Author ORCID Identifier


Date of Graduation


Document Type

Dissertation (PhD)

Program Affiliation

Biomathematics and Biostatistics

Degree Name

Doctor of Philosophy (PhD)

Advisor/Committee Chair

Ying Yuan

Committee Member

Amir A. Jazaeri

Committee Member

J. Jack Lee

Committee Member

Yisheng Li

Committee Member

Ruitao Lin


With the revolutionary achievement in molecular targeted therapies and cancer immunotherapies, the traditional drug development paradigm in phase II trials becomes increasingly inefficient due to its slow progress, high cost, and high failure rate. Fitting one standard strategy to all different trials also harms its reliability in decision-making because it doesn’t fully use all available resources and information in each trial. It’s crucial to develop novel phase II trial designs to accomplish different objectives for different types of trials. This research mainly focuses on Bayesian adaptive designs for phase II trials. Three types of trials are discussed in which traditional designs may encounter limitations, and three novel designs have been proposed to resolve these issues. Specifically, chapter 2 focused on randomized phase II trials. Compared to traditional single-arm trials, randomized trials could provide more consistent results even when bias in historical control is introduced. We proposed s Bayesian optimal design that has the capability of dealing with single and mixed endpoints for randomized trials. This design is optimized to maximize power while controlling type I error rate. Chapter 3 focused on platform trials. A platform trial could simultaneously screen multiple treatments on multiple indications, which reduces the length of the trial and saves operational resource. We proposed a Bayesian platform trial design that enables borrowing of information across subtrials with multiple primary endpoints. Chapter 4 discussed phase II trials with dual-criteria decision-making rules. Hypothesis testing in traditional trial designs could only show whether the investigational treatment is statistically better than historical control, which may not be sufficient to claim treatment is clinically effective. We proposed a Bayesian optimal design that takes both statistical significance and clinical significance into consideration. All these three proposed designs outperform the traditional designs by improving trial efficiency and providing desirable operating characteristics.


Bayesian adaptive design, phase II clinical trial, platform trial design

Included in

Biostatistics Commons



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.