Agreement between methods.

Abstract

Before new tests are implemented, it is important to compare their results with those of other measurement methods that are already in use. In the determination of this so-called agreement between methods, one may choose between several statistical approaches. The correlation coefficient is a popular approach to determine the agreement between measurement methods. It is easy to calculate, but has important limitations: it does not provide any information on the type of association and it is extremely sensitive to the range of values within the study. Finally, a correlation coefficient does not reveal whether any difference between two measurements is systematic or random. Therefore, it is highly preferable to use Bland-Altman plots instead, as these reveal both systematic and random errors. Bland-Altman plots are also preferable in case of repeated measurements and calibrations.