研發癌症新藥的第一期臨床試驗之主要目標為估計最大耐受劑量，其中服用該劑量的病患產生劑量限制毒性之機率最接近目標毒性機率。因為試驗的劑量升降過程也應該以保護病患為要，宜避免其施予過高的劑量。本文建議升降劑量過程針對可接受毒性進行貝氏檢定，其中可接受毒性為劑量限制毒性機率低於目標毒性機率之事件。當試驗病患人數用盡後，本文則建議選取低於最大耐受劑量後驗分布眾數的最高劑量水準，供下一期臨床試驗使用。本文藉由模擬在不同劑量毒性模型下的毒性機率決定合理的臨界點，進一步藉模擬研究本文所提出的貝氏檢定設計，相對於連續重評估方法與控制藥物過度增量設計，在劑量限制毒性之機率與最大耐受劑量之估計表現。此外，更藉由探討在誤用劑量毒性模型時，上述設計之相對表現。模擬研究顯示，本文提出的貝氏檢定設計除了在劑量升降過程中，能合理的控制過量給藥的可能性，也得到相對靠近真實最大耐受劑量之估計，同時，在錯誤的使用劑量毒性模型下，其表現為相對穩健。;In the development of new drugs for oncology, the primary objective of phase I clinical trials is to estimate the maximum tolerated dose (MTD) so that the probability of patients receiving the MTD and experiencing the dose-limiting toxicity (DLT) is closest to the target toxicity probability (TTP). Meanwhile, it is important to protect patients from overdosing in the dose escalation procedure. To do so, in this paper, Bayesian tests are proposed for the acceptable toxicity in the dose escalation procedure, where the acceptable toxicity is the event that the toxicity probability is less than the TTP. When the maximum number of patients is reached, the highest dose level under study but less than the posterior mode of the MTD is recommended for future clinical trials. The method is then denoted by BTMD design. Plausible cutoffs for Bayesian tests in the BTMD are investigated via a simulation study under a variety of dose-toxicity models. A simulation study is then used to explore the toxicity probability and the MTD estimation of the proposed BTMD design and some other competing designs, including the well-known continuous reassessment method (CRM) and escalation with over-dose control (EWOC) design. A sensitivity study is also implemented to investigate the performances of the competitive designs when the dose-toxicity model is misspecified. The results of the simulation study indicate that the proposed BTMD design reasonably controls the probability of overdosing and gives an estimated MTD which is relatively close to the true MTD. Moreover, relative to the competitors, the BTMD design is more robust to the misspecified dose-toxicity model.
