A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques

Partial CO2 rebreathing indirect Fick technique for non-invasive measurement of cardiac output

OBJECTIVE

Evaluation in animals of a non-invasive and continuous cardiac output monitoring system based on partial carbon-dioxide (CO2) rebreathing indirect Fick technique.

METHODS

We have developed a non-invasive cardiac output (NICO) monitoring system, based on the partial rebreathing method. The partial rebreathing technique employs a differential form of the Fick equation for calculating cardiac output (QT) using non-invasive measurements. Changes in CO2 elimination (deltaVCO2) and partial pressure of end-tidal CO2 (deltaPETCO2) in response to a brief period of partial rebreathing are used to measure pulmonary capillary blood flow (Q(PCBF)). A non-invasive estimate of anatomic and intrapulmonary shunt fraction (Q(S)/Q(T)), based on oxygen saturation from pulse oximetry (SpO2) and inspired oxygen concentration (FIO2), is added to compute total cardiac output [Q(T) = Q(PCBF)/(1 - Q(S)/Q(T))]. The performance of the NICO was compared with iced 5% dextrose bolus thermodilution cardiac output (TDco) measurements in 6 dogs. Cardiac output was varied using dobutamine, and halothane, and by clamping of the inferior vena cava. Two hundred and forty-six (n = 246) paired measurements of TDco and NICO over a range of cardiac outputs (TDco range = 0.60-8.87 l/min) were compared using Bland-Altman analysis and weighted correlation coefficient.

RESULTS

The Bland-Altman technique yielded a NICO precision of +/- 0.70 l/min (13.8%) with a mean bias of -0.07 l/min (-1.4%) compared to TDco. The weighted correlation coefficient between TDco and NICO values was: r = 0.93 (n = 246).

CONCLUSION

The partial CO2 rebreathing technique for measurement of cardiac output is non-invasive, automated, and based on the well accepted Fick principle. The limits of agreement between NICO and TDco is within the recommended value for NICO to be a clinically acceptable method for cardiac output measurement. The results of this canine study show that NICO performed as well, and in some cases better, than other currently available non-invasive cardiac output techniques over a wide range of cardiac outputs.

Lithium dilution cardiac output measurement: a clinical assessment of central venous and peripheral venous indicator injection

OBJECTIVE

The lithium indicator dilution technique has been shown to measure cardiac output (CO) accurately by using central venous injection of lithium chloride (Li-CCO). This study aimed to compare the measurement of CO by using peripheral venous administration of lithium chloride (Li-PCO) with Li-CCO.

DESIGN

Prospective, observational human study.

SETTING

Surgical intensive care unit.

PATIENTS

Thirty-one patients were studied after major surgery. All patients had arterial, central, and peripheral venous catheters. A total of 24 patients had pulmonary artery catheters.

MEASUREMENTS

Serial measurements of Li-CCO and Li-PCO were made during hemodynamically stable conditions. CO was also measured using thermodilution (TDCO) when a pulmonary artery catheter was present. Data were analyzed by linear regression, the generalized estimating equation, and the comparison method described by Bland and Altman.

MAIN RESULTS

There were 93 Li-CCOs, 93 Li-PCOs, and 216 TDCOs recorded. The ranges of COs were similar: Li-CCO, 2.36-11.52 L/min (mean, 5.22 L/min; n = 31); Li-PCO, 1.63-9.99 L/min (mean, 5.22 L/min; n = 31), and TDCO, 3.28-10.4 L/min (mean, 5.75 L/min; n = 24). There was good linear correlation between Li-CCO and Li-PCO (R2 =.845). The mean difference for Li-CCO-Li-PCO was very small and insignificant (p =.97), and the limits of agreement were acceptable (mean difference +/- sd, 0.0005 +/- 0.64 L/min). The mean difference for Li-CCO-Li-PCO was smaller if the peripheral injection site was proximal rather than distal to the wrist (p =.053). Li-PCO and Li-CCO values were lower than simultaneously obtained TDCO measurements (Li-PCO-TDCO, -0.538 +/- 0.95 L/min, p =.003; Li-CCO-TDCO, -0.526 +/- 0.67 L/min, p =.0001).

CONCLUSIONS

Li-PCO gives a measurement that agrees well with Li-CCO. Accuracy of Li-PCO is probably improved if a proximal arm vein is used. Li-PCO provides accurate measurements of CO without the risks of pulmonary artery or central venous catheterization.

Thermodilution cardiac output. Cold vs room temperature injectate and the importance of measuring the injectate temperature in the right atrium

BACKGROUND

The feasibility of thermodilution cardiac output measurements with the more convenient room temperature thermal indicator instead of cold injectates has been repeatedly investigated. However, the issue has not been addressed with the appropriate statistical approach advocated by Altman and Bland. Furthermore, we wished to determine if the incorporation of a second thermistor in the thermodilution catheter, to measure the temperature of the thermal indicator where it is delivered into the right atrium/superior caval vein, would result in more precise cardiac output measurements.

METHODS

Fifty patients were randomized to receive a single or dual thermistor pulmonary artery thermodilution catheter. Cardiac output was calculated as the average of four injections of 10 ml of isotonic saline. Precision (2 x SD of differences in replicate measurements) for the two catheters and injectate temperatures, and bias and limits of agreement between measurements, with cold and room temperature injectates, were determined.

RESULTS

Precision was (0 degrees C) 0.42 l/min and (20 degrees C) 0.90 l/min, and bias and limits of agreement -0.83 l/min and -1.93-0.27 l/min for the single thermistor catheter. For the dual thermistor system precision was (0 degrees C) 0.34 l/min and (20 degrees C) 0.58 l/min. Bias and limits of agreement were -0.03 l/min and -0.61-0.55 l/min.

CONCLUSION

The second thermistor is redundant if cold injectates are used. If one wishes to use room temperature injectates the single thermistor system is inadequate. A dual thermistor catheter is, on the other hand, acceptable.

A comparison of cardiac output derived from the arterial pressure wave against thermodilution in cardiac surgery patients

In three clinical centres, we compared a new method for measuring cardiac output with conventional thermodilution. The new method computes beat-to-beat cardiac output from radial artery pressure by simulating a three-element model of aortic input impedance, and includes non-linear aortic mechanical properties and a self-adapting systemic vascular resistance. We compared cardiac output by continuous model simulation (MF) with thermodilution cardiac output (TD) in 54 patients (18 female, 36 male) undergoing coronary artery bypass surgery. We made three or four conventional thermodilution estimates spread equally over the ventilatory cycle. In 490 series of measurements, thermodilution cardiac output ranged from 2.1 to 9.3, mean 5.0 litre min(-1). MF differed +0.32 (1.0) litre min(-1) on average with limits of agreement of -1.68 and +2.32 litre min(-1). Differences decreased when the first series of measurements in a patient was used to calibrate the model. In 436 remaining series, the mean difference became -0.13 (0.47) litre min(-1) with limits of agreement of -1.05 and +0.79 litre min(-1). When consecutive measurements were made, the change was greater than 0.5 litre min(-1), on 204 occasions. The direction of change was the same with both methods in 199. The difference between the methods remained near zero during surgery suggesting that a single calibration per patient was adequate. Aortic model simulation with radial artery pressure as input reliably monitors changes in cardiac output in cardiac surgery patients. Before calibration, the model cannot replace thermodilution, but after calibration the model method can quantitatively replace further thermodilution estimates.

Lack of agreement between thermodilution and carbon dioxide-rebreathing cardiac output

BACKGROUND

A continuous, accurate, non-invasive monitor of cardiac output would represent a major step forward in patient management. A cardiac output computer, NICO2, based on the Fick principle and an automatic partial carbon dioxide (CO2)-rebreathing technique has just become available. We compared the performance of this monitor with the standard thermodilution method.

METHODS

Thirty patients were investigated after cardiac surgery. Replicate measurements were performed simultaneously with the thermodilution and NICO2 techniques. An Altman-Bland analysis was used to assess repeatability of each of the two methods and to determine the agreement between the two techniques.

RESULTS

The repeatabilities of thermodilution and CO2-rebreathing cardiac output were excellent, with coefficients of repeatability of 0.35 l/min and 0.60 l/min. Mean thermodilution and NICO2 cardiac output were 4.4 l/min (SD 0.9, range 2.7-6.1) and 4.6 l/min (SD 1.3, range 1.6-6.9). A comparison of the methods, however, revealed excessive limits of agreement (+/-1.80 l/min).

CONCLUSION

The agreement between the NICO2 derived cardiac output and the de facto standard - thermodilution cardiac output - is poor. The methods are not interchangeable with the present version of the NICO2. The repeatability of the partial CO2-rebreathing technique holds promise that a sufficient accuracy may be obtained by suitable modifications of the monitor's algorithms.

Hemodynamic-induced changes in aortic valve area: implications for Doppler cardiac output determinations

Monitoring cardiac output (CO) by transesophageal echocardiography involves measurements of ascending aortic flow and an initial measurement of aortic valve area (AVA). Hemodynamic-induced changes in AVA are a potential source of error for this simplified method. Our goal was to quantify these changes in AVA and their effects on CO calculations. In 17 anesthetized patients, a dobutamine infusion was titrated to achieve a 50% increase in ascending aortic flow velocity (V(max)). Hemodynamic and echocardiographic variables, including V(max) and planimetry of AVA, were determined at baseline and at maximal dobutamine dose. Dobutamine produced a 3.0 +/- 1.4 L/min increase in CO, a 54.5% +/- 19.6% increase in V(max), and a 50.6% +/- 34.2% increase in systolic blood pressure. AVA increased by 4.3% +/- 2.6% during dobutamine infusion (P < 0.001). The simplified CO method, which does not account for increases in AVA, produced a 0.32 +/- 0.24 L/min underestimation of CO. This investigation demonstrates hemodynamic-induced changes in AVA. The use of a single AVA measurement for all subsequent CO calculations introduces a clinically acceptable degree of error, supporting a simplified CO protocol requiring less probe manipulation and reduced procedural time.

IMPLICATIONS

An intraoperative dobutamine infusion was used to increase aortic blood flow and demonstrate hemodynamic-induced changes in aortic valve area. These valve-area changes affect the accuracy of Doppler cardiac output determinations.

Is the placement of a pulmonary artery catheter still justified solely for the measurement of cardiac output?

OBJECTIVE

The authors compared four clinical techniques of measuring cardiac output (CO) in critically ill patients: pulmonary artery thermodilution (CO[PA]), transpulmonary aortic thermodilution (CO[AORTA]), Fick principle-derived (CO[FICK]), and continuous pulmonary artery (CCO) measurements.

DESIGN

Prospective clinical study.

SETTING

Surgical intensive care unit of a university hospital.

PARTICIPANTS

Twelve adult patients suffering from sepsis or septic shock.

INTERVENTIONS

All patients were deeply sedated and mechanically ventilated in a pressure-controlled mode. Each patient received a 7.5F five-lumen pulmonary artery catheter for the continuous measurement of cardiac output and a 4F aortic catheter with an integrated thermistor. The thermistors of the two different catheters were connected to one computer system (COLD-Z021, Pulsion Medical Systems, Munich, Germany). Whole-body oxygen consumption was measured by indirect calorimetry using a metabolic cart (Deltatrac, Datex-Engstroem, Helsinki, Finland) over a 5-minute period, at the end of which arterial and mixed venous blood gases were taken and measured by co-oximetry. During each measuring period, three bolus CO measurements were performed. A total number of 51 CO measurements was analyzed.

RESULTS

Linear regression analysis revealed the highest correlation between CO(AORTA) and CO(PA) (r = 0.98), whereas agreement between these two techniques and CCO was lower (r = 0.92 and r = 0.93). All three techniques correlated comparably with CO(FICK) (r = 0.85, r = 0.83, and r = 0.83).

CONCLUSION

The correlations among the four CO techniques were high and similar, with CO(PA) and CO(AORTA) techniques showing the highest agreement. Because CO with similar accuracy can be obtained from transpulmonary aortic thermodilution in a less-invasive manner, it appears that the placement of a pulmonary artery catheter solely for the measurement of CO is no longer justified, unless continuous CO measurements are needed.