Biostatistics and Programming, Sanofi-Aventis US, Bridgewater, NJ 08807, USA.
Stat Med. 2013 Jul 10;32(15):2529-43. doi: 10.1002/sim.5693. Epub 2012 Dec 5.
When several treatment arms are administered along with a control arm in a trial, dropping the non-promising treatments at an early stage helps to save the resources and expedite the trial. In such adaptive designs with treatment selection, a common selection rule is to pick the most promising treatment, for example, the treatment with the numerically highest mean response, at the interim stage. However, with only a single treatment selected for final evaluation, this selection rule is often too inflexible. We modified this interim selection rule by introducing a flexible selection margin to judge the acceptable treatment difference. Another treatment could be selected at the interim stage in addition to the empirically best one if the differences of the observed treatment effect between them do not exceed this margin. We considered the study starting with two treatment arms and a control arm. We developed hypothesis testing procedures to assess the selected treatment(s) by taking into account the interim selection process. Compared with the one-winner selection designs, the modified selection rule makes the design more flexible and practical.
当试验中同时使用多个治疗臂和一个对照臂时，早期放弃无前途的治疗方法有助于节省资源并加快试验进度。在具有治疗选择的适应性设计中，常见的选择规则是在中期选择最有前途的治疗方法，例如，数值上最高的平均反应治疗。然而，由于最终只有一种治疗方法被选中进行最终评估，因此这种选择规则往往过于僵化。我们通过引入灵活的选择裕度来判断可接受的治疗差异，从而修改了这种中期选择规则。如果两者之间观察到的治疗效果差异不超过此裕度，则可以在中期阶段选择另一种治疗方法，而不仅仅是经验上最好的治疗方法。我们考虑了从两个治疗臂和一个对照臂开始的研究。我们开发了假设检验程序，通过考虑中期选择过程来评估所选的治疗方法。与单一获胜者选择设计相比，修改后的选择规则使设计更加灵活和实用。
