隨著廉價的治療方案和微創手術的發展，我們面臨越來越多非劣效性的研究問題，早期用於檢定新治療與標準治療所使用的非劣效性統計方法，較多的研究內容均針對二元或常態的假設下進行研究。隨著研究不斷發展，近期則越來越多人研究在實驗指標為存活中位數的情況下，評估藥效的問題。 在醫學實務中，經常根據存活中位數衡量患者的生存經歷及治療效果，因此本論文的目的是根據兩存活中位數的比值比較藥物的療效。我們利用 Su & Wei (1993) 中所提出信賴區間的方法與傳統的log-rank 方法分別進行非劣性檢定，並透過模擬研究比較兩種方法對於非劣性的檢定能力，並且計算其需要多少樣本才能達到我們想要的檢定能力。最後根據一筆實際資料說明所提方法的應用。關鍵詞:非劣性檢定；存活中位數；信賴區間 With the development of inexpensive treatment regimens and less invasive surgical procedures, we are confronted with non-inferiority study objectives. In the early days, statistical methods for showing the non-inferiority of a new (test) treatment compared to an established standard (reference) one had been mainly developed for binary or normal outcomes. Recently, more and more people have studied the issue of assessing drug efficacy in the case where the experimental indicator is the median failure time.In medical practice, the median failure time is a popular measure for characterizing the survival experience and treatment effect of a group of patients. Therefore, the purpose of this paper is to compare the efficacy of the drug based on the ratio of the median failure time. We use the confidence interval method proposed by Su & Wei (1993) and the non-inferiority log-rank method to perform non-inferiority verification. Then we compare two methods for non-inferiority verification ability by simulation studies and calculate how many samples it needs to achieve the verification capabilities we want. Finally, the application of the method is explained based on a practical data.Key words: non-inferiority test; median failure time; confidence interval
