Sample Size Determination for a Three-arm Biosimilar Trial

Abstract

The equivalence assessment usually consists of three tests and is often conducted through a three-arm clinical trial. The first two tests are to demonstrate the superiority of the test treatment and the reference treatment to placebo, and they are followed by the equivalence test between the test treatment and the reference treatment. The equivalence is commonly defined in terms of mean difference, mean ratio or ratio of mean differences, i.e. the ratio of the mean difference of the test and placebo to the mean difference of the reference and placebo. In this dissertation, the equivalence assessment for both continuous data and discrete data are discussed. For the continuous case, the test of the ratio of mean differences is applied. The advantage of this test is that it combines a superiority test of the test treatment over the placebo and an equivalence test through one hypothesis. For the discrete case, the two-step equivalence assessment approach is studied for both Poisson and negative binomial data. While a Poisson distribution implies that population mean and variance are the same, the advantage of applying a negative binomial model is that it accounts for overdispersion, which is a common phenomenon of count medical endpoints. The test statistics, power function, and required sample size examples for a three-arm equivalence trial are given for both continuous and discrete cases. In addition, discussions on power comparisons are complemented with numerical results.