Abstract: Using Monte Carlo simulation to compare different test approaches, the results showed that Dunnett-Gent method was the best one of them both in type I error controlling and test power, and the statistical powers are very lower for all of 10 equivalence test methods when test group and reference group has only 100 cases.When establishing equivalence, 100 cases for each group suggested by State Food and Drug Administration (SFDA)is insufficiency, the required sample size should be estimated by statistical procedure.

Key words: test of equivalence on proportions, power, sample size