Abstract: To search the statistical approaches for evaluating and dealing with the center effects in multicenter clinical trial. Breslow-Day test was used for evaluating the differences of the virtual rating among the centers whose responses were binary, Cochran-Mantel-Haenszel test for dealing with the center effects whose responses were binary or ordinal, and logistic regression for evaluating and dealing with the center effects simultaneously. The results showed that Breslow-Day test is invalid in evaluating the center effects whose response is ordinal, and Cochran-Mantel-Haenszel test can not be used for dealing with the effects of other covariates meanwhile.

Key words: multi-center clinical trial, center effects, Breslow-Day test, Cochran-Mantel-Haenszel tests, logistic regression, ordinal logistic regression