存在共同终点的优效性临床试验样本量问题探讨  

Sample Size Considerations in Superiority Clinical Trials With Co-primary Endpoints

作　　者：吴振强[1,2] 李卫[1] 孙业桓[2] 王杨[1] 

机构地区：[1]中国医学科学院,北京协和医学院,国家心血管疾病中心,阜外心血管病医院,心血管疾病国家重点实验室,医学研究统计中心,100037 [2]安徽医科大学公共卫生学院

出　　处：《中国卫生统计》2013年第5期680-682,686,共4页Chinese Journal of Health Statistics

摘　　要：目的探讨当优效性临床试验存在多个共同终点时,各终点分别计算样本量取最大值的样本量计算方法的合理性。方法运用蒙特卡洛模拟的方法,验证实践中常用的多个主要终点分别计算取最大值的样本量计算方法的合理性;并进一步探讨主要终点间相关性与检验效能之间的关系。结果当多个终点分别计算的样本量相等时,分别计算取最大值的方法不能达到预期的检验效能(80%);当相关系数小于0时,检验效能在60%左右波动;大于0时,随着相关系数的增加,检验效能逐渐增大。结论当多个主要终点算得样本量相当时,取最大值的样本量计算方法并不适用于优效性临床试验;当多个主要终点算得的样本量差距较大时,可以达到试验所需的检验效能。因此,运用时需要考虑其适用条件。Objective To evaluate the method of selecting a maximum sample size of those obtained from testing individual primary endpoint. Methods Monte Carlo simulation was used to verify the power of this sample size calculated method. Furthermore, we assessed the effect of correlation coefficient among the co-primary endpoints. Results When the mean effect sizes are equal in individual primary endpoints, the study power cannot reach 80%. Study power fluctuates around 60% if the corre- lation coefficient is negative. However, the power increases with the incre- tion of correlation coefficient when the correlation coefficient is greater than 0. Conclusion The method of selecting a maximum sample size is un- der-power when the mean effect sizes are equal in individual primary end- point. When effect sizes largely varied, the study power is enough.

关 键 词：临床试验 样本量计算 检验效能 相关性 蒙特卡洛模拟 

分 类 号：R195.1[医药卫生—卫生统计学]