关键词: 条件Ⅰ型错误, 条件检验效能, 国际多中心临床试验, Monte Carlo模拟

Abstract:

AIM: To explore the feasibility to adjust the significance levels α' in the target region as the basis of data in normal distribution for the primary efficacy endpoint, the reasonable value of adjusted significance levels α' and reasonable proportions of sample size in target region，and to make a reference for new drug approvals of Multi-Regional Clinical Trial (MRCT) in target region.  METHODS: By using Monte Carlo simulation, we studied the conditional false positive rate (CFPR) and conditional power (CP) changing with the proportions K of the sample size in target region by adopting different significance levels given that the treatment effect was shown to be significant under the significance level α=0.05 based on different sized MRCTs. RESULTS:Simulations results showed that CFPR and CP increased with the increasing significance level α' and the bigger K, the bigger CFPR. In addition, CFPR could be well controlled under 0.5 if K was no more than 30% and α' is no more than 0.5. Though f values 1.0, CP still kept low and was no more than 0.76 when K was no more than 15%. If f could be more than 0.8, CP would exceed 0.70 when K valued 20%. If f can be more than 0.8, CP will exceed 0.75 when K values 25%. When K valued 30%, CP would exceed 0.80 as long as f≥ 0.7. Lastly, even though f valued only 0.5, CP would exceed 0.80 when K valued 50%. CONCLUSION: The method in this study is easy to understand and operate, and have a better performance especially when f≥0.7.

Key words:  conditional false positive rate, conditional power, multi-regional clinical trial, Monte Carlo simulation