In this single centre, observational study of 52 patients, diagnostic images and complete dataset were obtained on all participants. The mean CI values for AV CW-VTI and PiCCO were similar (2.70 L/min/m2 and 2.86 L/min/m2 respectively), with lower mean values identified for LVOT PW-VTI (2.33 L/min/m2). The mean difference (bias) between AV CW-VTI and PiCCO was − 0.16 L/min/m2 and between LVOT PW-VTI and PiCCO − 0.54 L/min/m2. This continued over a wide range of CIs suggesting that AV CW-VTI may offer a more accurate estimate than LVOT PW-VTI, using PiCCO as a reference standard, while LVOT PW-VTI underestimates CI as compared to both PiCCO and AV CW-VTI with a mean difference (bias) of -0.38 L/min/m2.
These results are similar to Evangelista [11] where CW mean values closely matched those obtained using PAC and PW underestimated CO as compared to PAC derived values (-28% +/-13%, P < 0.01).
There is no gold standard for clinicians to assess CI and the PAC remains the most widely used reference standard. PAC is rarely used so in this paper the PiCCO (thermodilution) was designated as the reference standard. PICCO as a reference device is open to question with a recent meta-analysis suggesting that the LVOT PW-VTI offered a better agreement:tolerability index and summary percentage errors to PAC than PICCO (using thermodilution) to PAC [21]. However, the paper did report a smaller spread of values for PICCO vs. PAC and using a reference device assists in the comparison between CI as assessed by AV CW-VTI and LVOT PW-VTI. The PAC (thermodilution) does not allow continuous CI monitoring, is influenced by ventilation, core temperature and valvular disease, and, traditionally, requires manual input. Previous work reported a PE of 25% for LVOT-VTI compared with PAC [10], and a bias of -0.2 L/min (CO not CI) with LOA − 1.2, 1.8 L/min, like this paper.
Most clinicians would not regard the AV CW-VTI difference of 0.16 L/min/m2 as clinically significant but may the 0.54 L/min/m2 mean difference between PiCCO and LVOT PW-VTI, as this value may be above the 10–15% increase in CI that defines a fluid responder to a 500ml fluid challenge. However, equally important to clinicians is the inter- and intra observer reliability, which were not addressed in this study. One possible explanation for AV derived CI values being closer to PICCO derived values is a more accurate tracing of red blood cells, the VTI “envelope”, using the CW doppler as opposed to using the gated PW doppler.
We showed a difference approaching statistical significance between the two methods in terms of their absolute mean difference from the PiCCO derived CI, p = 0.07. The absolute mean difference comparing AV CW-VTI CI to PiCCO CI was 0.47 ± 0.43 L/min/m2, as opposed to the absolute mean difference of 0.61 ± 0.41 L/min/m2 comparing LVOT PW-VTI CI to PiCCO CI. With no power calculation available the significance of this finding is open to debate. On average, the results again indicate CI values via the AV CW-VTI method to be closer to the PiCCO method CIs than those via the LVOT PW-VTI method.
There were two cases which were outliers where the LVOT PW-VTI CI values were considerably lower than the PiCCO CI, with differences being 1.80 and 1.70 L/min/m2. These are significant differences, and these outlier results may have impacted the accuracy of the PW derived method in overall differences from the PiCCO method.
In this study the spread of the values (standard deviations) were similar for all three methods, suggesting similar precision. The B-A limits for the AV CW-VTI and PiCCO measurements were approximately ± 1.2 L/m2 /min from the mean value. Thus, while the mean values for CI obtained suggest AV CW-VTI and PiCCO may be used interchangeably the wide LOA may not support this. 95% of the values obtained by AV CW-VTI will be within 1.2 L/m2/min of the mean value obtained by PiCCO, so potentially seeing the two methods classifying CI differently, one as normal and the other method as abnormal.
The 95% B-A LOA were slightly smaller for LVOT PW-VTI and PiCCO (± 1.0 L/m2 /min) than AV CW-VTI and PiCCO, indicating slightly increased precision of the PW method. However, due to the larger mean difference between LVOT PW-VTI and PiCCO, the B-A LOA are skewed towards lower values. The bottom range LOA suggests that the CI could be up to 1.5 L/m2 /min lower for the LVOT PW-VTI method than the PICCO derived assessment, risking misclassification of shock syndromes and instituting inappropriate care.
Previous authors have used a range of methods to assess and compare CI measuring devices. Accuracy (bias, the difference or systematic error between assessed techniques) and precision (the scatter or random error between techniques, LOA) are best expressed using B-A graphs [27]. Critchely and Critchely suggested acceptable LOA of 10–20% (errors consequent upon respiratory cycle variation in CI and lack of precision of device, with three to five averaged readings to minimise these), based on the assumed accuracy of PAC, or, 30%, based on the notional error of the two techniques under review [25]. The PE adjusts the LOA for CI and given the high proportion of low CI patients in this study enables wider comparison. It provides assessment of the changes in CI not linked to true changes in CI but to error within the two systems under comparison.
In the paper published today the PE were: 43.5% (AV CW-VTI – PICCO), 38.6% (LVOT PW-VTI – PICCO) and 46.0% (LVOT PW-VTI – AV CW-VTI). These are all higher than the recommended maximum value of 30% suggested by Critchely and Critchely in 1999. However, Peyton and Chong performed a meta-analysis in 2010 that concluded a limit of 30% was somewhat arbitrary and does not reflect in vivo PE in a range of invasive and non-invasive methods of CI monitoring devices. Their recommendation is that a PE of 45% may be more suitable threshold, which was met by all devices in this study [26].