Novel designs for multi-arm multi-dose multi-stage platform trials

Project Summary/Abstract The drug development process is an intricate and resource-intensive endeavor, particularly in the context of addressing complex diseases with unmet medical needs, such as cancer, serious rheumatic diseases, acute ischemic stroke, among others. There has been a surge in the development of new therapeutic agents and combination treatment strategies, driven by rapid advancements in biological knowledge. However, the conventional approach to clinical trials, where treatments are evaluated one at a time in a sequential manner, is fraught with several shortcomings, especially in the era of precision (personalized) medicine. This \one-treatment-at-a-time" paradigm in drug development is associated with notably low success rates. It not only consumes valuable time and resources between separate trials but also extends the overall drug development timeline. The discrete nature of these phases impedes the ecient exchange of information across di erent stages, potentially leading to a loss of overall eciency. Furthermore, the evaluation of multiple treatments individually presents challenges in accurately estimating and interpreting the relative treatment e ects of each drug due to the presence of \treatment{trial" confounding. Viewing the drug development process as a whole system, the multi-arm multi-stage platform trial provides an e ective way to eciently evaluate modern treatments. Platform trial can eciently evaluates treatment by quickly advancing promising ones and discarding ine ective or overly toxic ones. This reduces the time it takes to identify e ective treatments for patients in unmet medical need. Platform trials can easily add new treatment arms, enabling continuous adaptation and optimization. Additionally, they control family-wise false positives and increase eciency with methods like multiple testing procedures and adaptive randomization to allocate more patients to better treatment arms, among other prominent bene ts. During recent years, several well-known platform trials have been conducted to advance drug development. These trials include I-SPY2 for breast cancer, REMAP-CAP for pneumonia, and DIAN-TU for Alzheimer's disease, among others. Additionally, during the COVID-19 pandemic, more than 50 COVID-19 platform trials were registered globally between 2020 and 2021. In response to the growing prevalence of platform trials, this research aims to propose robust Bayesian adaptive designs and methods to address the practical challenges that arise in real-world platform trials. These challenges include optimizing sequential monitoring of multiple treatments and/or multiple endpoints, eciently establishing proof-of-concept and dose selection in multi-arm multi-dose platform trials, managing incompatibility issues due to non-concurrent controls, and addressing late-onset outcomes, among others. Each proposed method will be tailored to tackle a combination of these challenges in speci c platform trial settings. For each design, user-friendly software will be developed, which will include programs for trial simulation to establish design operating characteristics, facilitate trial conduct, and assist physicians in selecting optimal treatment for their patients. The overarching goal is to develop Bayesian adaptive methods to identify superior treatments or doses across various diseases and clinical settings, ultimately aiming to achieve greater anti-disease e ects, improved safety, and enhanced survival outcomes.