Comparing different methods of cardiac output determination

EDITOR: We read the article by Dr Bajorat and colleagues reporting their results with different methods of cardiac output (CO) determination with great interest [1]. The authors should be commended for this study having for the first time compared a broad array of clinically used methods of CO determination to a true reference standard, i.e. aortic transit time ultrasound. However, though their approach is ambitious, some methodological remarks are necessary. Firstly, the authors suggest that pulmonary arterial and transpulmonary thermodilution can be used interchangeably for CO determination even under acute haemodynamic changes. This conclusion is not supported by the data presented. Following the introduction of Bland Altman plots for method comparison in 1986 [2], for more than a decade the judgement of bias and limits of agreement was left to the clinician, and identical values were interpreted differently. The pivotal work by Critchley for the first time suggested a comprehensive mathematically derived criterion for assessment of observed variability [3]. Given an inherent variability of ±20% for each method under comparison, the combined variability (i.e. limits of agreement) should not exceed ±30% of the mean CO. Applying these strict criteria to the data, only transpulmonary thermodilution (defined as high flow conditions of CO > 4L min−1) was really interchangeable with the reference method. Secondly, the authors calculated a trend score to determine if the different methods investigated track changes of the instantaneous CO consistently. Despite the fact that this score does not account for a situation when one method yields an increasing or decreasing CO, respectively, while the other method yields an unchanged reading, the fruitful discussion following a publication reporting on the comparison of pulse contour derived with pulmonary arterial derived CO measurements suggests analysis of the change in CO after a specific intervention also with a Bland Altman plot, comparing the mean percent change of both methods against the difference [4]. This analysis reveals quickly if methods under comparison do track ensuing changes of CO in a comparable fashion. Thirdly, in this regard it would be of great interest how the pulse contour derived CO performed during the various interventions reported. Since the authors used the PiCCO® for transpulmonary thermodilution, these values could have been easily obtained. Unfortunately, there are no data provided with regard to the pulse contour CO throughout the study period. This is quite surprising since data regarding the performance of pulse contour derived CO determination during rapid haemodynamic changes are currently still lacking. Finally, to facilitate comparison of different studies, we would suggest a consensus for data analysis and interpretation. Judgement of bias and limits of agreement should be based on Critchley's recommendations and the analysis of CO changes should also be accomplished with a Bland Altman plot.