Twenty-three healthy subjects (24–43 yr old, 2 females) gave informed written consent to participate in the studies, which were approved by the hospital ethics committee.
Measurements of Cardiac Output
As described previously (2,17), the transducer was placed around the chest at or within 4 cm below the xiphoid process, secured with an adhesive tape, and connected to a digital inductance plethysmograph (Noninvasive Monitoring Systems, Miami Beach, FL). The inductance signal and the ECG were sampled at 200 Hz with 16-bit resolution and processed by a computer using custom software (7). Inductance waveforms were digitally bandpass filtered (passband 0.7 times the actual heart rate to 10 Hz (2,7)) and ensemble averaged with the ECG R-wave as trigger. This signal processing technique provided mean ventricular volume curves (Fig. 1) and automatically detected the peak to trough amplitude as a relative measure of stroke volume. This allowed tracking of changes in stroke volume and cardiac output (i.e., the product of stroke volume and heart rate) in relative units (percent) without the need for calibration.
To obtain estimates of cardiac output by inductance cardiography in absolute units (L·min−1), one paired measurement by an alveolar gas exchange method was required to calibrate inductance cardiography. Subsequent estimates of cardiac output in the same individual could then be measured by inductance cardiography in liters per minute.
For comparisons to the carbon dioxide rebreathing technique, 100–120 heart cycles occurring over 60 s immediately before the rebreathing maneuver were included into the ensemble averages. For comparisons to the oxygen uptake method, mean ventricular volume curves were computed by selecting ensemble averaging periods of three different lengths: 200, 100, and 50 heart cycles, respectively. This provided the opportunity to assess the effect of the number of heart cycles that were included into ensemble averages on precision of inductance cardiography.
Carbon dioxide rebreathing.
Cardiac output was determined based on the carbon dioxide (CO2) Fick equation by dividing exhaled CO2 by the arteriovenous CO2 content difference (8,12,15). Arterial pCO2 was estimated from end-tidal pCO2 in the breath immediately before rebreathing. Mixed venous pCO2 was obtained by equilibrium CO2 rebreathing (9) using a mixture of CO2 at a concentration of 15% and a volume of 1.5 times the tidal volume. The rebreathing period was initiated at the end of expiration and continued for 20 s. The criteria for equilibration were satisfied when pCO2 did not vary by more than ±0.1 kPa during at least two complete respiratory cycles and if equilibrium was obtained within 15 s of rebreathing. Venoarterial CO2 content difference was calculated using the McHardy equation (13). Hemoglobin concentration was assumed to be normal. Measurements were performed with a metabolic unit (Jaeger, Würzburg, Germany).
As described by Stringer et al. (18) cardiac output (CO) was calculated from the measured oxygen uptake (V̇O2) and the estimated arteriovenous oxygen content difference [C(a-vDO2)] as:MATH
V̇O2 was measured breath-by-breath using a metabolic measurement system (SensorMedics, Yorba Linda, CA) and averaged over successive time periods that corresponded to those analyzed with inductance cardiography. Maximal oxygen uptake (V̇O2max) was defined as V̇O2 averaged over the last 30 s of exercise before exhaustion.
Subjects performed exercise in upright position on an electronically braked cycle ergometer. They were connected to the metabolic measurement system via a mouthpiece, and the nose was occluded.
Eleven subjects performed a steady-state exercise protocol starting at 25–50 W with 25- to 50-W workload increments every 10–15 min over a time period of 45 min. Five minutes after each increment, one to two averages of duplicate cardiac output measurements taken at time intervals of two min by the CO2 rebreathing method and simultaneous cardiac output determinations by inductance cardiography were obtained.
Twelve subjects performed a progressive ramp exercise protocol with workload increases of 15–30 W·min−1 to exhaustion. Simultaneous cardiac output determinations were obtained by inductance cardiography and oxygen uptake methods over corresponding successive time periods.
Relative changes in cardiac output over successive measurements by inductance cardiography and gas exchange techniques over the course of exercise tests were expressed in percent. In addition, estimates of cardiac output by inductance cardiography were also obtained in absolute units (L·min−1) by individual calibration of the inductance cardiograph by comparison of one cardiac output estimated by the inductance cardiograph to a corresponding measurement estimated by the CO2 rebreathing or the O2 uptake methods, respectively. We assessed agreement between methods by computing the bias and limits of agreement (1).
Comparison of inductance cardiography to the CO2 rebreathing method.
Eleven subjects underwent the steady-state exercise protocol. A mean (±SD) of 7.1 ± 1.8 (range 4–10) paired cardiac output estimates were obtained for each subject by the two methods resulting in a total number of 78 comparisons. Mean maximal workload and heart rate were 141 ± 22 W (range 120–170 W), and 132 ± 17 bpm (range 106–160 bpm), respectively. Mean cardiac output derived from the CO2 rebreathing method was 15.2 ± 3.9 L·min−1 (range 8.7–24.8 L·min−1). Cardiac output estimates by the two methods agreed well with 76 of 78 paired measurements (97%) falling within 20% of each other (Fig. 2). Limits of agreement for cardiac output by the two methods were ±2.8 L·min−1 (Fig. 2) corresponding to ±18% of the individual mean cardiac output.
Successive paired estimates of cardiac output by inductance cardiography and CO2 rebreathing were plotted for individual subjects over the course of the exercise tests (Fig. 3). The trends of changes in cardiac output by the two methods were similar. The mean change (i.e., the mean value of all differences between two successive measurements) in cardiac output over the course of exercise tests was determined by CO2 rebreathing as 1.7 ± 1.6 L·min−1 (range 0.1–7.3 L·min−1; 67 observations). Comparison of relative (percent) changes in cardiac output measured by the two methods agreed well with a bias of 1% and limits of agreement of ±21% (Fig. 4).
Comparison of inductance cardiography to the oxygen uptake method.
Twelve subjects underwent a progressive ramp exercise protocol and reached a mean maximal workload (±SD) of 292 ± 50 W (range 235–375 W). Mean maximal heart rate was 178 ± 12 bpm (range 156–195 bpm). Mean V̇O2max (±SD) was 3240 ± 535 mL·min−1.
Stable linear relations among the arteriovenous oxygen content difference [C(a-vDO2)] and V̇O2 for determination of cardiac output by the oxygen uptake method were assumed at V̇O2 ≥ 30% of V̇O2max (18). Data analysis was thus performed over the time periods corresponding to 30–100% of V̇O2max.
The mean number (±SD) of paired cardiac output determinations per subject over successive 200 heart cycles was 8.1 ± 2.0 (range 5–12), resulting in a total number of 98 paired measurements. Mean cardiac output derived by the oxygen uptake method was 17.2 ± 3.4 L·min−1 (range 11.3–28.5 L·min−1). Mean heart rate was 147 ± 24 bpm (range 98–190 bpm).
The identity plot of cardiac output estimated by the two methods over successive 200 heart cycle periods (corresponding to 68–83 s at maximal exercise) revealed that 97 of the 98 paired values (99%) fell within ±20% of each other (Fig. 5). The limits of agreement were ±2.8 L·min−1 corresponding to ±16%. For averaging periods extending over only 100 heart cycles (corresponding to 32–39 s at maximal exercise), 164 of 171 paired values (96%) fell within ±20% of each other, and the limits of agreement were ±4.2 L·min−1 corresponding to ±23%. If the ensemble averaging periods were further reduced to 50 heart cycles (corresponding to 16–21 s at maximal exercise), then the inductance waveforms were distorted at exercise levels exceeding 80% of maximal oxygen uptake in all subjects, and even at lower work loads in three subjects. Therefore, these data had to be discarded and analysis had to be limited to the remaining nine subjects and to exercise levels up to 80% of maximal oxygen uptake. This provided 92 paired cardiac output estimates. Limits of agreement were ±3.3 L·min−1 corresponding ±19%.
The mean change in cardiac output estimated by the oxygen uptake method over consecutive 200 heart cycles was 0.9 ± 0.6 L·min−1 (0–3.2 L·min−1) corresponding to a mean relative change of 5% ± 3% (range 0–15%). Comparison of relative changes in cardiac output measured over successive 200 heart cycles by the two methods revealed a mean difference (bias) of 0% and limits of agreement of ±22% in 86 comparisons. If the ensemble averaging periods were reduced to 100 heart cycles, then the bias was 0% and limits of agreement were ±24% in 159 comparisons.
We have extended the validation of inductance cardiography, a noninvasive technique for monitoring of cardiac output, to the application during exercise in healthy subjects. Comparison of inductance cardiography with two other noninvasive reference methods revealed close agreement in estimates of cardiac output and of its changes during submaximal and maximal upright cycle ergometry.
Thermodilution and the direct Fick method are used for estimation of cardiac output in clinical practice and research (11). However, as these techniques are invasive and associated with considerable risks, they were not practicable as reference methods for use in healthy volunteers. Therefore, we decided to compare inductance cardiography to two other noninvasive techniques for estimation of cardiac output, the CO2 rebreathing and the indirect oxygen uptake methods, both of which have been validated in prior studies (10,18). Esperson and coworkers (10) found close agreement among cardiac output measurements by the CO2 rebreathing technique and the direct Fick method in 11 healthy subjects. The bias and limits of agreement at rest and during exercise were −0.1 ± 1.4 L·min−1. Using progressive cycle exercise with a ramp protocol similar to the protocol in the current study, Stringer and coworkers (18) estimated cardiac output by the indirect oxygen uptake and the direct Fick method in 5 subjects during 10 exercise tests. Their graphical analysis revealed that the bias among the two methods was close to 0 L·min−1, and that 108 of 112 (96%) of paired cardiac output estimates fell within 20% of each other.
The comparisons of inductance cardiography to the CO2 rebreathing and oxygen uptake methods in the current study revealed an agreement in cardiac output estimates within the limits expected from the variability of the reference methods: 96–99% of paired estimates by inductance cardiography and any one of the two reference methods fell within ±20% of each other. The limits of agreement were ±2.8 L·min−1 (±18%) (Fig. 2) and ±2.8 L·min−1 (±16%) (Fig. 5) for comparisons of inductance cardiography to the CO2 rebreathing and oxygen uptake methods, respectively. Furthermore, limits of agreement of inductance cardiography versus the reference methods in estimation of changes in cardiac output were ±21% (Fig. 4) to ±24%, respectively. These results corroborate our previous observations in critically ill patients where we compared estimates of cardiac output derived from inductance cardiography with thermodilution (4). The bias and limits of agreement in cardiac output obtained by the two methods were 0.2 ± 2.4 L·min−1 in that study, and bias and limits of agreement for comparisons of changes in cardiac output were 0 ± 22%.
The critical effect of the signal averaging period on agreement among inductance and oxygen uptake derived cardiac output during cycle ergometry is illustrated by greater limits of agreement with data collection over 100 as compared to 200 heart cycles (limits of agreement of ±22% vs ±16%, respectively). We assume that the precision of both methods is improved with longer averaging periods. If the averaging period was reduced to 50 heart cycles, the resulting mean inductance waveforms were clearly distorted preventing estimation of cardiac output in all subjects at the highest exercise levels (>80% of maximal oxygen uptake) and in 3 of 12 subjects even at lower exercise intensities. Therefore, our data suggest that satisfactory suppression of random noise in inductance waveforms for precise measurement of cardiac output during maximal bicycle exercise requires averaging periods over at least 100 or more heart cycles, which corresponded to 32–39 s. This compares favorably to methods like thermodilution (11,20) or gated blood pool ventriculography that require 60–90 s or even longer periods of data acquisition at peak exercise and yet are still accepted and highly utilized.
A limitation of inductance cardiography is the requirement for initial calibration by an independent technique to obtain subsequent cardiac output estimates in absolute units (L·min−1). Nevertheless, even without calibration, inductance cardiography allows accurate tracking of relative (percent) changes in cardiac output over the course of exercise (Figs. 3 and 4) as validated by the limits of agreement of ±21% and ±24% in comparison to corresponding values by the two reference methods. In terms of clinical decision-making, and for evaluation of the hemodynamic response to interventions such as exercise or pharmacological treatments, the trends from serial cardiac output measurements represents more relevant information than individual absolute values.
Our data suggest that inductance cardiography has the potential to serve as a valuable tool for unobtrusive monitoring of changes in stroke volume and cardiac output during bicycle exercise in healthy subjects. Combining inductance plethysmographic measurements of the rib cage (as in inductance cardiography) and corresponding measurements by an additional abdominal transducer (as in respiratory inductance plethysmography) may serve to investigate interactions of the cardiac and respiratory systems during exercise carried out with cycle ergometry.
We would like to thank Christina Spengler, Ph.D., Institute of Sports and Exercise Physiology, University of Zürich, Switzerland, for her support.
This study was supported by a grant from the EMDO foundation Zürich, Switzerland.
1. Bland, M. J., and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1: 307–310, 1986.
2. Bloch, K. E. Impedance and inductance monitoring of cardiac output. In: Principles and Practice of Intensive Care Monitoring, M. J. Tobin (Ed.). New York: McGraw-Hill, Inc., 1998, pp. 915–930.
3. Bloch, K. E., J. A. Adams, J. J. Bassuk, M. J. Mangino, and M. A. Sackner. Noninvasive monitoring of cardiac output in human neonates and juvenile piglets by inductance cardiography (thoracocardiography). J. Crit. Care 12: 259–266, 2002.
4. Bloch, K. E., P. C. Baumann, R. Stocker, and E. W. Russi. Noninvasive monitoring of cardiac output in critically ill patients with thoracocardiography. Am. J. Respir. Crit. Care Med. 155: 26–31, 1997.
5. Bloch, K. E., S. Jugoon, H. de Socarraz, K. Manning, and M. A. Sackner. Thoracocardiography: noninvasive monitoring of left ventricular stroke volume. J. Crit. Care 13: 146–157, 1998.
6. Bloch, K. E., S. Jugoon, and M. A. Sackner. Inductance cardiography (thoracocardiography): a novel, noninvasive technique for monitoring left ventricular filling. J. Crit. Care 14: 177–185, 1999.
7. Bucklar, G. B., V. Kaplan, and K. E. Bloch. Signal processing technique for non-invasive real-time estimation of cardiac output by inductance cardiography (thoracocardiography). Med. Biol. Eng. Comput.
8. Clausen, J. P., O. A. Larsen and J. Trap-Jensen. Cardiac output in middle-aged patients determined with CO2
rebreathing method. J. Appl. Physiol. 28: 337–342, 1970.
9. Collier, C. R. Determination of mixed venous CO2
tension by rebreathing. J. Appl. Physiol. 9: 25–29, 1956.
10. Esperson, K., W. Jenson, J. K. Thmsen, K. Eliasen, N. V. Olsen, and I. L. Kanstrup. Comparison of cardiac output measurement techniques: thermodilution, Doppler, CO2
-rebreathing and the direct Fick method. Acta Anesthesiol. Scand. 39: 245–254, 1995.
11. Jansen, J. R. C. The thermodilution method for the clinical assessment of cardiac output. Intensive Care Med. 21: 691–697, 1995.
12. Jernerus, R., C. Lundin, and D. Thomson. Cardiac output in healthy subjects determined with a CO2
rebreathing method. Acta Physiol. Scand. 59: 390–399, 1963.
13. Mchardy, G. J. H. The relationship between the differences in pressure and content of carbon dioxide in arterial and venous blood. Clin. Sci. 32: 299–309, 1967.
14. Sackner, M. A. Measurement of cardiac output by alveolar gas exchange. In: Handbook of Physiology, Section 3: The Respiratory System, Volume IV: Gas Exchange, A. P. Fishman, L. E. Farhi, and S. M. Tenney (Eds.). Bethesda, MD: American Physiologic Society, 1987, pp. 233–255.
15. Sackner, M. A., D. Greeneltch, M. S. Heiman, S. Epstein, and N. Atkins. Diffusing capacity, membrane diffusing capacity, capillary blood volume, pulmonary tissue volume, and cardiac output measured by rebreathing technique. Am. Rev. Respir. Dis. 111: 157–165, 1975.
16. Sackner, M. A., R. A. Hoffman, B. P. Krieger, M. Shaukat, D. Stroh, and J. D. Sackner. Thoracocardiography. Part 2: noninvasive measurements of changes in stroke volume; comparisons to impedance cardiograph. Chest 99: 896–903, 1991.
17. Sackner, M. A., R. A. Hoffman, D. Stroh, and B. P. Krieger. Thoracocardiography. Part 1: noninvasive measurement of changes in stroke volume; comparisons to thermodilution. Chest 99: 613–622, 1991.
18. Stringer, W. W., J. E. Hansen, and K. Wasserman. Cardiac output estimated noninvasively from oxygen uptake during exercise. J. Appl. Physiol. 82: 908–912, 1997.
19. Watson, H. L., D. A. Poole, and M. A. Sackner. Accuracy of respiratory inductive plethysmographic measurement of cross-sectional area of models. J. Appl. Physiol. 65: 306–308, 1988.
20. Yelderman, M. L., M. A. Ramsay, M. D. Quinn, A. W. Paulsen, R. C. Mckown, and P. H. Gillman. Continuous thermodilution cardiac output measurement in intensive care unit patients. J. Cardiothorac. Vasc. Anesth. 6: 270–274, 1992.