USCOM (Ultrasonic Cardiac Output Monitors) lacks agreement with thermodilution cardiac output and transoesophageal echocardiography valve measurements.
The USCOM (Ultrasonic Cardiac Output Monitors) device is a non-invasive cardiac output monitor, which utilises transaortic or transpulmonary Doppler flow tracing and valve area estimated using patient height to determine cardiac output. We evaluated USCOM against thermodilution cardiac outputs and transoesophageal echocardiography valve area measurements in 22 ASA PS4 cardiac surgical patients. Data collection commenced following pulmonary artery catheter insertion, with cardiac output measurements repeated after sternotomy closure. Failure to obtain transaortic Doppler readings using USCOM occurred in 5% of planned measurements. USCOM transaortic analysis was not planned for 11 patients with known aortic disease. Bias at the aortic window (n=20) was -0.79 l/min with limits of agreement from -3.66 to 2.08 l/min. At the pulmonary window, failure to obtain Doppler readings occurred in 24% of planned measurements. Bias at the pulmonary window (n=36) was -0.17 l/min with limits of agreement from -3.30 to 2.97 l/min. The USCOM estimates of valve area based on height showed poor correlation with the echocardiographic measurements of aortic and pulmonary valves (r=0.57 and r=0.17, respectively). It was concluded that USCOM showed poor agreement with thermodilution. The estimated valve area was identified as one source of error.
Key Words: non invasive cardiac output, USCOM, aortic valve area, pulmonary valve area, Doppler, flow velocity
Thermodilution determination of cardiac output using a pulmonary artery catheter is generally accepted as the clinical reference standard, however there are significant risks associated with its use. Various non-invasive monitors have been devised to minimise risks associated with cardiac output measurement, but each of these has limitations in terms of accuracy and clinical suitability (1) such that there is no one device which suits all patients.
Recently, an Australian company, Ultrasonic Cardiac Output Monitors (USCOM[R]) has licensed a device for measuring cardiac output based on transthoracic Doppler ultrasound that is totally non-invasive. Several studies using the monitor have been published (2-5), but only two evaluated it against other methods of cardiac output measurement in humans (3,5).
The present study was designed to compare USCOM derived cardiac output measurements from the aortic and pulmonary valve ultrasound windows with thermodilution cardiac measurements. Valve planimetry was planned using transoesophageal echocardiography (TOE) to test the validity of the USCOM biometric assumptions.
MATERIALS AND METHODS
Following approval by the Royal Adelaide Hospital Ethics Committee, informed consent for inclusion in the study was obtained from patients scheduled for elective cardiac surgery where the anaesthetic plan included TOE and a pulmonary artery catheter (PAC). Anaesthesia was induced with midazolam (0.05 to 0.1 mg/kg), fentanyl (1 to 2 mg) and rocuronium or pancuronium, and after intubation of the trachea, was maintained using end-tidal isoflurane concentrations of between 0.25 to 0.75% in an oxygen and air mixture. Mean arterial blood pressure was maintained at above 70 mmHg by infusion with adrenaline or noradrenaline or metaraminol bolus and intra-aortic balloon counterpulsation was continued if present. A balloon-tipped, double-thermistor catheter (CritiCath[TM] SP5507H, Becton Dickinson, Singapore) was positioned in the pulmonary artery. Thermodilution and USCOM cardiac output measurements were commenced simultaneously in the operating theatre immediately after PAC placement and again after closure of the chest. In the event that signal quality was insufficient for analysis using either method, a third set of measurements was taken after stabilisation in the intensive care ward. A period of 10 minutes was allowed for each measurement set, during which changes in patient position or haemodynamically active drugs were avoided. All measurements were done with the patient in the supine position. The USCOM operator was blinded to the thermodilution measurements until all traces had been analysed and recorded.
Thermodilution cardiac output was measured five times by injecting 10 ml of iced 5% dextrose into the right atrium after flushing the line. Observation of a smooth characteristic temperature waveform following injection was used as a marker of adequate signal quality. Cardiac output was calculated and averaged automatically (S/5 Anaesthetic Monitor, Datex-Ohmeda, Madison, U.S.A.).
Using USCOM methodology, cardiac output was measured at both aortic and pulmonary valves. For measurements at the aortic window the 2.2 MHz continuous wave Doppler probe was placed in the suprasternal notch or above the clavicle, and directed at the aortic valve. Once characteristic wave-forms with maximal peak velocity and minimal background noise were obtained, they were frozen on the screen and saved. The acoustic image quality of the waveform was graded on a scale of one to six using the Fremantle image score criteria, where six represents the maximum quality (4). A minimum image score of four was required for data inclusion. Using the Touchscreen[R] method incorporated in the device, in which six markers were placed on the screen defining the wave, heart rate (HR) and velocity time integral (VTI in cm) were calculated from the Doppler trace. Cardiac output was calculated using built-in USCOM software, estimating the diameter of the aortic annulus ([d.sub.AA] in cm) by a formula derived by Nidorf (6), based on height in cm:
Aortic valve area ([A.sub.AV] in [cm.sup.2]) was calculated as: [A.sub.AV]=0.785x[d.sup.2.sub.AA]
and cardiac output ([CO.sub.AV]) in l/min was obtained by the formula:
[CO.sub.AV] = VT [I.sub.AV] x [A.sub.AV] x HR / 1000
For cardiac output measurement at the pulmonary valve, the Doppler probe was placed parasternally in the second, third or fourth left intercostal space and directed at the pulmonary valve until the waveform was optimal. To obtain an estimate for pulmonary annulus diameter ([d.sub.PA] in cm) the USCOM used the formula:
and further calculations proceeded as above. All USCOM measurements were done in triplicate, accepting the measurement with the best trace optically, or when quality was equal the one with the largest stroke volume.
Between induction of anaesthesia and cardiopulmonary bypass, a dedicated anaesthetist completed an echocardiographic study, including measurements of the functional aortic valve area by direct planimetry and equilateral triangle-shaped planimetry of the aortic orifice in mid systole (using the aortic valve short axis view), left ventricular outflow tract aortic valve diameter (using the aortic valve long axis view) and pulmonary valve diameter (using the right ventricular inflow-outflow view), all in triplicate.
For each cardiac output method, repeatability coefficients (standard deviation of the percentage errors of all replicates) were calculated and are presented as a percentage. There is a 95% probability that the within-method measurement of cardiac output will fall within this percentage of the mean of a set of steady state measurements.
For clinically acceptable agreement between the two methods to be confirmed, the study protocol required a systematic error (bias) of less than 0.5 l/min and for 95% of observations by one method (limits of agreement, or precision (7)) to fall within [+ or -] 1.0 l/min of the other. This is consistent with previous studies comparing monitors of cardiac output (8) and is equivalent to a precision of +/-20% for a cardiac output of 5 l/min. The agreement between different methods of cardiac output measurement was tested by Bland-Altman analysis (7). Assuming that the differences between the two methods were normally distributed, with a variance [[sigma].sup.2], the Null hypothesis ([H.sub.0]: [sigma] [less than or equal to] 0.5 l/min, corresponding to limits of agreement of [+ or -] 1.0 l/min) for n pairs of observations was tested by computing [chi square]:
[chi square] = (n - 1)[s.sup.2]/[[sigma].sup.2]
in which s was the standard deviation of differences between two methods in our sample. P values were computed by one-tailed [chi square]-testing with (n - 1) degrees of freedom and tested against a critical value of [alpha] [less than or equal to] 0.05. Power analysis indicated a 95% power of correctly rejecting the null hypothesis if s >0.63, with 20 observations.
Accepting TOE as the gold standard for measuring cardiac dimensions, valve areas were compared with calculated USCOM values using two-tailed Pearson's correlation (product moment correlation) to indicate whether statistical agreement between the two measurements was present. Statistical significance was set at [alpha] [less than or equal to] 0.05.
Between August and December 2004, 22 patients (13 male, nine female) with ages ranging from 24 to 85 years (median 70) and height varying from 152 to 190 cm (median 170 cm) were enrolled. In our practice, pulmonary artery catheters were only placed when a complicated course was anticipated (e.g. poor left ventricular function, or difficult surgery). All patients were classified as ASA 4 and two patients had an intra-aortic balloon pump. Scheduled operations were coronary artery bypass grafting (14), aortic valve replacement (11) and mitral valve surgery (4), including seven combined procedures. In two patients echocardiography was not available.
Pulmonary artery catheter placement was uneventful in all patients. In one instance, the catheter was withdrawn on bypass and not repositioned because of surgical considerations. One patient died on bypass. The repeatability coefficient of 46 thermodilution CO measurements was 6.3%.
USCOM at aortic window
Patients with known aortic valve disease were excluded from transaortic analysis, but were considered suitable for pulmonary valve analysis. In the remaining 11 patients, valid aortic Doppler traces were obtained in 19 of 20 scheduled pre- and post-bypass measurements (95%); the one failure was successfully repeated four hours postoperatively. All included waveforms were of Fremantle image score criteria five or six. The repeatability coefficient for triplicates of 20 transaortic measurements was 7.2%.
Comparison with thermodilution (Figures 1 and 2) showed a bias of -0.79 l/min and a standard deviation of differences between methods of 1.43 l/min (Table 1). The hypothesis that precision was within [+ or -] 1.0 l/min was rejected with [chi square] of 156.3 (P <0.001). Using only traces with the highest possible waveform score standard deviation of differences remained 1.43 l/min, and eliminating measurements of patients in atrial fibrillation resulted in a standard deviation of differences of 1.41 l/min (Table 1).
Echocardiographic data were available for nine of the 11 patients fit for transaortic analysis. Pearson's correlation coefficient was calculated for aortic valve area calculated by USCOM and valve area using direct planimetry (r=0.57, P=0.052), area of left ventricular outflow tract (r=0.413, P=0.207) and aortic valve triangle planimetry (r=0.48, P=0.132).
Entering the available planimetric valve areas into the equation used by USCOM to calculate cardiac output resulted in 16 paired samples. In comparison with thermodilution, the bias increased to -0.54 l/min, but standard deviation decreased to 0.96 l/min, resulting in limits of agreement from -2.46 to 1.37 l/min and [chi square] of 55.0 (P <0.001), with rejection of [H.sub.0].
USCOM at pulmonary window
Valid Doppler waveforms were obtained in 32 of 42 scheduled measurements (76%). There were four failed measurements pre-bypass versus six post-bypass: four failed measurements were successfully repeated four hours postoperatively. All but one of the waveforms were of Fremantle image score criteria five or six. The repeatability coefficient for triplicates of 36 transpulmonary measurements was 9.7%.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
Comparison with thermodilution (Figures 3 and 4) showed a bias of -0.17 l/min and a standard deviation of differences of 1.57 l/min (Table 1). [H.SUB.0] was rejected with [chi square] of 354.2 (P <0.001). Using only traces with the highest possible waveform score standard deviation increased to 1.80 l/min and eliminating measurements of patients in atrial fibrillation resulted in a standard deviation of 1.55 l/min (Table 1).
Due to calcifications of the aortic root, visualisation of the pulmonary trunk by TOE was problematic in seven patients. Good quality views were obtained from the remaining 13 patients, with pulmonary valve diameters ranging from 1.73 to 2.48 cm (1.94 to 2.35 cm as calculated by USCOM). Correlation coefficient between calculated and measured valve diameters was 0.17 (P=0.572).
Entering the available TOE pulmonary valve diameters into the equation resulted in 23 paired samples. in comparison with thermodilution, the bias was 0.24 l/min and standard deviation was 1.40 l/min, resulting in limits of agreement from -2.56 to 3.04 l/min (Figure 4) and [chi square] of 172.5 (P <0.001), with rejection of [H.sub.0].
The main finding of this study was lack of agreement between thermodilution and USCOM cardiac output measurements. Both the transaortic and the transpulmonary approaches resulted in precision limits well outside [+ or -] 1.0 l/min, which was the standard set before the start of this study. Correcting the USCOM estimated valve areas for the actual valve areas measured by TOE improved the agreement with thermodilution slightly at both the aortic and pulmonary windows. This identified the estimated valve areas as contributing, but not the only sources of error.
Although a true 'gold' standard for cardiac output determination is lacking, thermodilution and the metabolic method based on the Fick principle have most often been the standard to which newer techniques were compared (1). Both techniques have limitations. Thermodilution has been noted to be influenced by homogeneity of the injection and mechanical ventilation (9). Though the use of larger volumes (10 ml or more) and iced injectate did not introduce systematic differences, both have been shown to reduce the variability of the measurements when compared with smaller volumes and room temperature injectates (10,11). It is noteworthy that these two standard methods have shown considerable disagreement in two separate studies (9,12). To compare a new method with an existing method in the absence of a gold standard, Bland and Altmann described the approach that was followed in this study, using the bias and precision calculations as described above. With precision limits of [+ or -] 1.0 l/min, a cardiac output of 5.0 l/min by one method would predict with 95% certainty a result by the other method between 4.0 and 6.0 l/min. For clinical decision making, a higher level of agreement between the 'old' and 'new' methods might be desirable, but in most studies comparing cardiac output devices a cut-off of [+ or -] 1.0 l/min has been used (8). Two earlier studies that compared USCOM with thermodilution (3,5) showed limits of agreement of [+ or -] 1.60 l/min and [+ or -] 2.02 l/min respectively. Without specifying their criteria for acceptance, both authors described the observed level of agreement as good. From a clinical perspective however, precision levels in this range are associated with considerable variance between methods.
A recent survey suggested that it took emergency physicians with no prior ultrasonographic experience only 20 measurements to become reliable USCOM operators (3). Subjectively, our experience with the device was less robust. Using USCOM involved blindly directing a Doppler beam at the aortic and pulmonary valves, and optimising flow traces by minute manipulations, aiming to obtain the highest peaks with the lowest background noise possible. However, the absence of any positive confirmation other than the shape of a curve, that the centre of the jet-stream has been scanned at a narrow angle, could undermine the operator's confidence, particularly at the pulmonary valve, where at least three windows need to be scanned. In the study by Tan et al (3), it required up to 45 min and frequent patient repositioning to obtain some of the transpulmonary traces. In our study, no more than 10 minutes for data acquisition and no patient repositioning were allowed, because that would be unrealistic in everyday practice. This may explain some of the difference in precision between the Tan et al (3) study and the present one. Because using the USCOM might be a skill-dependent technique, all measurements were done by the same operator. Even post-sternotomy, nearly all measurements at the aortic window were successful, but in 24% of transpulmonary attempts no suitable flow trace could be obtained. This corresponded to the 25% failure rate in supine position reported by Tan et al (3). in that study, a scoring system for waveform quality was introduced, which we applied to our traces post hoc. Limiting the sample to measurements with the highest score did not improve the test performance in either of the two windows. In addition, we found that irregular heart rhythms, such as atrial fibrillation or multiple extrasystoles made use of the algorithm cumbersome. In the validation study done by Knobloch et al (5), only patients in sinus rhythm were included for that reason. In our study however, elimination of patients in arrhythmia did not improve the precision. This latter problem has been addressed by the company and later models of USCOM are fitted with a flow-tracer, which automatically averages stroke volume.
The estimated valve dimensions used by USCOM might be an important source of error. USCOM has adopted a regression equation that was derived by Nidorf relating height to aortic annulus diameter in a cohort of 268 healthy volunteers (6). According to this equation and to the USCOM algorithm, a patient with a height of 170 cm would have an aortic annulus of 1.95 cm diameter. However, applying the confidence limits of Nidorf's regression analysis, the 95% confidence interval would range from 1.65 to 2.20 cm. The calculated aortic valve area for this patient would be 2.98 [cm.sup.2], with a 95% confidence interval ranging from 2.14 to 3.80 [cm.sup.2]. Thus, in theory, the accuracy of cardiac output measurements based on these estimates can not be expected to exceed [+ or -] 40%. Use of the Pearson correlation coefficient in this study indicated that both aortic and pulmonary valve areas calculated by USCOM did not correlate with measured values at a statistically significant level. Furthermore, the extrapolation of the linear relation between height and aortic diameter presumed in healthy individuals to other patient groups might not always be valid. in the present analysis all patients had undergone diagnostic cardiac work-up and measurements using transaortic flow tracing were not performed when significant aortic disease was present. Despite this, aortic valve dimensions correlated poorly with the USCOM estimates. Theoretically, the impact of unknown mild to moderate aortic stenosis and/or regurgitation on the Doppler trace could be large and could form a further source of error of this method that cannot be accounted for without prior knowledge of the aortic valve condition. The authors would suggest that an attempt be made to compare both aortic and pulmonary cardiac output figures in patients with unknown cardiac valvular status if this device is to be used: marked discrepancy between readings could indicate valvular pathology and a need for caution in interpreting output readings. This hypothesis was not tested within our study. Regarding the pulmonary valve, it is not really known whether any statistical relationship between pulmonary diameter and height exists. The formula USCOM has developed for cardiac output measurement at the pulmonary window has been neither independently established nor validated. Instead, the formula was derived by multiplying the formula for aortic valve diameter by a factor 1.1. This ratio resulted from comparing pulmonary and aortic flow traces of 39 adult and paediatric subjects (unpublished data, described in an abstract on the company website) (13). Though the number of patients was small (13) the pulmonary valve measurements from the TOE examinations in the present study showed variation over a much wider range than the USCOM estimates and statistical analysis did not suggest a linear correlation with height. The presence of subclinical aortic valvular disease could further limit the usefulness of transaortic USCOM valve area estimations.
While the USCOM monitor is non-invasive, its use may still lead to inappropriate clinical interventions if it does not monitor cardiac output reliably. Where a reliable waveform can be repeatedly obtained from a single patient, it is possible that the trend data would be helpful in assessing the response to interventions aimed at improving cardiac output, if the absolute values are not critical. This was not assessed in our study, but warrants further investigation.
In conclusion, this study showed considerably more disagreement between transpulmonary USCOM measurements and thermodilution than previous studies (3,5). Before the transpulmonary method can be relied on clinically, it would be recommended that the algorithm for pulmonary valve diameter receive thorough evaluation. In this study, the pulmonary valve dimensions as measured by TOE lacked correlation to the USCOM estimates and therefore to patients' height. Measurements at the aortic window agreed slightly better with thermodilution, especially when corrected for aortic valve area. However, there appear to be some fundamental limitations to the use of the present linear formula for aortic annulus diameter for the purpose of determining cardiac output.
We would like to thank the Royal Adelaide Hospital Department of Cardiothoracic Surgery for facilitating this research, Dr Dave Jarvis for his assistance with statistical analysis and USCOM Australia for providing operator training and a loan USCOM cardiac output monitor for the duration of the study.
Accepted for publication on July 18, 2007.
(1.) Chaney JC, Derdak S. Minimally invasive hemodynamic monitoring for the intensivist: current and emerging technology. Crit Care Med 2002; 30:2338-2345.
(2.) Critchley LA, Peng ZY, Fok BS, Lee A, Phillips RA. Testing the reliability of a new ultrasonic cardiac output monitor, the USCOM, by using aortic flowprobes in anesthetized dogs. Anesth Analg 2005; 100:748-753.
(3.) Tan HL, Pinder M, Parsons R, Roberts B, van Heerden PV. Clinical evaluation of USCOM ultrasonic cardiac output monitor in cardiac surgical patients in intensive care unit. Br J Anaesth 2005; 94:287-291.
(4.) Dey I, Sprivulis P. Emergency physicians can reliably assess emergency department patient cardiac output using the USCOM continuous wave Doppler cardiac output monitor. Emerg Med Australas 2005; 17:193-199.
(5.) Knobloch K, Lichtenberg A, Winterhalter M, Rossner D, Pichlmaier M, Phillips R. Non-invasive cardiac output determination by two-dimensional independent Doppler during and after cardiac surgery. Ann Thorac Surg 2005; 80:1479-1483.
(6.) Nidorf SM, Picard MH, Triulzi MO, Thomas JD, Newell J, King ME et al. New perspectives in the assessment of cardiac chamber dimensions during development and adulthood. J Am Coll Cardiol 1992; 19:983-988.
(7.) Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-310.
(8.) 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.
(9.) Dhingra VK, Fenwick JC, Walley KR, Chittock DR, Ronco JJ. Lack of agreement between thermodilution and Fick cardiac output in critically ill patients. Chest 2002; 122:990-997.
(10.) McCloy K, Leung S, Belden J, Castenada J, Erickson V, Koch K et al. Effects of injectate volume on thermodilution measurements of cardiac output in patients with low ventricular ejection fraction. Am J Crit Care 1999; 8:86-92.
(11.) Williams JE, Jr., Pfau SE, Deckelbaum LI. Effect of injectate temperature and thermistor position on reproducibility of thermodilution cardiac output determinations. Chest 1994; 106:895-898.
(12.) Gonzalez J, Delafosse C, Fartoukh M, Capderou A, Straus C, Zelter M et al. Comparison of bedside measurement of cardiac output with the thermodilution method and the Fick method in mechanically ventilated patients. Crit Care 2003; 7:171-178.
(13.) Phillips R, West M, Burstow D. A 2D independent, CW Doppler echocardiographic algorithm for determination of transpulmonary and transaortic flow volumes in adults and children. Asia Pacific Congress of Doppler and Echocardiography Jan 13 2004. (Abstract).
H. L. A. VAN DEN OEVER *, E. J. MURPHY [[dagger]], G. A. CHRISTIE-TAYLOR [[double dagger]]
Department of Anaesthesia, Royal Adelaide Hospital, Adelaide, South Australia, Australia
* M.D., Anaesthesia Fellow.
[[dagger]] M.B., B.S., F.A.N.Z.C.A., Specialist Anaesthetist.
[[double dagger]] M.B., B.Ch., F.A.N.Z.C.A., Specialist Anaesthetist.
Address for reprints: Dr E. J. Murphy, Department of Anaesthesia, Royal Adelaide Hospital, North Terrace, Adelaide, S.A. 5000.
TABLE 1 Bias and limits of agreement: transaortic and transpulmonary USCOM compared with thermodilution cardiac output measurements USCOM site Selection applied Number of Bias observations (l/min) Aortic window None 20 -0.79 Waveform score=6 16 -0.82 Regular rhythm 16 -0.82 Pulmonary None 36 -0.17 window Waveform score=6 21 -0.15 Regular rhythm 29 -0.21 USCOM site Selection applied Standard deviation Upper limit (l/min) of agreement Aortic window None 1.43 -3.66 Waveform score=6 1.43 -3.68 Regular rhythm 1.41 -3.64 Pulmonary None 1.57 -3.30 window Waveform score=6 1.80 -3.74 Regular rhythm 1.55 -3.31 USCOM site Selection applied Lower limit of agreement (l/min) Aortic window None 2.08 Waveform score=6 2.05 Regular rhythm 2.00 Pulmonary None 2.97 window Waveform score=6 3.44 Regular rhythm 2.89
|Printer friendly Cite/link Email Feedback|
|Author:||Van Den Oever, H.L.A.; Murphy, E.J.; Christie-Taylor, G.A.|
|Publication:||Anaesthesia and Intensive Care|
|Article Type:||Clinical report|
|Date:||Dec 1, 2007|
|Previous Article:||Critically ill children in non-paediatric intensive care units: a survey, review and proposal for practice.|
|Next Article:||The preoperative detection of risk of anaphylaxis during anaesthesia.|