在第二期臨床試驗時，我們不希望花費太多的時間和資源，因此我們需要有效率的隨機指派方法，挑出可能有療效的治療方法，趕緊進到第三期試驗做藥效的確認。傳統上使用存活時間當作主要終點(primary endpoint)，將花費太長的時間，無法有效地在試驗期間適應性地隨機指派不同治療給病人，然而使用短期的療效(例如腫瘤反應率)當作隨機指派的依據，可能病人短期療效好，但存活時間卻不長。由於主要的目的是希望新的治療可以使病人存活時間延長，因此Huang等人(2009)等人使用短期反應率的訊息來加速存活時間適應性隨機指派臨床試驗。本論文將推廣Huang等人(2009)的方法由兩組治療的比較推廣至多組比較，並將存活時間為指數分配的假設推廣至韋伯分配。由於Huang等人(2009)提出的模型是每一個世代(cohort)一位病人進案時都進行期中分析，下一個進到試驗的病人則是根據修正的隨機指派之機率進行隨機指派，為節省時間，我們考慮使用每一個世代為五個人進到試驗後再進行期中分析，並比較每一個病人和每五個病人的模擬結果。 We usually don’t want to waste substantial amount of time and resources in phase two clinical trials, therefore we need effective methods of randomization, so that we can chose a potential treatment on survival, and confirm its benefit in phase three trial quickly. Traditionally, using survival time as a primary endpoint, takes a long period of time to assign patients to different groups ineffectively during the adaptive-randomization trails. However, using short-term efficacy (e.g. tumor response rate) as a primary endpoint, may show that patients have good efficacy shortly, but do not have long period of survival time. Owing to prolonging survival time is a common goal of a better treatment, Huang et al. (2009) proposed a model which used short-term response information to facility adaptive randomization for survival clinical trials. In this article, we extend the method which Huang et al. (2009) proposed to multi-arms clinical trials. Moreover, we assume that the survival time of patients is following the mixture Weibull distribution. In addition, since the model of Huang et al. (2009) proposed conduct interim analyses each cohort of one patient, and following patient randomize into the trial is based on the revised probability of adaptive randomization. In order to save time, we consider to conduct interim analyses each cohort of five patients, then compare the respective simulation results.
