Statistical test of agreement between measurements in method-comparison study

In clinical settings, researchers often want to assess agreement between two measurements (or tests) of the same continuous variable. For example, when new point-of-care analyzer for testing blood glucose level were introduced clinicians need to compare results from standard or established laboratory method of measurement to those of new or point-of-care analyzer. The question in a method-comparison study would either of two different methods be used to measure the same variable equivalently. In this paper common misuse of statistical methodologies seen in the medical literatures such as correlation coefficient and paired t-test are discussed. The Bland-Altman technique has been widely used for this purpose and provides a graphic in presentation of the findings from a method-comparison study, with a mean value of measurement, this bias and the limits of agreement. For ease of application and interpretation of this technique we discussed the analysis procedure and illustrated with two worked examples. Finally, a number of alternative ways in which data can be analysed and reported in such studies were reviewed.