Statistic analysis in tests for equivalence

As is the case for any confirmatory statistical analysis, in assessing the significance of the results of an equivalence trial the use of testing procedures which are optimal with respect to power has much to recommend it. For three standard situations very frequently encountered in medical research, a concise account of such an optimal test for equivalence is given. Furthermore, it is pointed out that tests based on the well-known principle of confidence interval inclusion are valid in the sense of guaranteeing the significance level, but tend to have unnecessarily low efficiency.