1. Initial description of the lithium dilution method and clinical trial comparing it with thermodilution [Linton RAF, Band DM, Haire KM: A new method of measuring cardiac output in man using lithium dilution. British Journal of Anaesthesia 1993; 71:262-266.]

(Copyright of this paper remains with the British Journal of Anaesthesia. We are grateful for permission to reproduce some material from this paper.)

This was the first clinical trial and therefore many of the details of the method were slightly different from those of the present system.

Objective

The objective of this study was to demonstrate the feasibility of using this new indicator dilution method of measuring cardiac output in man.

Study population

9 patients were studied in the immediate postoperative period after either coronary artery bypass grafting (7 patients) or aortic valve replacement.

Design

In each patient, each lithium sensor was used to make 2 LDCO measurements, one immediately before and one immediately after a set of thermodilution (TDCO) readings. Up to 3 sensors were tested in this way in each patient; results were obtained from 22 sensors. The average of the 3 closest thermodilution values was compared with the mean of the corresponding 2 LDCO measurements.
One of the disadvantages of this design was that measurements by the two methods were not simultaneous (although as close together as possible - usually within 1 min), so some temporal variation in cardiac output would have reduced the correlation between the two methods.

Evaluation

The results of the two methods were compared by Bland and Altman analysis. The mean of the differences (thermodilution - LDCO) = 295 ml/min with a standard deviation of the differences = 498 ml/min. This discrepancy between the two methods would have had three components: first, a true temporal variation in cardiac output (mentioned above), secondly an error due to the thermodilution method and thirdly an error due to the LDCO method. Stetz et al. (1982) analysed 14 reports of the reliability of thermodilution and concluded that there should be a minimum difference of 12-15% between the averages of triplicate measurements (and 20-26% between single measurements) for it to be considered that a true change in cardiac output had occurred. The figure shows the XY plot of LDCO against TDCO.

There are several potential sources of error in the LDCO method. The wrong dose of lithium chloride might be injected; this would produce a proportional error in the derived cardiac output. Lithium might have been lost between the point of injection (central vein) and the point of detection by the sensor. This could have been due to loss in the lungs, binding to protein or uptake by red cells. If such loss occurred, then the lithium dilution curves produced would have been too small and consequently cardiac output would have been overestimated. By adding aliquots of lithium chloride to whole blood in vitro we showed that there was no measurable reduction in plasma lithium concentration over a period of 30 min. In a further study in which lithium chloride was injected into the right or left atrium of patients and the curves compared, no significant loss of lithium was found. Another potential source of error is non-constancy of flow past the sensor, which might occur if the arterial catheter kinked. This would cause the recorded curve to deviate from the lognormal and this would be apparent on inspection of the curves.

Deviations from protocol

Results from 5 sensors were rejected either because of excessive drift or because they had a reduced span when calibrated after use. This failure rate would not be acceptable for a commercial product, but the design and reliability of the sensors have been improved since this study was carried out.

Adverse events

There were no adverse events. The dose of lithium chloride used was 3 to 6 times the currently used dose.

Comment

This initial study was carried out with an early prototype of the device. Since then improvements have been made in sensor design, the blood withdrawal system, signal processing and curve analysis.