Clinical agreement between two quantitative measurements on a group of subjects is generally assessed with the help of the Bland-Altman (B-A) limits. These limits only describe the dispersion of disagreements in 95% cases and do not measure the degree of agreement. The interpretation regarding the presence or absence of agreement by this method is based on whether B-A limits are within the pre-specified externally determined clinical tolerance limits. Thus, clinical tolerance limits are necessary for this method. We argue in this communication that the direct use of clinical tolerance limits for assessing agreement without the B-A limits is more effective and has tremendous merits. This nonparametric approach is simple, is robust to the distribution pattern and outliers, has more flexibility, and exactly measures the degree of clinical agreement. This is explained with the help of two examples, including setups where clinical tolerance limits can be set up to follow varying trends if required in the clinical context – a feature not available in the B-A method.
Indrayan, Abhaya Prof, "A simple and robust alternative to Bland-Altman method of assessing clinical agreement" (January 2022). COBRA Preprint Series. Working Paper 121.