| 摘要: | 本文針對癌症治療研發合併使用兩種藥物的複合藥設計早期臨床試驗，目的是找尋複合藥的最佳生物劑量組合，讓服用該劑量組合的病患發生劑量限制毒性的機率低於設定的目標毒性機率，且有最大機率出現藥效反應。本文提出兩種早期臨床試驗設計，第一種為兩階段設計，其中第一階段根據二元毒性反應資料初尋可接受毒性的劑量組合，然後利用二元藥效反應資料，在可接受毒性的劑量組合中，初尋有效的劑量組合即為可行的劑量組合；第二階段則分別使用調適性或平衡設計隨機配置世代病人服用持續修正的可行劑量組合。當受試者達到預定的人數時，則中止試驗，估計最佳生物劑量組合。第二種是單劑量組合漸增設計，同時根據二元毒性及藥效反應尋找可行的劑量組合，然後調適性的分配世代病人服用該劑量組合。本文進一步討論試驗中若有毒性晚發與藥效延遲的情形，如何在有效縮短試驗時間之下，修正試驗的設計。因為本文建立的調適性試驗設計無需使用特定的劑量組合與毒性或藥效關係的模式，僅務實地假設複合藥毒性會隨劑量組合提升而增強，藥效會隨劑量組合提升而增強或是在高劑量組合後維持不變，故上述的兩階段設計或單劑量組合漸增設計為模型輔助設計。本文最後在各種不同的毒性機率與藥效機率的組合情境下進行模擬研究，結果顯示兩階段設計能適當的控管試驗時間，單劑量組合漸增設計耗時良久，但是在正確估計最佳生物劑量組合與病人配置方面表現較佳。因此本文建議在單劑量組合漸增設計之前加入啟動階段，且在最大受試人數不變下，增加每一世代病人數以降低世代試驗次數。修正的單劑量組合漸增設計將能合理的管控試驗時間且維持鑑別最佳生物劑量組合的效率。;This thesis considers early clinical trial designs for cancer treatment by using the combined drugs as a combination of two drugs. Therefore, the purpose of early clinical trial designs is to find the optimal dose combination (ODC) so that the combined drugs preserve the maximum efficacy, while maintaining allowable probability of dose-limiting toxicity (DLT). Two different early phase clinical trial designs are considered for combined drugs. The first one is a two-stage design, where, in the first stage, dose combinations with admissible toxicity are initially estimated and then an initial estimate of dose combinations with feasible efficacy is obtained, and, in the second stage, cohorts of patients are randomly assigned adaptively or in balance to the selected ones from the continuously reassessed feasible dose combinations. When patients reaches the maximum number, the trial is stopped and the ODC is estimated. The other one is a single dose-combination ascending design that adaptively assigns feasible dose combinations to the next cohort of patients where the feasible ones are reevaluated based on the available binary data of toxicity and efficacy, simultaneously. The proposed trial designs are further modified for the situation with late-onset toxicity or delayed efficacy. The proposed two adaptive trial designs are, in fact, the model-assistant ones since both do not rely on any specific model for describing the relationship between the dose combination and toxicity or efficacy, but only employ a priori knowledge that the toxicity increases with dose combinations, and the efficacy increases and then remains the same at some high dose combinations. Finally, a simulation study is conducted to investigate the performances of the proposed designs under a variety of scenarios of toxicity and efficacy probabilities. The results show that the two-stage design well controls the trial time, while the time consuming single dose-combination ascending design performs better in correct ODC selection and patient assignment. However, after adding a start-up phase and increasing cohort size, hence, reducing number of cohorts under the same total number of patients, the modified single dose-combination ascending design would have a reasonable control of trial time and remain the efficiency on the ODC identification. |
