Critical Conversation

Thank you all for participating in this discussion today. I’d like to start out with a little background information about the general principles underlying the field. Can you give me some historical perspective on noninvasive hemodynamic techniques? What do you see as the major impediments to the development of devices that can accurately measure cardiac output noninvasively?

There are a number of noninvasive hemodynamic monitors, including Vigilio, LiDCO, ECOM, PiCCO, ICG as well as Deltex, and NICO. Monitors are developed, tested, and then marketed. Each new device is thought initially as a great advance and then they run into the same barriers. The first is simply that it is reasonably hard to measure cardiac output accurately. When one starts comparing the new device with the thermodilution cardiac output monitoring, one finds that there is substantial error in thermodilution measurements and not tremendous variability in the true signal. Thermodilution cardiac output measurements done with room temperature injection have a limit of agreement of ±2 L/min. When one compares the correlation of 2 measurements taken 1 minute apart, the R² is 0.5. With very carefully done iced injectates the limit of agreement is ±1 L/min with an R² of 0.7. The variation in cardiac output is very small with 70% of measurements being within 20% of a cardiac index of 2.0 L/min. What this variability and accuracy of the pulmonary artery (PA) thermodilution means is very little signal with lots of noise. When a new device is compared with the PA thermodilution measurements, one must take the variability of the PA data into account. Critchley et al ¹ have described statistical tests to allow this comparison, but the bottom line is simply that there is a lot of noise and not a lot of signal, which makes comparisons difficult.

To show you the difficulty in measuring cardiac output accurately, when we developed ECOM, we used a porcine model with chronically implanted Transonic transit time flow probes on the ascending aorta as the gold standard with occlusion of the inferior vena cava and PA to change cardiac output. The chronically implanted flow probes were extremely accurate real-time monitors with beat-to-beat measurements of stroke volume. When we use exactly the same type of probes in patients undergoing cardiac surgery, we found they were not clinically useful. Sometimes we would get the correct cardiac output while at other times it would be off by 50%! This is a flow probe clamped on the ascending aorta and when it is used clinically it is off by 50%! Imagine how inaccurate measurements are when a Doppler signal is obtained from the descending aorta with a guesstimate of aortic diameter.

When one gets a new device that is accurate enough to use clinically, adoption by physicians is slow. Many of the devices require training to use, continuous adjustments, or calibration. If a system requires calibration to get accuracy, such as many of the pulse contour systems, there is limited value. Why should a clinician use a device that requires a second device to calibrate the first device prior to use? If the calibration system requires a central line and a femoral arterial catheter, the risk is greater than the risk of a PA catheter, so why bother. If the system requires injection of lithium every time hemodynamics change, why bother. The basic problems with cardiac output monitors include: accuracy is difficult to achieve, the “gold standard” PA thermodilution measurement has noise, there is very little true variation in cardiac output because clinicians are very good at maintaining cardiac output, and then adoption is slow because it takes enormous time and effort to change clinical practice.

All noninvasive hemodynamic techniques are based on assumptions with regard to anatomic and functional “normalities.” Any pathology or even “outliers” to normal thus increase the bias of calculation. Since these devices are by definition “noninvasive,” there is no potential to either detect or correct these deviations. This impossibility is in my opinion also one of the major impediments.

What are the physiological limits for accurately measuring cardiac output? What is the best at high cardiac output? At low cardiac output?

This depends on how measurement is estimated or calculated.

All these systems use specific algorithms based on different variables either based on flow or anatomical/physiological assumptions (impedance, compliance, point of measurement). In general, algorithms are more accurate at high flow states (high cardiac output). However, determining cardiac output has changed from the differentiation of high and low toward adequate and inadequate, which relativizes the accuracy of high and low cardiac output states.

I agree, the physiologic limits depend on the device. Chronically implanted Transonics Transit Time Flow probes work over the range of 0 to 20 L/min. Thermodilution cardiac output using a pulmonary artery catheter is reasonably accurate between 2 and 12 L/min. Most cardiac outputs measured in cardiac surgery are between 2 and 7 L/min with 70% of measurements with a cardiac index within 20% of 2.0 L/min/m².

Have you all done some comparisons of different noninvasive devices against each other? What is the correlation between devices?

We have tried Transonics Transit Time Flow meters, PA thermodilution catheters, ECOM cardiac output monitors, and CardioDynamics BioImpedance monitors. ECOM was correlated with PA data at 101 male patients undergoing cardiac surgery. Patients were studied with an average age of 64 ± 9 years, weight 91 ± 19 kg, height 175 ± 8 cm, body surface area of 2.1 ± 0.2 m², and ejection fraction 56% ± 12%. Linear regression between thermodilution and ECOM cardiac output measurements had a correlation of R² = 0.63, slope = 0.90, intercept = 0.14 L/min, limit of agreement −2.30 to 1.68 L/min, bias = −0.31 L/min, and standard deviation = 1.02. CardioDyanmics BioImpedance did not work well enough to study. Transonics Transit Time Flow meters worked well when they worked and then suddenly were off by 50% and we stopped using them after 125 patients were studied clinically.

We have compared double indicator dilution techniques and pulse contour devices with PA catheter. All showed good correlation within clinically acceptable limits of agreement. However, devices without potential to calibrate cannot be recommended, since they shift or exceed the boundaries of accuracy without indicating and impossibility to detect or correct.

Art, what is the principle behind the ECOM? Are there certain patient parameters that affect the accuracy?

ECOM measured electrical impedance in the chest near the aortic arch. It measures impedance in 3 dimensions. There are 2 types of phenomena that change electrical impedance. The first is a change in conductor volume. The CardioDynamics Volume Conductance catheter uses this effect. With volume conductance systems, a multi-electrode catheter is placed in the left ventricle. As conductor volume (blood) changes, the signal changes. One integrates the change in impedance over five electrode combinations and sums the result. You then multiply by a fudge factor that converts ohms into milliliters to get the volume.

ECOM uses a different effect. If you change the velocity of a conductor, there is a small change in electrical resistance. The current is carried by ions in an ionic solution. The ionic velocity is on the same order of magnitude as the fluid velocity. When you change fluid velocity, you change ionic velocity, and the impedance changes a small amount. We measure the impedance which changes with flow. Initially we integrated the signal, multiply by a factor to convert from ohm second into milliliters, to get volume. The impedance waveform, measured perpendicular to flow, looks very much like the aortic flow measured with a transit time flow probe. The impedance waveform measured parallel to flow, shows persistent effects when there is turbulence. To correct for the fact that the aorta curves through 270° in the chest, we measure everything in 3 dimensions and then pick the best signals. This picking of signals eliminates the need to adjust the tube to improve signal quality. The electronics does all the adjusting. To improve accuracy we made a multiparamenter model that measures a number of factors from the impedance waveform. These factors include a number of timing signals. We multiply them by a factor to get stroke volume. We then multiply by the heart rate to get cardiac output. Since the signal measured parallel to flow is affected by turbulence, we can tell when there is turbulent flow. The occurrence of turbulence gives us a measure of flow high enough to exceed the Reynolds number and it is used for calibration.

One of the medical conditions that confuses the ECOM system is severe aortic stenosis. In patients with severe aortic stenosis with valve areas of 0.6 to 0.8 needing valve replacement, ECOM will read a liter/minute too high. Once the valve is replaced, it will read accurately. Turbulence is read as high cardiac output.

So what would you consider the gold standard, invasive or noninvasive for measuring output? How about echocardiographic measurements of cardiac output?

I believe any measurement that is able to “measure” cardiac output could be used as standard. For a long time, thermodilution using a PA catheter has been the standard but more recently, publications of calibrated pulse contour devices have supplemented. Echocardiography is also perfect tool for measuring cardiac output but is limited by the ability to give numbers in real time and by interobserver interpretation.

I think the accepted “gold” standard is iced thermodilution measurements. Measurements must be made in large numbers such as 6 to 15 measurements. If any measurement is greater than 15% from the mean it should be discarded. If there are fewer than 3 measurements, all should be discarded and started again. When clinicians make a single room temperature injection they are getting data that is ±2 L/min and should be taken with great suspicion.

What is your experience with measuring cardiac output with the dye indocyanine green (ICG)? Are there devices that can give real-time hemodynamic parameters using ICG? The dye dilution method using ICG was an early method of measuring cardiac output, but seems quite laborious.

There are currently no clinically relevant noninvasive bedside cardiac output devices using ICG measurement. At one time ICG cardiac output determination was among the most precise methods and the underlying assumptions have been implemented into the Pulsion pulse contour device with clinically acceptable worsening of precision.

I do not see a particular advantage using pulse spectrophotometer for determination of cardiac output. Cardiac output estimates appeared to be reasonable when determined during steady state. However, in a non–steady state, for example, during massive blood loss, estimates seemed not always reliable. I personally do not see any added value using this technology for a single point/single determination of the cardiac output. My enthusiasm is further tempered in view of newer technology that allows continuous cardiac output monitoring. Having said this, it is important to recognize that these new technologies also appear to have a significant performance bias during unstable periods in liver transplantation. An additional shortcoming of pulse spectrophotometer is the need for repeated injections of ICG, which can be quite costly and may lead to allergic reactions. The need for sufficient ICG clearance eliminates the ability to measure cardiac output serially within several minutes.

Does the fact that ICG is cleared by first-pass metabolism by the liver prove helpful in evaluating liver function in these patients? Could it be used to track liver function in patients with fulminant hepatic failure? Are there any serious side effects of ICG?

The test using several different endpoints (ie, 15 minute retention rate) has been extensively used in Asia for patient suitability scheduled to undergo hepatic resection. It is generally not used in North America or Europe. Several small studies demonstrated its usefulness in predicting postoperative liver function. However, on closer examination, concerns remained regarding the discriminatory power of the test, especially in patients with mild to moderate disease.

I have no experience using ICG in fulminant hepatic failure but I doubt that performance of the test would change clinical decision making.

As already eluded to, ICG has a real risk of allergic reaction with repeated exposure and also cross-reacts with shellfish allergies. Overall, it is my impression that ICG remains largely a useful research tool, at least in North America and Europe.

Yes, ICG measurement has become useful as one of the most (and only) reliable bedside test of “dynamic” liver function for various hepatic diseases. A possible exception that may eventually become a reliable dynamic test is a recently published ¹³C methacetin breath test method. ² ICG gives a reliable estimate of “remnant” liver function as well as the potential to recover. There are rare allergic reactions described—personally never seen.

What devices use arterial waveform analysis or pulse-pressure variation to determine cardiac output? How accurate are they at determining cardiac output?

There are a number of these devices. They are divided into 2 groups—calibrated and uncalibrated. Calibrated systems include PICO and LiCO. PiCO uses a central line for injection, and a femoral thermodilution sensor to calibrate using thermodilution. LiCO uses lithium injections to calibrate. The results are acceptable but the systems must be recalibrated every time resistance or compliance changes. One is essentially assuming resistance and compliance are unchanged since the last calibration. The correlations are average. Vigileo is uncalibrated and uses the arterial waveform to calculate the resistance, compliance, and then derive a cardiac output. The correlations are poor with some having R² of 0.2. With an R² of 0.2 it would be reasonable to just guess the cardiac output.

Most pulse contour devices use changes of the area under the curve to determine changes in cardiac output. However, the changes (variations) themselves are much more interesting, since they reflect the changes in preload to inspiratory pressure changes and thus are an estimate of volume responsiveness. Precision in cardiac output measurement is not their first aim but the potential to predict the reaction to fluid boluses (will there be an increase in cardiac output?)

Well thank you very much for you participation in this discussion. This seems like it’s an ever-evolving field and one worth keeping track of into the future.