Sample Size Calculation For An Agreement Study

1. Bland JM, DG Altman. Statistical methods to assess the agreement between two methods of clinical measurement. Lancet 1986;i:307-310. If μ-0, the calculation of sample size can be written as follows: Recent studies of the agreement between two instruments or clinical trials in the ophthalmological literature have multiplied. McAlinden et al. used a method of calculating sample size for agreement studies based on the method proposed by Bland [19]. The sample size was calculated without taking into account the effectiveness of the statistical method, so the probability of obtaining the required width was only 0.50 [20]. During the study phase, consideration of performance in sample size calculations could result in expected conclusions below the pre-established level of performance. Cesana et al. provided another estimate of sample size that was needed to establish a pearson correlation coefficient between the differences and the means of the measurements [20], and we feel that this method is not appropriate.

Indeed, the correlation coefficient indicated by Cesana reflected proportional distortion. As we know, one of the hypotheses of application of the Bland-Altman method is not a proportional bias. In the absence of the hypothesis, this method would not be applicable. If you know exactly how to estimate the limits of the match, you can use it to determine the sample size. We set α-0.05, β 0.20, μ 0.4 0.4, δ -2.7 and predetermined power – 80%. Figure 2 shows the sizes and strengths of the B-A sample and the new method under different parameter parameters. For the Bland-Altman method, the sample size is calculated without taking into account the effectiveness of the statistical method, so the probability of obtaining the required width is only 0.50. With the new method, the resulting performance is generally close to 80% performance. The first important step in the Bland-Altman method is to draw the data and verify its model and distribution. The differences for both methods are presented against their means, and if the data behave well, then the construction of the different boundaries and the interpretation of the data are simple and simple. Assumptions about the limits of the agreement method are that differences resulting from two measures should have an approximate normal distribution, a constant variance of differences and proportional distortion [10].

There is proportional distortion when differences from average values increase or decrease [11].