Reproducibility of Impedance Cardiography Hemodynamic Measures in Clinically Stable Heart Failure Patients


  • Barry H. Greenberg MD,

    1. From the Heart Failure and Cardiac Transplantation Program, UCSD School of Medicine and UCSD Medical Center, San Diego, CA;1 the CardioDynamics International Corporation, San Diego, CA;2 and Statistical Consultant, Palo Alto, CA3
    Search for more papers by this author
  • 1 Denise D. Hermann MD,

    1. From the Heart Failure and Cardiac Transplantation Program, UCSD School of Medicine and UCSD Medical Center, San Diego, CA;1 the CardioDynamics International Corporation, San Diego, CA;2 and Statistical Consultant, Palo Alto, CA3
    Search for more papers by this author
  • 1 Maryann F. Pranulis RN, DNSc,

    1. From the Heart Failure and Cardiac Transplantation Program, UCSD School of Medicine and UCSD Medical Center, San Diego, CA;1 the CardioDynamics International Corporation, San Diego, CA;2 and Statistical Consultant, Palo Alto, CA3
    Search for more papers by this author
  • 2 Lucia Lazio RN, BSN,

    1. From the Heart Failure and Cardiac Transplantation Program, UCSD School of Medicine and UCSD Medical Center, San Diego, CA;1 the CardioDynamics International Corporation, San Diego, CA;2 and Statistical Consultant, Palo Alto, CA3
    Search for more papers by this author
  • and 2 David Cloutier BS 3

    1. From the Heart Failure and Cardiac Transplantation Program, UCSD School of Medicine and UCSD Medical Center, San Diego, CA;1 the CardioDynamics International Corporation, San Diego, CA;2 and Statistical Consultant, Palo Alto, CA3
    Search for more papers by this author

Barry H. Greenberg, MD, Director, Heart Failure and Cardiac Transplantion Program, UCSD School of Medicine and UCSD Medical Center, 200 West Arbor Drive, San Diego, CA 92103


Background and Aims. One of the greatest challenges confronting physicians who are managing the care of patients with heart failure is to acquire objective data that signals treatment effectiveness and/or disease progression. The aim of this study was twofold: 1) to determine the extent to which (real time) impedance cardiography measurements obtained with a specific medical device (the BioZ®) were reproducible in outpatients with clinically stable heart failure; and 2) to establish “normal” ranges of one week hemodynamic variability in this population of patients. Information of this nature would help clinical cardiologists and primary care practitioners to evaluate the implications of their patient's visit-to-visit hemodynamic variability.

Methods. A one group, prospective, time series design was used. The sample consisted of 62 patients who had clinically stable heart failure and who were being treated in an outpatient heart failure clinic at a university medical center. BioZ® hemodynamic measures of cardiac output, contractility, and after load were obtained at five points in time: two, 10, and 60 minutes resting following a 40–50 foot walk on the first day and at two and 10 minutes resting following a 40–50 foot walk on the second day, one week later.

Results. Small but significant changes in cardiac output and cardiac index (mainly due to changes in heart rate) were seen during the 60 minute period on week one. Stroke index did not change during this period. In general, reproducibility between measurements taken on the same day and between days was quite good. Establishment of 95% confidence intervals helped define boundaries of variability in this population. Further clinical evaluation of the four patients whose values exceeded the 95% confidence intervals revealed unexpected, potentially relevant changes that could have accounted for their interday variability. Conclusion. The BioZ® impedance cardiography measurements are responsive to hemodynamic activity-rest changes and are reproducible at a one week interval in clinically stable heart failure patients being treated in an outpatient clinic. Stroke index is a better measure of patient status than cardiac output or cardiac index.

Heart failure has remained one of the most difficult medical conditions to manage throughout recorded medical history.1 Despite significant advances in scientific knowledge and technology in the latter half of the 20th century, one of the greatest challenges in managing patient care results from the need for readily accessible, objective data that signals disease progression and/or treatment effectiveness.2–4 Obtaining, recording, and trending this data is dependent upon technology that produces valid, reproducible, and cost effective measurements of cardiac function in a timely manner. While both invasive and noninvasive technologies have been developed and used effectively in the assessment, diagnosis, and evaluation of treatment outcomes, most require specialized environments, costly equipment, and specially trained medical personnel to obtain and/or interpret the data. Because of the cost and/or risk associated with these technologies, repeated hemodynamic measures that would enhance medical management and fine tuning of care are not obtained in ambulatory care settings (the locus of care for a major portion of the heart failure patient population for a significant period of the illness).

One technology that appears to meet the criteria of safety, cost effectiveness, and ease of use is thoracic electrical bioimpedance (TEB). The application of TEB technology in the assessment of cardiac function is known as impedance cardiography (ICG). ICG was introduced in the late 1930s;5 however, systematic study of TEB technology was delayed until the late 1960s when the National Aeronautics Space Administration considered it for use in monitoring astronauts during the Apollo manned flight into space.6 TEB technology is based on Ohms law, which states resistance (impedance or Z) to alternating current flow is inversely proportional to voltage (U) when the current (I) remains constant (Z=U/I).7 Because “…the properties of a conductor are related to the inherent resistance of the conducting medium, the length of the conduit, and its mean cross sectional area…,”8 changes in impedance (delta Z or dZ) over changes in time (delta t or dt), when current is held constant, reflect a change in the properties of the conducting medium (inherent resistance, length or cross sectional area).9,10

ICG works by: 1) introducing a low energy, high frequency alternating current (I) of a specific frequency and amplitude, which is held constant, through the thorax; 2) measuring the voltage drop (U); and 3) calculating the impedance (Z) to the flow of energy. The delivered current is between 20–200 kHz, which cannot be felt and is not harmful. Because the thorax is composed of both conductive (intracellular, extracellular, and interstitial fluid and blood) as well as less conductive or nonconductive substances (bone and air), the thorax is a good conductor of energy. Baseline impedance (Z0), the amount of impedance/conductivity of all of the conductive matter in the thorax, is measured. Changes in impedance over changes in time are calculated (dZ/dt) and displayed numerically and graphically as a wave form to reflect the dynamic state of fluid (the most conductive medium) in the thorax. Because blood is a highly conductive medium and the aorta is highly distensible, changes in thoracic impedance reflect changes in the pulsatile volume of blood flow through the aorta in response to electrical and mechanical events in the myocardium.11–14

The accuracy, precision, and reproducibility of ICG measures of hemodynamics have been demonstrated in a variety of situations and subjects ranging from normal healthy subjects, during normal pregnancy and/or the intrapartal period to acute and/or critical illness, including during general or regional anesthesia, dialysis, open heart surgery, post coronary artery bypass surgery, and cardiac rehabilitation with generally positive results.15 However, it has not been extensively tested with heart failure patients being managed in ambulatory care settings. In order to successfully apply knowledge about ICG from the populations previously studied to clinical monitoring of patients' responses to treatment for heart failure in an outpatient setting, it is important to determine the ability of ICG to measure hemodynamic changes in heart failure patients as well as the reproducibility of ICG measurements in clinically stable heart failure patients. This information would facilitate interpreting ICG measured hemodynamic changes that are trended over time and identifying the need for modifications in the treatment plan. The following questions were addressed in this study.

“In clinically stable, heart failure patients treated in an ambulatory care setting:

  • Are ICG measurements of hemodynamic function reproducible at a one week interval?
  • How variable are hemodynamic parameters instable heart failure patients over a one week interval?”

The purposes of this study were: 1) to establish the reproducibility of ICG hemodynamic measurements obtained with a specific medical device (the BioZ® Noninvasive Hemodynamic Monitor, CardioDynamics International Corporation, San Diego, CA) on patients with clinically stable heart failure; and 2) to establish normal ranges for hemodynamic variability in this patient population when measurements are obtained with the same instrument, under the same or similar conditions, at a one week interval.


Study Design. A one group, prospective, time series (repeated measures) design was used. Patients were clinically evaluated on Day 1 and on Day 2 (one week later) by their examining cardiologist to establish their suitability (clinical stability) for inclusion or continuation in the study. ICG hemodynamic measurements using the BioZ® ICG device were obtained at five points in time: on Day 1 at two, 10, and 60; on Day 2, one week later, at two and 10 minutes rest following ICG sensor placement after a 40–50 foot walk from the waiting area to the exam room. Environmental conditions were consistent for both observation points. The protocol was reviewed and approved as a minimal risk study by the UCSD Medical School Human Subjects' Committee.

Inclusion Criteria. Adult patients who where being followed in the Cardiomyopathy Clinic at a major university medical center in Southern California and who appeared to be clinically stable were considered for inclusion. Clinical stability was defined as: 1) absence of clinically significant changes in physical signs of heart failure; 2) absence of clinically significant changes in self reported symptoms; and 3) no changes in prescribed medical therapy during the period under consideration.

Exclusion Criteria. Patients were excluded if their body surface area estimates exceeded the ranges for BioZ® standardized impedance algorithm calculation: height <47 in (120 cm) or >91 in (230 cm) or weight <66 lbs (30 kg) or >342 lbs (155 kg). In addition, patients were excluded if they: 1) refused to participate; 2) reported a change in their heart failure status or their medical regimen since the most recent previous clinic visit prior to the initial data collection; 3) reported experiencing these changes between the first and second observation periods; 4) had a minute ventilation pacemaker (which also uses impedance technology and, theoretically, can cause pacing and/or monitoring interference); or 5) had aortic valve incompetence.

Laboratory Analysis. ICG data were recorded digitally in the BioZ® system at the time of measurement. These data were transferred to a floppy disk and printed out at the CardioDynamics engineering laboratory for examination of recording adequacy. Numerical data were transferred electronically to an off site statistician for analysis.

Measurement Strategies. The registered nurse (RN) study coordinator (L.L.) approached 64 eligible patients in the clinic waiting area prior to a regularly scheduled clinic appointment. The study procedures were explained and consent was obtained from 100% of the patients who were approached. After signing the consent form, the patient was escorted, walking approximately 40–50 feet, to the examination room where the BioZ® sensors were placed on the root of the patient's neck and on the lower thorax at the mid axillary line at the level of the xyphoid process (Fig. 1) and attached to the BioZ® monitor by cable. The BioZ® system introduces an alternating current (AC) of constant low amplitude (I=2.5 mA) and high frequency (f=70 kHz) into the thorax through the outermost sensors and senses voltage through the innermost sensors. The electrical activity of the myocardium is also sensed and recorded as a nondiagnostic quality ECG. The real time comparison of dZ/dt to ECG events produces measures of hemodynamic parameters using the Z MARC™ (Modulating AoRtic Compliance) proprietary algorithm (CardioDynamics manufacturer's specifications).

Figure 1.

Thoracic electrical bioimpedance (TEB) sensor placement.

Systemic arterial blood pressure was taken by auscultation on the patient's left upper arm by the RN research coordinator (L.L.) at pre-established intervals (see time intervals in following paragraph) which, along with the patient's identifying information (gender, height, and weight) were entered into the BioZ® system. After verifying accuracy of the information on the data screens, continuous patient monitoring was initiated with specific events marked and saved on the continuous recordings.

BioZ® and blood pressure measurements were obtained with patients in a supine, resting position with the head of the exam table elevated between 30°–45° according to patient comfort and breathing ease. Continuous BioZ® measurement was initiated and data were saved at five points in time: on Day 1 at two, 10, and 60 minutes resting, and on Day 2 at two and 10 minutes resting. The same nurse (L.L.) obtained measures using the same procedures on all subjects on both days, which eliminated measurement error due to interobserver variability.

Data Analysis. Two of the 64 patients who agreed to participate, did not return for follow up one week later. Data analysis was carried out for the 62 patients who completed both the first and second days of observation. Data analysis examined the values, mean values, standard deviations, and confidence intervals for each hemodynamic parameter at each point in time and the pattern of change over time. Intrasubject variability over time and the relationships between and among the hemodynamic parameters and patient characteristics were examined. Hemodynamic values for each parameter were compared as follows: two minutes D1 to 10 minutes D1, two minutes D1 to 60 minutes D1, 10 minutes D1 to 60 minutes D1; two minutes D1 to two minutes D2 and 10 minutes D1 to 10 minutes D2. Outlier data were examined to identify characteristics that might be explanatory and might generate hypotheses for future testing.

Statistics. Descriptive statistics, Chi square, two factor repeated measures analysis of variance (ANOVA), paired t-tests, and regression analyses were performed using Systat version Bland-Altman bias plots were created using Analyse-It version 1.44.17


Patient Characteristics. The subjects (N=62) were patients with previously diagnosed heart failure secondary to a variety of etiologies who appeared to be clinically stable, who agreed to participate in this study, and who returned for the follow up visit (Table I). Women were more likely to have nonischemic heart failure (p<0.05), were shorter (p<0.001), and weighed less (p=0.001) than men, which accounted for their significantly smaller body surface area (BSA) (p<0.001). The majority of patients were on digitalis/diuretic/ACE inhibitor combined therapy with or without ß-blocker therapy. Men had higher cardiac output (CO) (p=0.001) and stroke volume (p<0.01). However, the between group differences for cardiac index (CI) and stroke index (SI), which include adjustments for BSA, were not significant. Because there were no other discernible, gender related differences in medication therapy, NYHA Class, or ejection fraction and because the BioZ® proprietary algorithms correct for gender and body size differences in calculating hemodynamic data, subsequent hemodynamic data analysis was carried out with a one group design.

Ejection Fraction  
Body Surface Area (p<0.001)  
Range1.58–2.53 m21.29–2.23 m2
Mean2.017 m21.724 m2
SD±0.194 m2±0.210 m2
NYHA Class  
Etiology of HF (p<0.05)  
Medication Therapy  
Digitalis, diuretic, ACE inhibitor2110
Digitalis, diuretic, ACE inhibitor and ß-blocker136
Other combinations66

Safety . No patients reported any adverse reactions to the procedures associated with BioZ® hemodynamic monitoring. No arrhythmias attributable to the ICG monitoring or skin reactions to the sensors were observed during monitoring or immediately following sensor removal after 60 minutes on Day 1 or after 10 minutes on Day 2.

Efficacy . There were a total of 310 sets of observations for the 62 subjects. Adequate ICG wave forms and complete numerical records were obtained for 308 (99%) of the observation points. The two minute, Day 2 data for one subject and the 10 minute Day 2 data for another subject were incomplete. However, neither subject was an outlier on any of the measures at any of the other points in time.

Findings. Eleven parameters were determined (or calculated using the Z MARC™ proprietary algorithm) at each of the five points in time: heart rate (HR, bpm), mean arterial pressure (MAP, mm Hg), thoracic fluid content (TFC, Ω−1), stroke volume (SV, mL/heart beat) and stroke index (mL/heart beat/m2), CO (L/min) and CI (L/min/m2), systemic vascular resistance (SVR, dynes/sec/cm−5) and systemic vascular resistance index (SVRI, dynes/sec/cm−5/m2), left systolic work index (LSWI, kg m m2), systemic stroke resistance index (SSRI), acceleration index (ACI as 1/second2), index of contractility (IC as 1/second), and systolic time ratio (STR). The summary statistics are presented in Table II.

Parameter2mD1 & 2mD2 r (R2) Bias10mD1 & 10mD2 r (R2) Bias
HR0.84 (0.708) −1.73886 (0.743) −1.217
MAP0.76 (0.572) 0.1150.69 (0.472) −2.050
TFC66 (0.435) −0.0020.65 (0.423) −0.001
Cardiac Ooutput Measures
SV84 (0.707) −1.7270.86 (0.732) −1.998
SI0.76 (0.575) −0.7210.78 (0.602) −0.867
CO0.90 (0.782) −0.2270.86 (0.736) −0.181
CI0.79 (0.628) −0.10578 (0.601) −0.085
Workload Measures
SVR0.73 (0.54) 40.90.71 (0.51) −9.8
SVRI0.66 (0.44) 83.90.67 (0.44) −18.0
LSWI0.75 (0.568) −0.8660.75 (0.564) −2.030
SSRI0.70 (0.485) −0.3280.68 (0.462) −5.500
Contractility Indices
ACI0.68 (0.468) −0.0080.77 (0.593) −0.019
IC (VI)0.76 (0.578) 0.0010.77 (0.589) 0.000
STR0.53 (0.283) −0.0110.58 (0.333) −0.026
HR=heart rate; MAP=mean arterial pressure; TFC=thoracic fluid content; SV=stroke volume; SI=stroke index; CO=cardiac output; CI=cardiac index; SVR=systemic vascular resistance; SVRI=systemic vascular resistance index; LSWI=left stroke work index; SSRI=systemic stroke resistance index; ACI=acceleration index; IC(VI)=index of contractility; STR=systolic time ratio

There was sufficient intersubject variability in each of the parameters to conduct a regression analysis; however, a clustering of values was noted for TFC (even though the range was adequate for statistical analysis), which may have contributed to lower interday correlations for this variable (r=0.65–0.66) as presented in Table II and illustrated in Figure 2.

Figure 2.

Thoracic fluid content (TFC) scatter plot and bias plot for two minutes on Day 1 and Day 2.

BioZ® Responsivity to Hemodynamic Changes . Prior to addressing the study questions, the hemodynamic change patterns over a 60 minute period of time on the first day of observation and a 10 minute time period on the second day, were examined to determine if the BioZ® measurements were responsive to hemodynamic changes during rest following delimited physical activity. Heart rate, CO/CI, SV/SI, and LSWI decreased from two to 10 minutes on both days. On the first day, these parameters continued to decrease between 10–60 minutes. The greatest amount of change was recorded between the two and 60 minutes measures on the first day.

The two factor, repeated measures analysis of variance for intrasubject changes over time, demonstrated a statistically but not clinically significant change in HR (p<0.02), CO (p<0.03), CI (p<0.03), SSRI (p=0.004), and LSWI (p<0.05). Paired t-tests were performed to determine which of the changes were most significant. The results indicated that the two to 10 minutes changes were significant only for HR (p<0.01), CO (p=0.02), and CI (p<0.03). However, the two to 60 minutes differences in HR (p<0.001), CO (p<0.001), CI (p<0.001), and SV (p<0.05) were greater than the two to 10 minutes changes. However, SI did not change significantly following two, 10, or 60 minutes rest. These findings are compatible with what is known about the effect of neurohormonal responses to physical activity and postural changes on these sensitive parameters. It is also known that SI, workload, and measures of contractility should be less sensitive to such limited changes in activity and posture. Thus, ICG measures obtained with the BioZ® were responsive to the anticipated hemodynamic changes that occur over a period of rest following delimited physical activity among clinically stable heart failure patients.

Reproducibility . To address the first question: “Are ICG measures of hemodynamic functions reproducible at a one week interval?” the “2 to 10 minutes change patterns” on Day 1 were compared to the “2 to 10 minutes change patterns on day 2.” The direction and magnitude of change were similar for all parameters except MAP and TFC, which revealed significant day by time interaction effects. MAP increased slightly on Day 1 but decreased over the same time interval on Day 2 (p<0.02). TFC decreased on Day 1 but remained constant on Day 2 (p<0.03). As noted previously and illustrated in Figure 2, the values for these patients clustered around a narrow range (0.03–0.045 for TFC), which confounds analysis of this parameter.

The strength of the linear relationships between the measures taken on Day 1 and those taken on Day 2 at both two and 10 minutes were examined. As illustrated in Table II, the interday correlations were moderate (r=0.5 for STR) to strong (r=0.9 for CO indicating that the variability in each of the parameters on the first day accounted for 25%–80% of the variability on the second day. Paired t-tests revealed that only the interday differences for CO and CI at two minutes were significant (p<0.03 and p<0.05) but the differences were not clinically significant. None of the interday differences at 10 minutes were either statistically or clinically significant.

Finally, bias and precision were calculated using the Bland-Altman formula18 and considering the Day 1 measures as the independent variable (or “gold standard”) and the Day 2 values as the dependent variable. Table III presents the bias, as well as the interday correlations for each parameter. These findings illustrated that ICG measures obtained with the BioZ® are highly reproducible over a one week interval in clinically stable heart failure patients.

Hemodynamic Variability in Clinically Stable Heart Failure Patients . The second question: “How variable are hemodynamic parameters in stable heart failure patients over a one week interval?” is dependent on adequate reproducibility of the measurement strategies in a “stable” population when the measurements are taken under similar conditions. The reproducibility of the BioZ® ICG measures was demonstrated in this study and in other studies with other patient populations.19 However, because the same data set were used to address the first and second study questions, it would be important to replicate this study with another, larger sample of clinically stable heart failure patients before the variability parameters identified in this study are used as the primary or sole indicator for modifying therapy in clinical practice.

Data for outliers who exceeded the 95% confidence intervals for the amount of interday variability they demonstrated were examined. There were two men and two women who were variability outliers on multiple (four or more) variables for both the two minute and 10 minute interday comparisons. These patients did not differ from the nonoutliers in relation to age, gender, BSA, ejection fraction, or NYHA Class. Two of the four variability outlier patients failed to disclose that they had medication changes: one had been self medicating with herbals which contained ephedrine; the other patient had a vasoactive medication added to her therapy by her private physician between the first and second measurement days without notifying the clinic. The third variability outlier patient developed overt symptoms of deterioration within one month after the study measurements and had an urgent transplant. The fourth variability outlier patient was the only patient whose heart failure was secondary to primary pulmonary hypertension. A review of existing clinic records (approximately six months after the study measurements) revealed that one patient died from metastatic breast cancer; her hemodynamic variability was within the 95% confidence intervals. No other patients developed problems that could be identified through the existing clinic records.

Based on the mean values, standard deviations, bias, confidence levels for the bias, and the follow up information obtained about the outliers, it was concluded that if a patient's hemodynamic values fell within the range of the 95% confidence interval for the bias, it would be equivalent to “no change” in the interday values. If the patient's values changed, but were within the 95% confidence intervals for the upper and lower levels of agreement for the value, this would be considered “normal” interday variability for this population. Figure 3 presents the profile of stable patients based on the 10 minute, Day 1 mean values ± one standard deviation.

Figure 3.

Profile of hemodynamic values at 10 minutes resting on Day 1. MAP=mean arterial pressure; SV=stroke volume; SI=stroke index; CO=cardiac output; CI=cardiac index; TFC=thoracic fluid content; ACI=acceleration index; VI=velocity index; STR=systolic time ratio; SVR=systemic vascular resistance; SVRI systemic vascular resistance index; LSWI=left stroke work index.


This study demonstrated that ICG is a useful technology for evaluating the status of heart failure patients in an outpatient setting. The technology is noninvasive with minimal risk, provides real time hemodynamic data, and is easy to use. Further, the BioZ® provides trended data, which facilitates evaluating changes over time. The finding that measures SI, contractility, and workload are more stable indices (less influenced by environmental conditions and/or neurohormonal responses to activity rest or postural changes) than HR, arterial pressures, and CO which encourages the use of these measures over time to evaluate responses to treatment and disease progression. Previously, these measures were available only through the use of invasive technologies or complex, noninvasive technologies, such as echocardiography. Because those operator dependent procedures involve either risk for patient safety or are costly, their use in routine patient evaluation has been limited. Thus, changes in clinical signs and symptoms become the main criteria for evaluating patient and disease progress in the outpatient setting. It is known that there is a substantial (but unclear duration) time lag between structural changes at the cellular level, and functional changes at the organic level, and further time lags between functional changes and the onset of clinical signs and symptoms of deterioration. Thus, sole reliance on clinical signs and symptoms produces a deleterious, and perhaps irreparable, delay in appropriately modifying the treatment program and the course of the disease. Obtaining important, real time data about hemodynamic functional status may improve the response time to indications of deterioration. Modifications in treatment may be initiated earlier in the disease trajectory and may improve the outcomes.

Limitations. The central limitation of this study lies in incomplete data for the short and long term outcomes for this group of patients and the lack of serial measurements over time. These data sets are needed to clearly establish the “natural” history of clinically stable heart failure and to establish the predictive validity of the variability “norms” identified in this study. This study is also limited by the fact that it was a single site study with limited variability in treatment modalities and environmental conditions. However, these limitations can also be considered controls on extraneous variables that might influence the variability of hemodynamics, which (although considered a limitation) also adds strength to the findings. The data from the outliers is highly suggestive that if the patient's interday changes exceed the limits established in this study that there is a need for further evaluation.


ICG hemodynamic evaluation using the BioZ® is appropriately responsive to anticipated hemodynamic changes from delimited physical activity to the resting state over a period of one hour. The BioZ® ICG values are reproducible in a one week interval in clinically stable heart failure patients being treated in an outpatient setting. This study established ranges of hemodynamic variability in stable heart failure patients seen in a clinic setting. Additionally, it demonstrated the usefulness of ICG monitoring over time in research as well as in clinical practice.