Comparing methods of clinical measurement: Reporting standards for bland and altman analysis

In this era of medical technology assessment and evidence-based medicine, evaluating new methods to measure physiologic variables is facilitated by standardization of reporting results. It has been proposed that assessing repeatability be followed by assessing agreement with an established technique. If the 'limits of agreement' (mean bias ± 2SD) are not clinically important, then one could use two measurements interchangeably. Generalizability to larger populations is facilitated by reporting confidence intervals. We identified 44 studies that compared methods of clinical measurement published during 1996 to 1998 in seven anesthesia journals. Although 42 of 44 (95.4%) used the limits of agreement methodology for analysis, several inadequacies and inconsistencies in reporting the results were noted. Limits of agreement were defined a priori in 7.1%, repeatability was evaluated in 21.4%, and relationship (pattern) between difference and average was evaluated in 7.1%. Only one of the articles reported confidence intervals. A computer macro for the Minitab statistical package (State College, PA) is described to facilitate reporting of Bland and Altman analysis with confidence intervals. We propose standardization of nomenclature in clinical measurement comparison studies. Implications: A literature review of anesthesia journals revealed several inadequacies and inconsistencies in statistical reports of results of comparison studies with regard to interchangeability of measurement methods. We encourage journal editors to evaluate submissions on this subject carefully to ensure that their readers can draw valid conclusions about the value of new technologies.