Method Comparison in Cardiovascular Trials Using the Difference Plot Method

Abstract

In cardiovascular trials investigators often want to compare two different measuring methods and decide on the equivalence of the two measurements. The differences between association and agreement are emphasised and the plot of difference against the average, ‘the difference plot method’, is demonstrated using a practical example. In addition, the interpretation of bias and limits of agreement are discussed and logarithmic transformation is suggested if the difference between the measurements by the two methods is related to the magnitude of the measurement. Finally, it is emphasised that no method will be appropriate in all circumstances and that care should be taken to ensure sufficiently large samples, to pre-specify the meaning of agreement and to avoid systematic variation, that can be caused by observer, time of day, position of subject etc. These are precautions valid to all method comparison studies irrespective of the statistical methods used in the final analyses.