Method comparison: Statistical measurement correlation or agreement-most appropriate tool?

Abstract

In laboratory, a reliable method is very important to report precise and accurate results which are very important for clinical diagnosis and management decisions. In case the existing method needs to be replaced by new method due to poor results or financial issues, method comparison of a new measurement system/method with an established/referral method is required to see whether they agree sufficiently for the new to replace the old method or use the two interchangeably. Most commonly Pearson’s product-moment correlation is used to ascertain whether two methods agree sufficiently so that the old method can be replaced by a new method. However, it is the most inappropriate method with number of limitations. In this paper, other methods are discussed with their merits and demerits with examples.

Keywords

• method comparison; pearson’s product-moment correlation; bland–altman method; intra class correlation coefficient; clinical tolerance limit