnson NL, Tetley H. Statistics, an Intermediate Text Book. 2nd Edition. Cambridge (UK): Cambridge University Press, 1962.
30. Barnett HAR. The variance of the product of two independent variables and its application to an investigation based on sample data. Journal of the Institute of Actuaries 1955;81:190.
31. Storer T, Kearney JT, Alexander HA. Comparison of the BodyGem to a Mechanical Simulation Device. Healthtech, Inc.: Golden, Colorado; 2001.
32. Takala J, Keinänen O, Väisänin P, Kari A. Measurement of gas exchange in intensive care: Laboratory and clinical validation of a new device. Crit Care Med 1989;17:1041–1047.
33. Makita K, Nunn JF, Royston B. Evaluation of metabolic measuring instruments for use in critically ill patients. Crit Care Med 1990;18:638–644.
34. Fields DA, Kearney JT, Copeland, KC. MedGem hand-held indirect calorimeter is valid for resting energy expenditure measurement in healthy children. Obesity 2006;14:1755–1761.
35. Miller RD (Ed). Anesthesia. 5th Edition. Philadelphia: Churchill Livingstone; 2000.
36. Lorell BH, Grossman W. Dynamic and isometric exercise during cardiac catheterizadion. In: Baim DS, Grossman W (Eds). Cardiac Catheterization, Angiography, and Intervention. 5th Edition. Baltimore: Williams & Wilkins. 1996, pp. 281–296.
37. Lindahl SG. Oxygen consumption and carbon dioxide elimination in infants and children during anaesthesia and surgery. Br J Anaesth 1989;62:70–76.
38. Khayat RN, Xie A, Patel AK, et al. Cardiorespiratory effects of added dead space in patients with heart failure and central sleep apnea. Chest 2003;123:1551–1560.
39. Selzer A, Sudrann RB. Reliability of the determination of cardiac output in man by means of the Fick principle. Circ Res 1958;6:485–490.
40. Thomasson B. Cardiac output in normal subjects under standard basal conditions. Scand J Clin Lab Invest 1957;9:365–376.
41. Reddy PS, Curtiss EI, Bell B, et al. Determinants of variation between Fick and indicator dilution estimates of cardiac output during diagnostic catheterization. Fick vs. dye cardiac outputs. J Lab Clin Med 1976;87:568–576.
42. Hillis LD, Firth BG, Winniford MD. Analysis of factors affecting the variability of Fick versus indicator dilution measurements of cardiac output. Am J Cardiol 1985;56:764–768.
43. Enghoff E, Michaëlsson M, Pavek K, Sjögren S. A comparison between the thermal dilution method and the direct Fick and the dye dilution methods for cardiac output measurements in man. Acta Soc Med Ups 1970;75:157–170.
44. Grossman W. Blood flow measurement: The cardiac output. In: Baim DS, Grossman W. Cardiac Catheterization, Angiography, and Intervention. 5th Edition. Baltimore: Williams & Wilkins. 1996, pp. 109–124.
45. Douglas CT. A method for determining the total respiratory exchange in man. J Physiol 1911;42:17–18.
46. Warburton DER, Haykowsky MJF, Quinney HA, et al. Reliability and validity of measures of cardiac output during incremental to maximal aerobic exercise. Part 1: conventional techniques. Sports Med 1999;27:23–41.
47. Bailey SR, Russell EL, Martinez A, Evaluation of the AVOXimeter: Precision, long-term stability, linearity, and use without heparin. J Clin Monit 1997;13:191–198.
48. Gong, AK. Near-patient measurement of methemoglobin, oxygen saturation, and total hemoglobin: Evaluation of a new instrument for adult and neonatal intensive care. Crit Care Med 1995;23:193–201.
49. Freeman GL, Steinke JM. Evaluation of two oximeters for use in cardiac catheterization laboratories. Cathet Cardiovasc Diagn 1993;30:51–57.
50. Hedlund KD, Sanford DM. Clinical experience with the Oxicom 3000 whole blood oximeter. Am Acad Cardiovasc Perfusion 1993;14:57–61.
51. Russell EL. Correct and spurious calibration of the Oxicom oximeter. Cathet Cardiovasc Diagn 1994;33:264–266.
52. Shepherd AP, McMahan CA. Role of oximeter error in the diagnosis of shunts. Cathet Cardiovasc Diagn 1996;37:435–446.