Presenter: Ziqian Zhuang

Supervisory Committee: Wei Xu (Supervisor), Amy Liu, Dehan Kong

Date and Time: Thursday, May 30, 2024, 12-2pm EST

Location: 155 College Street, Health Sciences Building, Room 734

Abstract: Endpoints in research studies and clinical trials serve as critical markers or outcomes that help evaluate the effectiveness, safety, and overall impact of interventions or treatments. Analyzing endpoints not only aids in understanding the efficacy and safety of interventions but also guides decision-making processes in healthcare and pharmaceuticals. Multiple types of endpoints, including continuous, binary, and survival data, are commonly employed to assess treatment effects or drug efficacy from diverse perspectives. However, many existing studies tend to disregard the correlations among multiple endpoints, analyzing each endpoint independently and thus leaving considerable room for exploring models that integrate various endpoint types. Additionally, determining appropriate metrics for feature selection during joint analysis poses a worthwhile direction in the face of the dual challenges of different types of endpoints and high-dimensional features.

We first propose a novel feature selection method for ensemble analysis of multivariate outcomes mixed with continuous and binary variables, thereby addressing an existing gap in the literature. This method holds the potential for extension to incorporate variables of other types, thereby broadening its applicability. Then we specifically focus on the joint model of endpoints comprising multiple adverse events, wherein we consider the severity of adverse events (AE) alongside distinguishing between long term and short-term adverse events in our model. Our objective is to offer a comprehensive understanding of the impact of treatments on multivariate event occurrences. We intend to further develop a more advanced model to accommodate data featuring interval censoring and terminal events.