中文摘要：

在肿瘤新药Ⅱ期临床试验中，往往要对多个瘤种、多种剂量或用法进行探讨，目的是淘汰无效剂量、筛选敏感瘤种，以便进一步深入研究。当某试验组疗效未达到预期效果时，研究者希望尽可能早地终止该试验组的研究，避免更多的受试者接受无效的治疗。本文介绍多阶段设计，包括单阶段设计、二阶段设计和三阶段设计的统计学原理和设计思路，并提供各阶段设计所需样本含量和早期终止的价值，并以实例说明。

英文摘要：

In phase Ⅱ clinical trials, multiple kinds of tumors were treated with anti-cancer drugs at different dosages and through different administration pathways to exclude the ineffective dosages and screen sensitive tumor （ s ） for further study. When expectant effects are not in fact observed in the experimental groups, it is essential to terminate the trials to minimize exposure of subjeers to ineffective interventions. The trials are usually carried out with a multi-stage design. The present paper introduces the multi- stage design including the statistic principles and design ideas of single-stage, two-stage and three-stage design, provided the sample sizes needed in clinical trials at different stages, and estimated the value of early termination. All are illustrated with an example.