Average treatment effect evaluation with time-to-event data in randomized clinical trials and observational studies
[EMBARGOED UNTIL 5/1/2024] The average treatment effect (ATE) is defined as the difference in the expected outcome between individuals receiving the treatment and those not receiving it. As a measure of the impact of a treatment or intervention on an outcome of interest, it has been used in various fields such as economics, epidemiology, and clinical trials. Current research in the field of ATE focuses on developing new methods for estimating the ATE in various study designs, including both randomized controlled trials (RCT) and observational studies. However, existing methodologies in survival analysis encounter multiple challenges such as the violation of the strict proportional hazard (HR) assumption and practical issues in oncology and chronic disease trials. The first part of this dissertation focuses on realistic scenarios in oncology trials, where event-free survival (EFS) is considered as an endpoint. The mixture cure model with cause-specific hazards (MCM-CSH) is fitted in real data and employed in simulations to mimic real oncology trials. We apply the restricted mean survival time (RMST) model without any assumptions to estimate ATE and investigate practical considerations, including data recording ways, predetermined assessment schedules, length of visit windows, size of treatment effects, and choices of the restricted time in RMST. When it comes to chronic diseases such as cancer and acquired immune deficiency syndrome (AIDS), they are featured in slow progression with multiple stages, giving rise to doubly-censored data. Although many methods have been proposed for regression analysis of such data, most of them are conditional on the occurrence time of the initial event and ignore the relationship between the two events or the ancillary information contained in the initial event. Corresponding to this, Chapter 3 presents a sieve maximum likelihood approach that makes use of the ancillary information contained in the initial event. The logistic model and Cox proportional hazards model are employed to model the initial event and the failure time of interest, respectively. A simulation study is conducted and suggests that the proposed method works well in practice and is more efficient than the existing methods as expected. The approach is applied to an AIDS study that motivated this investigation. Chapter 4 focuses on the estimation of ATE of a functional treatment in observational studies involving time-to-event data and presents a functional framework including a functional accelerated failure time (FAFT) model and three causal estimators. The regression adjustment estimator is based on conditional FAFT with subsequent confounding marginalization, while the weighted estimator is based on marginal FAFT. The double robust estimator combines the strengths of both methods and achieves a balance condition through the weighted residuals between observations and regression adjustment outcomes. The effectiveness of the proposed framework is demonstrated through a simulation study and is applied to investigate the causal effect of magnetic resonance imaging (MRI) on survival outcomes in Alzheimer's Disease.