Abstract: AIM: To evaluate the effects on the test power and type I error of center effect, treatment-by-center interaction and imbalanced sample sizes among centers in a multi-center clinical trials. METHODS: We focused on binary outcome and limited our study on the comparison of two groups. 60 scenarios (10 allocated proportions, 3 types of center effects with and without main effect of treatment) were considered. Using Monte-Carlo simulation studies, we estimated the test power or type I error in each scenario. RESULTS: When there was no central effect or no treatment-by-center interaction, the simulated powers were approximately equal to 0.80 and the type I errors were well controlled at 0.05 in various scenarios. The imbalance of sample sizes among centers had no affected on the power and type I error in these scenarios. However, when there was an interaction between treatments and centers, the power decreased, and the type I error inflated. The imbalance of sample sizes among centers slightly affected the power and type I error. CONCLUSION: In a multicenter clinical trial, the treatment-by-center interaction has a large effect on the test power and type I error, while the imbalance of sample sizes among center has no or only a limited effect. Therefore, we should pay more attention to the interaction between center and treatment in designing a multi-center clinical trial.

Key words: Multi-center clinical trial, Sample balance, Center effect, Treatment-by-center interaction