Testing Limits Of Agreement

Over-compliance has been interpreted by both a BA LOA and an equivalence approach. Both methods are favourable to acceptance on the basis of a clinical decision; However, in the case of equivalentzane, this indication must be expressly indicated a priori (reference Wellek 9). Compliance limits assess the range within which some of the differences between the measurements fall. Often, the match between two food methods is assessed on the basis of the BA analysis, and the decision to determine whether compliance should be determined or not is based on a dependent sampling test (t-paired test or Wilcoxon matched pair test). This approach has not been endorsed by BA and its original document, which describes the method, does not refer to hypothesis tests regarding distortion. Rather, the first manuscript of Bland & Altman (reference Bland and Altman 1) states: “The distance between measurements without creating difficulties will be a matter of evaluation. Ideally, it should be defined in advance in order to assist in the interpretation of the comparison of methods and to choose the sample size. The BA method (reference Bland and Altman 1) includes the representation of the difference between the two methods from the mean of the two methods and the study of the mean distortion, the determination of the 95% CIs of the distortion and a trend in the distortion. The accuracy of limit values is rarely taken into account in the interpretation of diagram BA. The interpretation of the accuracy of the limit values includes the calculation and interpretation of the 95% CI of the upper and lower limits and is described in detail with an example in the first document Bland and Altman (reference Bland and Altman 1). Another reference on the website by Martin Bland (www-users.york.ac.uk/~mb55/meas/sizemeth.htm, accessed August 28, 2015) clearly shows the impact of the sample on these estimates and emphasizes the importance of considering not only the width of the LOA, but also the accuracy with which they were estimated. To compare the measurement systems with the Bland Altman method, the differences between the different measurements of the two different measurement systems are calculated, and then the mean and standard deviation are calculated.

The 95% “concordance limits” are calculated as the mean of the two minus and plus values 1.96 standard deviation. This 95 percent limit should contain the difference between the two measurement systems for 95 percent of future measurement pairs. Table 2 presents the results of the BA comparisons and equivalence tests for the simulated data in tabular form. Fig. 1 represents the BA diagrams with equivalence intervals and 90% AI of the difference. Figure 2 shows the CI plot. Figure 2 (a) shows a diagram ci with the x axis that shows the difference between the two means, as is the traditional approach of pharmaceutical studies. Figure 2(b) shows a diagram of CI, expressed relative to the mean absorption of I using 3×24HR; both diagrams (Figures 2 (a) and (b) are identical in interpretation and, in this case, the methods are equivalent if the 90% CI is within the specified equivalence range.

All equivalence methods show that the FFQ is considered equivalent to 3×24HR only if the equivalence margin is set at 10% of the average of 3×24HR (12.24 μg) or alternatively at 15 μg. . . .

Comments are closed.