dc.description.abstract | Early phase clinical trials for cancer research aim to estimate the optimal biologic dose (OBD) of the targeted drug under study where the OBD is the dose gives the largest efficacy probability but the toxicity probability is less than the targeted toxicity probability (TTP). Two-stage designs are considered for the early phase clinical trials. In the first stage, the mTPI-2 design is used to adaptively assign doses to patients based on the binary data of toxicity. In the second stage, cohort design is employed where the set of dose levels with acceptable toxicity is determined, and then the OBD is found in the set based on the efficacy response. The trial is early terminated and declared invalid if the toxicity is too strong or the effect is insufficient. If the trial continues, appropriate dose levels are determined and patients are randomly assigned adaptively or balancedly. When the number of patients reaches a predetermined number, we stop the trial and estimate the OBD. Since there are two different randomized designs in the second stage, the two-stage designs are denoted as mTPI2-ARD and mTPI2-BRD, respectively. This article also modify the two-stage designs for the situation with late-onset toxicity or delayed effect in order to shorten the trial time as denoted by mTPI2L-ARD and mTPI2L-BRD, respectively. The proposed two-stage designs, adaptively escalating doses based on the accumulated toxicity and efficacy responses, not only reduce the risk of assigning over-toxic doses to patients but also increase the probability of patients receiving effective doses. Moreover, the proposed designs do not assume any specific dose-toxicity or dose-efficacy model. In fact, the proposed designs are model-assisted two-stage designs because they use only a conjugate prior distributions for the toxicity probability and the efficacy probability in the Bayesian analysis. Finally, the performances of the two-stage designs comparing to the single-dose ascending designs or the previous two-stage design are investigated in a simulation study under various scenarios of toxicity and efficacy probabilities. The results of the simulation study show that the proposed two-stage designs are competitive to other early clinical designs in estimating the OBD with more rational dose assignment and outperform the competitive single-dose ascending designs in time management. | en_US |