| Summary: | 碩士 === 國立成功大學 === 統計學系碩博士班 === 101 === To ensure the drugs are therapeutic and will not cause serious side effects on human, the drugs are required to go through four phases of clinical trials before they are approved for sale and marketing. Phase I clinical trial is the study where a drug is initially given to human. The main objective of phase I clinical trial is to identify the maximum tolerated dose (MTD). The purpose of phase I oncology clinical trial is not only to estimate MTD but also to treat the subject at a therapeutic dose since the participants in a phase I oncology trial are patients at advanced disease stages rather than healthy volunteers. Continual reassessment method (CRM) is the first model-based designs for phase I oncology clinical trial. The advantage of CRM is that it assigns few patients are treated at low and potentially non-therapeutic doses. But, the investigators consider it assigns too many patients to receive the doses which have high toxicity. Therefore, in this thesis, we propose an asymmetric loss function with fixed and flexible penalties to reduce the chance that patients receive overdosing assignment. We conducted simulations to compare the proposed methods with CRM. The results showed that relative to the CRM, the proposed methods efficiently decrease the proportion of patients who received overdoses and exhibited fewer toxicities, asymmetric loss function with fixed penalty especially. Additionally, both the standard deviation of number of patients who received an overdose and the proportion of overestimated MTD under CRM are lower than that under the proposed methods. Hence, the proposed methods improve the shortcomings of CRM.
|
|---|
