论文部分内容阅读
This thesis presents three novel statistical designs for treatment identification and consistency assessment in clinical trial.We consider the problem in phase Ⅱ clinical trial of determining whether an experimental treatment is sufficiently promising to warrant further investigation,and the problem in phase Ⅲ multi-regional clinical trial to test the overall efficacy of the drug across subgroups and evaluate the possibility of applying the overall trial result to some specific subgroup.We propose a two-stage procedure for selecting the best treatment among multiple candidate treatments for phase Ⅱ clinical trial where the heterogeneity of study population is taken into account by pre-defined subgroups.We define the best treatment as the one with the highest response rate.The procedure screens both treatments and subgroups at the same time that are promising at the first stage and selects the best treatment as winner based on a subgroup-weighted effective rate.We obtain the least favorable configuration to compute the lower bound of the probability of correct selection.The early stopping rules are proposed to control pre-specified allowable errors.Sample size determination is proposed to meet the pre-specified probability of correct selection as well as early stopping error control.Extensive simulations reveal that the resulting sample size saving can be significant in comparison with the traditional Simon’s design.We illustrate the application of the proposed method to a breast cancer study.Two-stage designs using a data based decision boundaries have been used widely.For different interim analysis,boundaries are set up so as to stop or continue the trial to next stage/phase.In this thesis,we propose a testing procedure for phase Ⅱ trial with heterogeneous population,in which the activity of a new treatment is anticipated to vary across subpopulations.The proposed strategy includes both stratum specific and combined analysis where the stratum specific analysis is followed by overall combined efficacy test.As a result of stratum specific analysis（at both stages）,subgroups are categorized as efficacious,ambiguous or inefficacious.The subgroup specific analysis is considered as a preliminary test for assigning weights to be used in the combined analysis.In this design,the trial is terminated early only if the decision from the combined test rejects the overall alternative hypothesis.When the drug is claimed efficacious at the end of stage 1 or stage 2,subgroups worth for further investigation are also recommended.Simulations are carried out to illustrate performance of the proposed design.We demonstrate the application of the proposed method to a hypothetical example.Thirdly,we present a novel approach for consistency assessment in multi-regional clinical trial（MRCT）.In the proposed design,we focus on the problem of evaluating applicability of a drug to a specific region of interest under the criterion of preserving a certain proportion of the overall treatment effect in the region.We propose a variant of James-Stein shrinkage estimator in the empirical Bayes context for the region-specific treatment effect.The estimator has the features of accommodating the between-region variation and finiteness correction of bias.We also propose a truncated version of the proposed shrinkage estimator to further protect risk in the presence of extreme value of regional treatment effect.Based on the proposed estimator,we provide the consistency assessment criterion and sample size calculation for the region of interest.Simulations demonstrate the performance of the proposed estimators in comparison with some existing methods.A hypothetical example is presented to illustrate the application of the proposed method.In summary,the proposed designs are well motivated by practical needs and shown to be more efficient than some existing designs from different perspectives.They can be further explored and extended to meet more challenges.It is also believed that the results in this thesis will serve as a useful addition to the growing literature on statistical designs in clinical trial.