Delay time adjustments to minimize errors in breath-by-breath measurement of V O2 during exercise

Transcription

1 Delay time adjustments to minimize errors in breath-by-breath measurement of V O2 during exercise DAVID N. PROCTOR AND KENNETH C. BECK Department of Anesthesia and Division of Pulmonary and Critical Care Medicine, Mayo Clinic and Foundation, Rochester, Minnesota Proctor, David N., and Kenneth C. Beck. Delay time adjustments to minimize errors in breath-by-breath measurement of V O2 during exercise. J. Appl. Physiol. 81(6): , If the delay time between gas concentration and flow signals is not adequately corrected during breath-bybreath analysis of expired gas, an error in calculation of oxygen consumption (V O2 ) will result. To examine the frequency and delay time dependences of errors in V O2 measurement, six healthy men exercised at 100, 200, and 250 W on a cycle ergometer while breath-by-breath assessment of V O2 was made simultaneously with collection of expired air. Subjects breathed first at normal rates (15 30 breaths/min) and then at 70 breaths/min. Each subject performed each level of exercise twice by using erroneous values for the delay time between gas concentration and flow signals. At normal breathing frequencies, errors in V O2 measurement were 610% over the full range of delay times used, and the errors were not tightly correlated with variations in delay times from optimum. However, at 70 breaths/min, errors approached 630% as the variations in delay times deviated 60.1 s from the optimal, and the errors were highly correlated with the variations in delay times. We conclude that there is greater potential for errors in V O2 measurement with incorrect delay time at higher breathing frequencies. These findings suggest that the optimal delay time for breath-bybreath systems should be adjusted by using high breathing frequencies. human performance; indirect calorimetry; hyperpnea; gas exchange; equipment validation; oxygen consumption OXYGEN CONSUMPTION (V O2 ) and carbon dioxide production (V CO2 ) are commonly used to characterize the response to exercise in both the research and clinical settings (13). Originally, the measurements were made by using timed collections of expired gas (8). Recently, rapid gas analyzers and computer technology have made breath-by-breath analysis of gas exchange practical and commonplace by using commercially available equipment (5). Although these systems offer the advantage of being able to present the data in real time, there are some technical errors that can make breath-bybreath measurements inaccurate (13). One of these errors is the delay time between the expired gas and expiratory flow signals (6, 9, 13). The gas concentration signal is delayed by, s compared with the flow signal in most commercially available systems. This is due to delays in transport of gas in the sampling tubing and delays introduced by electronic signal processing (2). This delay is handled in the computer by shifting the signals representing gas concentrations by an appropriate number of sample periods to properly align the gas concentration and flow signals before performing the integrations necessary to determine V O2 and V CO2. The delay time is usually determined during calibration by introducing a rapid change in gas concentration and measuring the response time of the gas analyzer. Errors can occur in this calibration procedure. For instance, a small part of the total delay time consists of valve-actuation times that can only be determined empirically and that can change due to component failure. If the delay time of the gas concentration signal is underestimated by the computer, V O2 and V CO2 will be underestimated, whereas if the shift is too large, V O2 and V CO2 will be overestimated (5, 9, 13). Because a constant time shift represents a larger fraction of the total breathing cycle of shorter breaths, the error will be magnified at higher breathing frequencies (4, 6, 13). Thus it is important that breath-bybreath systems be validated over an acceptable frequency range. To determine the frequency range of interest in the clinical setting, we retrospectively reviewed the exercise responses of 113 adult patients referred to our laboratory (Fig. 1). At rest, breathing rates ranged from,5 to just over 30 breaths/min. At maximum exercise, the breathing rates were considerably higher, ranging from 20 to.60 breaths/min. About 14% of the patients breathed at frequencies.55 breaths/min near maximal exercise. Because of the emphasis often placed on V O2 measurements made near maximal exercise (8, 11), these data emphasize the need to ensure proper equipment performance up to,60 or even 70 breaths/min. In this paper, we examine the errors in breath-bybreath measurements of V O2 during exercise by systematically varying the gas-sampling delay times and breathing frequency. Because of the frequency dependence of the error, validations performed at relatively low breathing rates may be within acceptable limits, whereas validations at higher but still physiologically reasonable breathing rates may show significant errors. METHODS Subjects. Six healthy men (age yr) who were laboratory personnel served as subjects. Each was recreationally active, but no subject was a competitive athlete. The subjects were familiar with gas-exchange measurements during cycle ergometry, and they provided verbal informed consent according to guidelines of the Mayo Clinic Institutional Review Board. Equipment. Exercise was performed with the subjects in the upright posture on an electrically braked cycle ergometer (Lode Excalibur Sport). Breath-by-breath measurements of exercise V O2,V CO2, and expired minute ventilation (V E) were made by using a commercially available automated system (Medical Graphics, model CPX/D, St. Paul, MN) modified to interface to a respiratory mass spectrometer (Perkin Elmer /96 $5.00 Copyright r 1996 the American Physiological Society 2495

2 2496 ERRORS IN BREATH-BY-BREATH V O2 Fig. 1. Histogram showing no. of patients with observed breathing frequencies measured at rest and near maximal (Max) exercise during incremental cycle ergometer-testing protocol in a clinical exercise- testing laboratory. Note that 14% of patients breathed at.55 breaths/min at peak exercise (vertical dotted line) and that there were a few patients in this limited retrospective review who breathed.70 breaths/min. These data emphasize need for adequate equipment validation over a wide range of breathing frequencies. 1100). Subjects breathed through a disposable pneumotachograph that contained the mass spectrometer gas-sampling port. The pneumotachograph registered flow by comparing impact and stagnation pressures in a region of slight narrowing of the flowing gas stream. Linearity of the flowmeter system was ensured by the manufacturer, who applied multiple correction factors to adjust for nonlinearity (12). The sampling rate of the mass spectrometer was 60 ml/min, and the sampling line was 200 cm long. The pneumotachograph was connected to a Hans-Rudolph non-rebreathing valve (80 ml total dead space volume) so that expired air could be periodically collected in meteorologic balloons (latex rubber) and subsequently analyzed for expired O 2 and CO 2 concentrations (mass spectrometer) and volumes (Tissot gasometer). Calculations of V O2 and V CO2 were performed by using the Haldane transformation (8). This permitted the comparison of breath-by-breath and bag (Bag) determinations of V O2 and V E. The breath-by-breath measurement system was calibrated by using the system s standard computer programs and precision-grade gas mixtures. The pneumotachograph was calibrated by using a 3-liter calibration syringe, again following procedures dictated by standard calibration routines. Multiple syringe strokes were performed over a range of flows to check the linearity of the pneumotachograph response. Calibrations were performed with the pneumotachograph attached to the non-rebreathing valve, exactly as configured during data collection. Protocol. Subjects completed separate bouts of light (100 W, n 5 6), moderate (200 W, n 5 6), and heavy (250 W, n 5 4) exercise at self-selected pedaling rates (55 75 revolutions/ min) while breath-by-breath gas-exchange measurements were continuously monitored. To examine the effect of altering the delay time between gas concentration and flow signals on breath-by-breath measurements, two bouts of exercise were performed at each intensity, each with a randomly chosen delay time varying between 0.11 and 0.25 s. Two bouts with different delay times at each of three power outputs gave six bouts total for each subject. The delay time of the computerized system was adjusted during a 10- to 15-min rest between bouts of exercise by manual change of the internal correction factors and recalibration of the system, as outlined in the APPENDIX. Each bout of exercise lasted 6 min and was preceded by 2 min of warm-up at 50 or 100 W. Subjects breathed normally during the first 4 min. A 1-min Bag collection was made during the fourth minute. To examine the effect of high breathing rates on gas-exchange measurements, subjects performed the final 2 min of each bout while breathing at 70 breaths/min. A 1-min Bag collection was made during the second minute of the high breathing rate (i.e., sixth and final minute of the bout). The subjects had all attained a near steady state in the minute of data collection, as judged from the breath-by-breath display of V O2 and V CO2 over time. A metronome was used to pace the subjects breathing, and they were coached to reduce their tidal volume to avoid lightheadedness associated with hypocapnia. The two Bag collections (minutes 3 4 and 5 6) were compared to assess the oxygen cost of the hyperpnea. Data analysis. The breath-by-breath V O2 and V E were calculated over the exact period of Bag collection. The mean V O2 for the interval was computed by first summing the VO 2 contributions from each breath and dividing by the sum of breath times for the interval. Percent errors in V O2 measurement were calculated as (breath-by-breath 2 Bag)/Bag 3 100, where breath-by-breath and Bag data were obtained over the same sampling period. The optimal delay time was determined from the intercept of %errors in V O2 measurement regressed against delay time. Analysis was performed on data obtained at normal breathing rates and during hyperpnea (70 breaths/min). RESULTS The average V E and V O2 responses obtained from the 1-min Bag collections are shown in Table 1. At all three power outputs, V E was 42 90% higher while subjects were breathing at 70 breaths/min than during normal breathing rates. Bag-derived V O2 values did not differ significantly between normal and high breathing rates while subjects were cycling at any of the exercise levels, indicating low work of breathing in these normal subjects. The increases in V O2 with increasing exercise Table 1. Oxygen uptake and minute ventilation responses measured by using timed bag collections n Normal- Frequency High- Frequency 100 W 6 V O2, ml/min 1, , (NS) V E, l/min Frequency, breaths/min W 6 V O2, ml/min 2, , (NS) V E, l/min Frequency, breaths/min W 4 V O2, ml/min 3, , (NS) V E, l/min * Frequency, breaths/min Values are means 6 SD; n no. of subjects. No. of observations were 12, 12, and 8 for 100, 200, and 250 W, respectively. V O2, O 2 consumption; V E, minute ventilation; NS, not significant. Highfrequency-breathing data significantly different from low-frequencybreathing data (paired t-test): *P, 0.05; P, 0.01.

3 ERRORS IN BREATH-BY-BREATH V O intensity were very close to values predicted for cycle ergometry, according to equations published by the American College of Sports Medicine (1). During the heavy work intensity (250 W), V O2 responses of all four subjects studied were higher during hyperpnea (70 breaths/min) than during normal breathing, though the difference failed to reach significance. Figure 2 shows the induced errors in breath-bybreath determinations of V O2 expressed as the %error compared with the corresponding Bag values. When subjects breathed at normal frequencies (A), induced errors in breath-by-breath determinations of V O2 were 610% across the range of delay times used (i.e., s). By contrast, during high-frequency breathing (B; 70 breaths/min), breath-by-breath measurement errors induced by variations in delay times ranged from 230 to 30%. The corresponding mean errors in V E (data not shown) ranged from 0 to 22.7%. The V E errors were only statistically significantly different from zero (P, 0.05) during the high-frequency breathing at the 200- and 250-W work intensities. However, the errors in V E measurement were small in comparison with the V O2 measurement errors and, as expected, were not correlated with the delay times (P. 0.05). As shown in Fig. 2 and Table 2, at high breathing rates, errors in V O2 measurement were much more Fig. 2. Percentage of error in O 2 consumption measurements plotted as function of delay time for 100, 200, and 250 W. Thick lines, linear regression; thin lines, 95% confidence interval for %error estimates from regression equation. Note steeper slope and narrower confidence limits of this relationship at high-frequency breathing (70 breaths/min; B) than at normal breathing frequencies (A). Results from regression analysis are presented in Table 2. Table 2. Regression analysis of %errors in V O2 vs. delay time tightly correlated with delay times compared with low breathing rates. The correlation coefficients between delay times and %error in V O2 measurement were lower at normal breathing frequencies compared with hyperpnea. The lower correlation coefficients resulted in wider confidence intervals for the delay times intercept at 0% error. In fact, at 100 and 200 W, the delay time intercept was virtually indeterminate from our data. Average delay time intercept obtained from data for high breathing rates was 0.18 s, which is shorter than the s reported initially by the equipment calibration routines. The use of s gave small errors in V O2 measurement at normal breathing rates and 10 15% errors at 70 breaths/min (Fig. 2). The 0.18 s is also shorter than the delay time for the mass spectrometer alone (Fig. 3 as discussed in the APPEN- DIX). DISCUSSION Normal-Frequency High-Frequency 100 W Slope, %/s Error intercept, % Time intercept, s R (NS) 0.92* 200 W Slope, %/s Error intercept, % Time intercept, s R (NS) 0.91* 250 W Slope, %/s Error intercept, % Time intercept, s R * 0.94* Regression coefficients for %error in V O2 5 error intercept 1 slope x (delay time). Time intercept was computed from error intercept/slope. *Significantly different R 2 from 0, *P, The important findings of this study are that errors in V O2 measurement by using a breath-by-breath gas exchange measurement system can be 30% at high, but still physiological, breathing frequencies when errors are close to an acceptable range at lower breathing frequencies. Although errors as high as 30% are unlikely when equipment is calibrated properly, this study emphasizes the magnification of error at high breathing rates. It should be noted that the manufacturer of the CPX/D system we used (Medical Graphics) routinely performs validation testing at high respiratory rates when in development and on each unit before it is shipped. Our system required revalidation because we used a mass spectrometer in place of supplied gas analyzers. These errors were not due to errors in measurement of V E or errors in the gain of the gas analyzers but were due to induced errors in delay times used to align flow and gas concentration signals. These results stress the importance of using high breathing

4 2498 ERRORS IN BREATH-BY-BREATH V O2 Fig. 3. Pop test to determine response of O 2 and CO 2 channels of mass spectrometer (see text for details). Average delay times in 2 trials were and s for CO 2 and O 2, respectively, and average 10 90% rise times were and s for CO 2 and O 2, respectively. frequencies during initial validation of automated breath-by-breath systems used to measure V O2. The effect of delay time on measurements of V O2 and V CO2 by using breath-by-breath methods has been discussed by others (2, 4, 6, 7, 9). Although Beaver et al. (4) made the point that the errors would be magnified at a higher breathing frequency, no studies have systematically examined the effects of breathing frequency on the error in measurement of V O2. Huszczuk et al. (7) demonstrated lack of frequency dependence of V O2 in a well-tuned system by using a gas-exchange simulator. We found, as predicted (4), that the errors in V O2 measurement were magnified by subjects breathing at high frequencies and that these errors were not due to mismeasurement of V E. The optimal delay time is the time shift that produces zero error in measurement of V O2. At normal breathing rates, the width of the statistical 95% confidence limits for delay time at zero error was wide, being essentially indeterminate at rest and,60 ms at 250 W. Hughson et al. (6) showed a similar confidence limit in one subject exercising at 200 W. However, their technique of recalculating the same data by using varying delay times should have produced somewhat less scatter in the data than our technique of making multiple measurements in different subjects. The slope of the regression of %error against delay time ranged from 39 %/s at 100 W to 124 %/s at 250 W. Our slope at 200 W, expressed in ml min 21 ms21 [0.68 s 21 3 (2,578 ml/min)/ (1,000 ms/s) ml min 21 ms21 ], compares favorably with a slope that can be derived from data presented by Noguchi et al. (9) of 1.4 ml min 21 ms21 on the basis of measurements in three subjects. In our study, the slope of %error against delay time increased and the confidence limits narrowed considerably at higher breathing rates. The delay time intercepts were similar over a wide range of exercise intensities. Because optimal delay times that minimized the errors were similar across exercise intensities, adjustment of delay times at low-intensity exercise may be safely extrapolated to higher intensities. These results emphasize the importance of initial validation of gas-exchange measurement systems by using high breathing frequencies. An alternative to Bag collection of human expired air is use of a mechanical gas-exchange simulator (7). High breathing frequencies are important during validation because errors observed during normal breathing frequencies were not as strongly correlated with delay time and were close to acceptable limits over a range of delay times. It should be stressed that these errors will be produced by any breath-by-breath measurement system when delay times used are in error and are not a deficiency in the design of commercial products. Both initial validation and periodic monitoring of measurement errors are recommended because there are a number of factors that can cause delay times to change, such as buildup of debris or moisture in gas-sampling lines, variations in the gas-sampling rates through the analyzer, or changes in length of the sampling tubing. It should be noted that stable daily calibrations (including a procedure to determine delay time) should be sufficient to ensure accuracy once equipment is initially validated. However, when equipment is serviced, repeat validation is useful to ensure adequate system performance. In summary, we found that errors in V O2 measurement due to varying the delay time factor by using a breath-by-breath gas-exchange system reached 30% at high, but still physiological, breathing frequencies when the induced errors were close to acceptable range at lower breathing frequencies. These findings emphasize the need for adequate initial and periodic validations of breath-by-breath gas-exchange measurement systems covering the full range of breathing frequencies likely to be encountered in practice. APPENDIX To perform these studies, we used CPX-D system from Medical Graphics (St. Paul, MN) by using calibration software version 3.8 and breath-by-breath data-acquisition software version 3.3. On consultation with engineers from Medical Graphics, we edited the CAL.CFG file, changing the numbers at the end of lines 202 and 203 to vary the delay times reported. It should be noted that Medical Graphics performs validation studies at high respiratory rates both in development and before each unit is shipped. We repeated validation because we used their system with a mass spectrometer in place of standard gas analyzers. The delay time reported by the calibration routine of the commercial software represents a sum of many components. To see how the reported delay time corresponded with the delay time of the mass spectrometer by itself, we performed the following pop test. We arranged an electrical circuit to produce a voltage when the sampling line of the mass spectrometer was sampling calibration gas. A thin wire that was part of the circuit was arranged so that it would break at the precise moment the tip of the sampling line from the mass spectrometer exited flowing calibration gas. In this way, we could rapidly pop the sampling line out of the flowing gas to

5 ERRORS IN BREATH-BY-BREATH V O sample room air, obtaining an electrical event signal when the sampling line began sampling room air. Results from this test are shown in Fig. 3. The delay in the gas concentration signal measured from the point the sampling line exited the calibration gas to the point where the O 2 signal had increased by 10% was 240 ms. This time is,50 60 ms longer than the delay time reported by the commercial software. This difference in delay time represents the time needed to process the flow signal and actuation times for small switching valves that are used in the determination of delay time. The manufacturer ordinarily includes a correction for these signalprocessing delays. This work was supported by the Mayo Foundation, National Institute of General Medical Sciences Grant GM-08288, and a General Clinical Research Center grant (Division of Research Resources Grant M01-RR00585). Address for reprint requests: K. C. Beck, Pulmonary Function Lab, S-3 Plummer Bldg, Mayo Clinic, Rochester, MN Received 7 July 1995; accepted in final form 22 July REFERENCES 1. American College of Sports Medicine. Guidelines for Exercise Testing and Prescription (4th ed.). Philadelphia, PA: Lea & Febiger, Arieli, R., and H. D. Van Liew. Corrections for the response time and delay of mass spectrometers. J. Appl. Physiol. 51: , Astrand, P., and K. Rodahl. Evaluation of physical work capacity on the basis of tests. In: Textbook of Work Physiology. New York: McGraw-Hill, 1977, chapt. 10, p Beaver, W. L., N. LaMarra, and K. Wasserman. Breath-bybreath measurement of true alveolar gas exchange. J. Appl. Physiol. 51: , Beaver, W. L., K. Wasserman, and B. J. Whipp. On-line computer analysis and breath-by-breath graphical display of exercise function tests. J. Appl. Physiol. 34: , Hughson, R. L., D. R. Northey, H. C. Xing, B. H. Dietrich, and J. E. Cochrane. Alignment of ventilation and gas fraction for breath-by-breath respiratory gas exchange calculations in exercise. Comput. Biomed. Res. 24: , Huszczuk, A., B. J. Whipp, and K. Wasserman. A respiratory gas exchange simulator for routine calibration in metabolic studies. Eur. Respir. J. 3: , Jones, N. L. Clinical Exercise Testing (3rd ed.). Philadelphia, PA: Saunders, Noguchi, H., Y. Ogushi, I. Yoshiya, N. Itakura, and H. Yamabayashi. Breath-by-breath V CO2 and V O2 require compensation for transport delay and dynamic response. J. Appl. Physiol. 52: 79 84, Roston, W. L., B. J. Whipp, J. A. Davis, D. A. Cunningham, R. M. Effros, and K. Wasserman. Oxygen uptake kinetics and lactate concentration during exercise in humans. Am. Rev. Respir. Dis. 135: , Wasserman, K., J. E. Hansen, D. Y. Sue, and B. J. Whipp. Principles of Exercise Testing and Interpretation. Philadelphia, PA: Lea & Febiger, Yeh, M. P., R. M. Gardner, T. D. Adams, F. G. Yanowitz, and R. O. Crapo. Anaerobic threshold : problems of determination, and validation. J. Appl. Physiol. 55: , Zeballos, R. J., and I. M. Weisman. Behind the scenes of cardiopulmonary exercise testing. Clin. Chest Med. 15: , 1994.