Dmitrienko Alex, Offen Walt, Wang Ouhong, Xiao Dan
Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, IN 46285, USA.
Pharm Stat. 2006 Jan-Mar;5(1):19-28. doi: 10.1002/pst.190.
This paper discusses multiple testing procedures in dose-response clinical trials with primary and secondary endpoints. A general gatekeeping framework for constructing multiple tests is proposed, which extends the Dunnett test [Journal of the American Statistical Association 1955; 50: 1096-1121] and Bonferroni-based gatekeeping tests developed by Dmitrienko et al. [Statistics in Medicine 2003; 22:2387-2400]. The proposed procedure accounts for the hierarchical structure of the testing problem; for example, it restricts testing of secondary endpoints to the doses for which the primary endpoint is significant. The multiple testing approach is illustrated using a dose-response clinical trial in patients with diabetes. Monte-Carlo simulations demonstrate that the proposed procedure provides a power advantage over the Bonferroni gatekeeping procedure. The power gain generally increases with increasing correlation among the endpoints, especially when all primary dose-control comparisons are significant.
本文讨论了具有主要和次要终点的剂量反应临床试验中的多重检验程序。提出了一个用于构建多重检验的通用把关框架，该框架扩展了邓尼特检验[《美国统计协会杂志》1955年；50: 1096 - 1121]以及德米特里恩科等人开发的基于邦费罗尼的把关检验[《医学统计学》2003年；22:2387 - 2400]。所提出的程序考虑了检验问题的层次结构；例如，它将次要终点的检验限制在主要终点显著的剂量上。使用糖尿病患者的剂量反应临床试验说明了多重检验方法。蒙特卡罗模拟表明，所提出的程序比邦费罗尼把关程序具有检验功效优势。检验功效增益通常随着终点之间相关性的增加而增加，特别是当所有主要剂量对照比较都显著时。
