Clinical trials of a new treatment may proceed through three phases. Phase I trials are small studies that evaluate toxicity with a specific task to determine the maximum tolerated dose. Once a safe dose of the treatment is chosen, its therapeutic efficacy will be tested in a phase II trial. Regimens shown promising in phase II trials will then be moved to large, multi-institutional phase III studies that compare their effectiveness to standard treatments. With many candidate regimens available, it is imperative to identify the most promising therapies for the expensive phase III testing. This has become increasingly important because of the limited subject availability and funding resources, and an ever increasing number of new compounds due to high throughput screening. In this research, we propose novel statistical designs and strategies that utilize the complex clinical data in an efficient manner, which is hoped to translate into equally accurate clinical conclusions with fewer resources. Specifically, this renewal application covers the following three clinical scenarios. First, we propose methods for phase I dose-finding trials with multiple safety endpoints under heteroscedasticity and multiple objective constraints. Existing designs collapse the endpoints into a dichotomized indicator of toxicity or no-toxicity, and may do so at the expense of not utilizing all information available and over-simplifying the complex clinical objectives. Our proposed methods will retrieve the information loss by using all endpoints and achieve clinical relevance by accommodating multiple objective constraints. Second, we propose methods for phase II dose- finding trials based on both safety and efficacy endpoints, in which patients will be enrolled in two stages. Having an interim analysis, we can shut down ineffective or unsafe doses and reduce the number of patients treated at these doses. Third, we propose designs to select treatments in phase II trials based on both clinical and biologic endpoints. This work extends our ongoing research on sequential selection boundaries for trials with a single biologic endpoint. While a biologic endpoint is typically less noisy than a clinical endpoint such as the modified Rankin scale in stroke patients, the primary therapeutic objective is to improve the clinical outcomes. Our bivariate approach will improve the efficiency in treatment selection by using the less noisy biologic endpoint while assuring the design is clinically relevant via its use of the clinical outcomes. These designs will be applied to design various clinical trials in patients with neurological disorders.
Despite the efforts in the past decade, additional therapies for neurological disorders such as acute ischemic stroke are sorely needed. Upon successful completion of this research, we will extend our capacity to design early phase investigation of new treatments and enhance the statistical efficiency of selection and screening process in a variety of clinical trial settings.
Showing the most recent 10 out of 20 publications
