## Example

The analysis of the repeated DBP measurements in the hypertension trial using a compound symmetry covariance pattern model gave a residual variance of a2 = 76 and the repeated measures had correlation p = 0.53 (see Section 6.3).

var j

We calculate the sample size required for a future study involving four post-treatment visits required to detect a difference in DBP of 5 mmHg at the 5% significance level with 80% power. The number of patients required per treatment group is n = 2(1.96 + 0.84)2x76 x (1 + 3x0.53)/(4x25) = 31.

Had no account been taken of the repeated measurements, then n would have been n = 2(zW2 + zpfo 2/A2 = 2(1.96 + 0.84)2 x 76/25 = 48.

If there is flexibility in the number of repeated measurements, then it might be worth calculating sample sizes for varying numbers of repeated measurements. For example, if 10 repeated measures were used, then n = 2(1.96 + 0.84)2 x 76 x (1 + 9x0.53)/(10x25) = 28.

However, this reduction in the number of patients required would be unlikely to justify the use of six additional repeated measures.