Share this article on:

The Effect of Peripheral Resistance on Impedance Cardiography Measurements in the Anesthetized Dog

Critchley, Lester A. H. MD, FFARCSI*; Peng, Zhi Y. MB BS, MD*; Fok, Benny S. BSc*; James, Anthony E. BVSc, MSc

doi: 10.1213/01.ANE.0000150602.40554.EB
Technology, Computing, and Simulation: Research Report

In the vasodilated and septic patient, the impedance method of measuring cardiac output (CO) may underestimate the true value. In this study, we sought to determine whether impedance CO (COIC) measurements are influenced by total peripheral resistance (TPR). In eight anesthetized and ventilated dogs, a high-precision flowprobe was placed on the ascending aorta, and direct CO was measured (CO flowprobe (COFP)). Mean arterial blood pressure was measured from the femoral artery. Simultaneous COIC measurements were made. TPR (mean arterial blood pressure × 80/COFP) was varied over 1–2 h by using infusions of phenylephrine and adrenaline and inhaled halothane. The bias between methods of CO measurement (COIC − COFP) was calculated and compared with TPR by using correlation and regression analysis. A total of 547 pairs of CO measurements were collected from the 8 dogs as TPR was varied. COFP changed by a mean of 190% (range, 89%–425%), and TPR changed by a mean of 266% (range, 94%–580%) during the experiment. The impedance method underestimated CO when TPR was low and overestimated CO when TPR was high. There was a logarithmic relationship between the CO bias and TPR. Correlation coefficients (r) between the CO bias and TPR ranged from 0.46 to 0.89 (P < 0.0001). The bias changed by 0.62 ± 1.8 L/min, or by 34%, every time TPR halved or doubled. This finding explains the poor agreement between COIC and other methods of CO measurement found in validation studies involving critically ill patients.

IMPLICATIONS: Impedance cardiac output measurements are influenced by peripheral resistance such that cardiac output will be underestimated by the impedance method when resistance is low and overestimated when resistance is high.

*Department of Anaesthesia and Intensive Care and †The Laboratory Animal Services Centre, The Chinese University of Hong Kong, Shatin, Hong Kong, China

Accepted for publication October 29, 2004.

Address correspondence and reprint requests to Lester A. H. Critchley, MD, FFARCSI, Department of Anesthesia and Intensive Care, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, Hong Kong, China. Address e-mail to

The measurement of cardiac output (CO) in a clinical setting is reserved for the high-risk patient because of the complexity, invasive nature, and expense of currently available techniques (1,2). However, there can be little argument that the availability of a simple and reliable method of measuring CO would make CO measurement more accessible to more patients (2). Impedance cardiography is an inexpensive and noninvasive approach to continuous CO measurement (3–5). The impedance method also detects blood-flow changes in the aorta (6), but how this information is translated to valid CO measurements remains a subject of much controversy. Ever since the method was developed by Kubicek et al. (5) more than 30 years ago, serious doubts have been cast on its reliability (7–9). In particular, comparisons with the thermodilution method in the intensive care setting have been disappointing (10–13).

The impedance method involves detecting small changes in electrical resistance during the cardiac cycle to a high-frequency low-amperage current passed through the thorax. The thoracic impedance changes cyclically as a function of time, and this results in a repetitive wave form. Several impedance variables are measured from the wave form, and these are used to estimate aortic blood flow (3–5). One of the most notable shortcomings of the impedance method is that we do not fully understand how the impedance changes during the cardiac cycle relate to stroke volume and CO. Several factors are known to affect the size and shape of the impedance wave form, such as changes in hemoglobin concentration (14), excessive lung fluid (15), and body habitus (16). Some investigators have also commented that impedance cardiography underestimates CO in the septic shock and vasodilated patient (11,17,18). Thus, changes in peripheral vascular resistance may also influence the size and shape of the impedance wave and CO measurements.

To investigate this hypothesis further, we developed an anesthetized dog model in which CO was accurately measured with an aortic flowprobe. We used this model to study the effects of pharmacologically induced changes in peripheral vascular resistance on impedance CO measurements. The impedance device used in the study was the RheoCardioMonitor (ACMA, Singapore) (19,20).

Back to Top | Article Outline


Ethical approval for the study was obtained from the Animal Research Ethics Committee of the Chinese University of Hong Kong. Male mongrel dogs were provided by the Laboratory Animal Services Centre of the Chinese University of Hong Kong. Anesthesia was induced by using IM ketamine 10% (5 mg/kg) and xylazine 2% (2 mg/kg) and was maintained throughout the experiment by using inhaled halothane 0.5%–1.0% in oxygen. The trachea was intubated, and the lungs were ventilated with a tidal volume of 10–15 mL/kg at a frequency of 12–15 breaths/min. Muscle relaxation was provided by rocuronium. IV access was secured in the forelimb and used to administer IV fluids (warmed saline at 2 mL · kg−1 · h−1) and drugs. The right femoral artery was cannulated to allow mean arterial blood pressure (MAP) recordings. Body temperature was maintained by covering the dog with an insulated blanket.

A thoracotomy at the left fourth intercostal space was performed. The pericardium was incised longitudinally to expose the aortic root. The ascending aorta was separated from the pulmonary artery for 2–3 cm by blunt dissection using the finger. The free fat surrounding the aorta was carefully removed. A snugly fitting flowprobe, either a 16- or 20-mm A-series ultrasonic probe (Transonic Systems Inc., Ithaca, NY), was placed around the ascending aorta, and ultrasonic gel was applied. A probe size was chosen that neither compressed the vessel wall nor was so loose that it allowed kinking. The probe cable was brought out of the thorax posteriorly. The pericardium was closed with sutures, a chest drain to underwater seal was inserted, the collapsed lung was reexpanded, and the chest wall was closed with sutures.

Aortic blood flow was measured by the flowprobe, which used a high-precision four-crystal array. It was connected to a T106 single-channel flowmeter (Transonic Systems) that also processed the transduced arterial blood pressure wave. The arterial blood pressure system was kept patent with a 10 mL/h saline infusion. A laptop computer collected the data and displayed it by using the software program WinDaq (DataQ Instruments, OH).

The RheoCardioMonitor consisted of six leads to be connected to the patient, a cardiograph module, a visual display unit, and a dedicated computer that analyzed the signal and stored data. An alternating 100-kHz 2-mA current was applied to the thorax of the dogs by 2 electrodes placed on the head and lower limb. The impedance to this current was detected by two pairs of opposing electrodes placed laterally on the mid neck and lower thorax, at the level of the diaphragm. Subcutaneous needle electrodes were used to improve the skin contact.

Stroke volume (SV) was calculated by the RheoCardioMonitor by using a modified Kubicek equation (5):



where L is the distance between the two current-detecting electrodes, Zo is the mean thoracic impedance, and ρ is the resistivity of the thorax. LVET is the left ventricular ejection time, dZ/dt(max) is the impedance index of flow, and Z(s − q) compensates for the asynchronism between the right and left ventricles, which are derived from the impedance wave form. The factor k in the calculation represents a complex formula based on the subject’s body habitus (20). The system was calibrated before use by imputing the dimensions of the thorax (thoracic length and the circumferences of the neck and lower thorax). CO was derived by multiplying by heart rate.

The anesthetized dog and measurement systems were allowed to stabilize for 30 min after preparation. The dog’s circulation was then first vasoconstricted by using an incrementally increasing infusion of phenylephrine 1–5 μg · kg−1 · min−1. Vasodilation was achieved by increasing the concentration of inhaled halothane (2%–4%) until the MAP decreased to 60 mm Hg. Finally, vasoconstriction was achieved by using adrenaline 0.1–1.0 μg · kg−1 · min−1. The dog’s circulation was allowed to return to baseline for 5 min between treatments. Data collection was restricted to 1–2 h to minimize any drift in measurements by the RheoCardioMonitor.

The flowmeter and RheoCardioMonitor measurements were recorded at 10-s intervals, which were subsequently transferred to Microsoft Excel (Redmond, WA), where they were averaged to provide data at 1-min intervals. These data were inspected for any obvious outliers that differed by >50% from adjacent values. They were replaced by values that were more in keeping with the general trend of the data: the average of adjacent values. Outliers accounted for 1%–5% of data points depending on the particular experiment. The total peripheral resistance (TPR) was calculated by the equation TPR = MAP × 80/COFP (resistance units), where COFP was the CO from the flowprobe. The offset or bias between CO measurements was calculated by finding the difference between the paired measurements (COIC − COFP) (L/min), where COIC was the CO from the RheoCardioMonitor.

For each experiment, CO was measured by each method for the range of circulatory variables tested. The percentage changes for these variables were calculated from the range of values over the minimum value. Time plots and accompanying scattergrams were created to investigate the relationships between TPR and the offset between the two methods of measuring CO (Fig. 1). Correlation and regression analysis was used to determine the relationship between the offset and TPR. This was performed with StatView for Windows (SAS Institute Inc., Cary, NC).

Figure 1

Figure 1

Back to Top | Article Outline


Eight dog experiments were performed. The mean (range) weight of the dogs was 19.5 kg (16–26 kg), and data collection for each experiment lasted 81 min (58–106 min). A total of 547 data sets were collected. The range of hemodynamic variables tested in each dog is shown in Table 1. The average percentage change [mean (range)] in the values measured was 190% for COFP (89%–425%), 129% for MAP (89%–194%), 136% for heart rate (20%–210%), and 266% for TPR (94%–580%).

Table 1

Table 1

COIC measurements underestimated CO when TPR decreased and overestimated CO at higher values for TPR. This dependence on TPR is demonstrated in Figure 1, in which COIC underestimated CO between 20 and 50 min, when the dog in Experiment 1 was vasodilated by the administration of halothane. A scattergram of measurement bias against TPR confirms this effect. There was a significant linear correlation in all eight experiments between the measurement bias and the logarithm of the TPR (Fig. 2). Correlation coefficients for regression of measurement bias against TPR ranged from 0.46 to 0.89 (P < 0.0001) (Table 2). Overall, there was a mean (sd) 0.62 L/min (0.18 L/min) change in the measurement bias each time the TPR was halved or doubled (Table 2). This was equivalent to a 34% shift in COIC, derived from the mean offset (0.62 L/min) divided by the overall mean experimental COFP (1.8 L/min).

Figure 2

Figure 2

Table 2

Table 2

Back to Top | Article Outline


The main finding of this study was that COIC measurements are altered by changes in peripheral resistance, such that vasodilation causes CO to be underestimated and vasoconstriction results in overestimation (Figs. 1 and 2). Previous investigations in this area have mainly evaluated the degree of agreement between impedance measurement and a reference method without regard for the potential influence of peripheral resistance (21).

In all eight experiments, the bias between the impedance and reference methods was strongly correlated with the peripheral resistance (r = 0.46–0.89; P < 0.0001) (Table 2). This finding helps to explain why the impedance method generally does not yield good agreement with other methods of CO measurement, such as thermodilation, when studied in critically ill patients, in whom the peripheral resistance may be very variable. Such patients may be excessively vasodilated because of sepsis or vasoconstricted because of stress or vasopressor administration. Furthermore, we found that halving or doubling the peripheral resistance offset the readings by more than 0.6 L/min, thus introducing an error of 34%. When this error is added to the errors due to other factors that influence impedance measurements (14–16), it provides an explanation for the excessive measurement bias of 40%–60% found in some validation studies (10–12). Clinically acceptable differences in the measurements should not exceed 20%–30% (13).

In studies involving healthy volunteers, peripheral resistance tends to vary less. This accounts for studies that indicate that impedance cardiography works reliably in healthy subjects. The greatest changes in TPR normally seen in volunteer studies are due to postural reflexes evoked by head-up tilting or standing erect, which accounts for up to a 50% increase in TPR (22–24). However, estimates based on data from the present study suggest that standing would change the measurement bias by <20% (i.e., halving or doubling the TPR shift the bias by 34%). Thus, peripheral resistance probably does not vary sufficiently in most volunteer studies to significantly affect impedance measurements.

When using the RheoCardioMonitor, we found that the 10-second measurements of CO varied greatly. Whether this is a problem specific to the RheoCardioMonitor or whether it is present in impedance cardiographs in general is unclear, because the impedance signal from which the impedance data are derived is very fragile and thus difficult to measure with any precision and consistency. The quality of the impedance signal was improved in this study by the use of subcutaneous needle electrodes, and the variability between measurements was also reduced by averaging the data at one-minute intervals and eliminating any obvious outliers, because they were deemed false or artifactual data. These measures helped to improve the resolution of any trends within the data, such as the dependence on TPR. The wide variability of the data can be clearly seen (Fig. 2).

Aortic blood flow is one of the main sources of the impedance signal (4). A study in which the heart was electrically isolated from the thorax showed that contraction of the heart contributed little to the overall impedance change, while the main impedance signal originated from volume changes in the aorta and pulmonary vessels (25). Wang and Patterson (26) mathematically modeled the thorax and estimated that arterial and venous blood flow contributes 57%, the lungs contribute 39%, and structural changes contribute 4% to the overall signal. Although the source of the impedance signal is known, the mechanism by which the impedance signal is produced has not been fully elucidated. Both Kubicek (6) and Visser (27) have suggested that expansion of the aorta during systole is responsible for the impedance change. Visser (27) has further suggested that as blood accelerates into the aorta, its resistivity decreases, thus also contributing to the impedance changes (27). However, these theories on the origin of the impedance signal have never been confirmed through in vivo experiments.

The expansion theory of the aorta as an explanation of the source of the impedance signal is in keeping with our findings. The aorta is a prominent structure within the thorax. It has a high conductivity relative to other tissues, because it contains blood. Thus, any change in aortic diameter should alter thoracic impedance and, during systole, should cause a transitory change in thoracic impedance. Impedance cardiography detects the impedance change and uses it to estimate CO. The validity of this measurement is based on the assumption that aortic distention is directly related to aortic blood flow, and, to an extent, this is true because of the pulsatile nature of aortic blood flow. However, other factors also influence aortic distension, such as vessel wall elasticity and the rate of increase in intraluminal pressure. The latter is dependent on the afterload or compliance of the arterial tree, of which the peripheral vascular resistance is a major determinant. Therefore, if aorta expansion is the main source of the impedance signal, one should not be surprised if variations in TPR alter COIC.

In summary, this study shows the effect of peripheral resistance on COIC measurements. It supports the theory that distention of the aorta is the main source of the impedance signal, and it provides an explanation for the failure of reliable COIC in the critically ill patient. For impedance cardiography to become a useful clinical tool for CO measurement in sick patients, the algorithms used to calculate COIC will need to be adapted to compensate for significant changes in MAP and peripheral resistance.

Back to Top | Article Outline


1. Connors AF Jr, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996;276:889–97.
2. Van De Water JM, Miller TW, Vogel RL, et al. Impedance cardiography: the next vital sign technology? Chest 2003;123:2028–33.
3. Critchley LA. Impedance cardiography: the impact of new technology. Anaesthesia 1998;53:677–84.
4. Bernstein DP. A new stroke volume equation for thoracic electrical bioimpedance: theory and rationale. Crit Care Med 1986;14:904–9.
5. Kubicek WG, Karnegis JN, Patterson RP, et al. Development and evaluation of an impedance cardiac output system. Aerosp Med 1966;37:1208–12.
6. Kubicek WG. On the source of peak first time derivative (dZ/dt) during impedance cardiography. Ann Biomed Eng 1989;17:459–62.
7. Clarke DE, Raffin TA. Thoracic electrical bioimpedance measurement of cardiac output: not ready for prime time. Crit Care Med 1993;21:1111–2.
8. Tremper KK. Continuous noninvasive cardiac output: are we getting there? Crit Care Med 1987;15:278–9.
9. Weil MH. Electrical bioimpedance for noninvasive measurement of cardiac output. Crit Care Med 1997;25:1455.
10. Sageman WS, Amundson DE. Thoracic electrical bioimpedance measurement of cardiac output in postaortocoronary bypass patients. Crit Care Med 1993;21:1139–42.
11. Wong DH, Tremper KK, Stemmer EA, et al. Noninvasive cardiac output: simultaneous comparison of two different methods with thermodilution. Anesthesiology 1990;2:784–92.
12. Young JD, McQuillan P. Related comparison of thoracic electrical bioimpedance and thermodilution for the measurement of cardiac index in patients with severe sepsis. Br J Anaesth 1993;70:58–62.
13. Critchley LA, Critchley JA. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput 1999;15:85–91.
14. Quail AW, Traugott FM, Porges WL, White SW. Thoracic resistivity for stroke volume calculation in impedance cardiography. J Appl Physiol 1981;50:191–5.
15. Critchley LA, Calcroft RM, Tan PY, et al. The effect of lung injury and excessive lung fluid, on impedance cardiac output measurements, in the critically ill. Intensive Care Med 2000;26:679–85.
16. Critchley LA, Leung DH, Short TG. Abdominal surgery alters the calibration of bioimpedance cardiac output measurement. Int J Clin Monit Comput 1996;13:1–8.
17. Bernstein DP. Continuous noninvasive real-time monitoring of stroke volume and cardiac output by thoracic electrical bioimpedance. Crit Care Med 1986;14:898–901.
18. Doering L, Lum E, Dracup K, Friedman A. Predictors of between-method differences in cardiac output measurement using thoracic electrical bioimpedance and thermodilution. Crit Care Med 1995;23:1667–73.
19. Barin E, Haryadi DG, Schookin SI, et al. Evaluation of a thoracic bioimpedance cardiac output monitor during cardiac catheterization. Crit Care Med 2000;28:698–702.
20. Haryadi DG, Westenskow DR, Critchley LA, et al. Evaluation of a new advanced thoracic bioimpedance device for estimation of cardiac output. J Clin Monit Comput 1999;15:131–8.
21. Raaijmakers E, Faes TJ, Scholten RJ, et al. A meta-analysis of published studies concerning the validity of thoracic impedance cardiography. Ann N Y Acad Sci 1999;873:121–7.
22. Blomqvist CG, Stone HL. Cardiovascular adjustments to gravitational stress. In: Handbook of physiology: the cardiovascular system. Baltimore: Williams & Wilkins, 1983:1025–63.
23. Critchley LA, Conway F, Anderson PJ, et al. Non-invasive continuous arterial pressure, heart rate and stroke volume measurements during graded head up tilt in normal man. Clin Auton Res 1997;7:97–101.
24. Smith JJ, Hughes CV, Ptacin MJ, et al. The effect of age on hemodynamic response to graded postural stress in normal men. J Gerontol 1987;42:406–11.
25. Bonjer FH, van der Berg JW, Dirken MN. The origin of variations of body impedance occurring during the cardiac cycle. Circulation 1952;6:415–20.
26. Wang L, Patterson R. Multiple sources of the impedance cardiogram based on 3-D finite difference human thorax models. IEEE Trans Biomed Eng 1995;42:141–8.
27. Visser KR. Electric properties of flowing blood and impedance cardiography. Ann Biomed Eng 1989;17:463–7.
© 2005 International Anesthesia Research Society