Numerous cardiac output (CO) monitors have been produced that provide continuous rather than intermittent readings. Bland and Altman has become the standard method for validating their performance against older standards. However, the Bland and Altman method only assesses precision and does not assess how well a device detects serial changes in CO (trending ability). Currently, there is no consensus on how trending ability, or trend analysis, should be performed. Therefore, we performed a literature review to identify articles published between 1997 and 2009 that compared methods of continuous CO measurement. Identified articles were grouped according to measurement technique and statistical methodology. Articles that analyzed trending ability were reviewed with the aim of finding an acceptable statistical method. Two hundred two articles were identified. The most popular methods were pulse contour (69 articles), Doppler (54), bioimpedance (38), and transpulmonary or continuous thermodilution (27). Forty-one articles addressed trending, and of these only 23 provided an in-depth analysis. Several common statistical themes were identified: time plots, regression analysis, Bland and Altman using change in CO (ΔCO), and the 4-quadrant plot, which used direction of change of ΔCO to determine the concordance. This plot was further refined by exclusion of data when values were small. Receiver operating characteristic curves were used to define the exclusion zone. In animal studies, a reliable reference standard such as an aortic flowprobe was frequently used, and regression or time plots could be used to show trending. Clinical studies were more problematic because data collection points were fewer (8-10 per subject). The consensus was to use the 4-quadrant plot with exclusion zones and apply concordance analysis. A concordance rate of >92% when using a 15% zone indicated good trending. A new method of presenting trend data (ΔCO) on a polar plot is proposed. Agreement was shown by the angle with the horizontal axis and ΔCO by the distance from the center. Trending can be assessed by the vertical limits of the data, similar to the Bland and Altman method.

When using bias and precision statistics, cardiac output, bias, limits of agreement, and percentage error should be presented. Using current reference methods, acceptance of a new technique should rely on limits of agreement of up to +/-30%.

No abstract