Accuracy and precision of the USCOM: does a meta-analysis provide the answer?

Abstract

Chong and Peyton's recent meta-analysis of bias and precision studies on the clinical performance of the Ultrasonic Cardiac Output Monitor (USCOM Ltd., Sydney, Australia) 1 provides a percentage error for the device against thermodilution cardiac output of 42.7%, which is well outside the current benchmark for clinical acceptability set at 30% 2. However, the authors' discussion of this result is notable because, rather than accepting the obvious conclusion that the USCOM is simply too unreliable to be used clinically, they provide several arguments to excuse their poor result and conclude that further studies to investigate the tracking ability of the USCOM are needed. And yet their meta-analysis data in no way leads to such a conclusion. Perhaps an explanation lies in the first paragraph of their publication where they propose the need for devices that help to optimise haemodynamics, and Bland and Altman analysis may not the best way of investigating this property 3. If one looks more deeply into the precision of the USCOM one finds that there are two main sources of error: (a) that arising from the Doppler measurement of aortic blood flow, the velocity time interval (vti); and (b) that arising from the USCOM estimation of aortic valve cross-sectional area (CSA), which is derived from an algorithm based on patient size and has nothing to do with the use of Doppler. Thus, as no two individuals are similar in shape or size, this estimation of CSA accounts for much of the error found in USCOM cardiac output readings. Although important when absolute values are required, the CSA component remains constant and plays no part in the measurement error when tracking changes in cardiac output. Thus, do we really know how precisely the USCOM measures changes in aortic blood flow from Chong and Peyton's meta-analysis? As Chong and Peyton acknowledge, percentage errors between studies varied enormously. In liver transplant patients, Su et al. found percentage errors of 15%, whilst in cardiac surgery patients, Van den Oever et al. found it to be 62% 4, 5. Furthermore, Chong and Peyton questioned the experience of the person performing the USCOM scan 6. Unlike other minimally invasive cardiac output monitoring technologies, such as bioimpedance and pulse contour analysis, the USCOM is extremely operator dependant and substandard technique can lead to low and inaccurate readings, a point that Chong and Peyton make their paper. Unfortunately, the level of skill needed to obtain reliable USCOM scans appears to be very high and 20 examinations to gain competence as suggested by some authors 7 is, in our opinion, just the starting point on a long learning curve. Furthermore, patient factors such as a low cardiac output and old age also adversely affect one's ability to perform good quality USCOM scans. Our earliest impression of using the USCOM in elderly anaesthetised patients was that good quality scans were usually difficult to obtain. However, in a younger population of obstetric patients with much higher cardiac outputs my colleagues had no such complaints about the ease and quality of scans and have made good use of the USCOM as a clinical research tool 8. Recently, we completed a study that investigated the effect of age on USCOM readings following induction of anaesthesia. The quality of each USCOM scan was assessed using the six-point Fremantle score described by Dey and Sprivulis 7, where the presence of key features in the USCOM scan are scored. A Fremantle score above 4 is considered acceptable. We found that above the age of 60 years, over half of our patients had unacceptable USCOM scans based on Fremantle criteria (Fig. 3). Could the variation between studies found by Chong and Peyton be due to differences in patient populations and varying abilities to use a device that is clearly very operator dependant? We agree with Chong and Peyton's final conclusion that the USCOM needs its ability to track cardiac output over time properly evaluated, as simple comparisons against thermodilution cardiac output using Bland and Altman analysis does not give us the answers that we need.