May, SusanneOthus, MeganSelukar, Subodh Rajesh2022-01-262022-01-262022-01-262021Selukar_washington_0250E_23628.pdfhttp://hdl.handle.net/1773/48208Thesis (Ph.D.)--University of Washington, 2021This dissertation comprises three projects that span the design, conduct and analysis of contemporary clinical trials, presented in individual chapters. The first project extends methods for stratified randomization to account for the possibility that experimental arms of a platform trial differ in eligibility criteria. The second project proposes a novel approach for evaluating cure model appropriateness in studies with long-term survivors. The third project develops a framework for sequential monitoring in one N-of-1 trial and joint analysis of a series of sequentially-monitored N-of-1 trials. Project 1: We extend methods for stratified randomization to the setting of differing experimental arm eligibility in platform trials. We suggest modifying block randomization by including experimental arm eligibility as a stratifying variable, and we suggest modifying the imbalance score calculation in dynamic balancing by performing pairwise comparisons between each eligible experimental arm and standard of care arm participants eligible to that experimental arm. We also derive a formula to quantify the relative efficiency loss of platform trials with varying eligibility compared to trials with non-varying eligibility. Project 2: We develop a novel approach for evaluating cure model appropriateness in studies with long-term survivors. We propose a method that assesses the proportion of uncured remaining at the time of analysis. We demonstrate that this method has desirable asymptotic and finite-sample properties with parametric models and that it displays superior performance over existing methods. Project 3: We propose a framework for the sequential monitoring of one N-of-1 trial and the joint analysis of a series of sequentially-monitored N-of-1 trials. We suggest considering design blocks (repeated units of time with fixed numbers of each treatment allocation) as independent units for use with existing monitoring boundaries when analyzing continuous data with a linear mixed-effects model. To jointly analyze several trials together, we propose computing a bias-adjusted estimate for each trial and then combining the estimates with a random-effects model with inverse-variance weighting. We show that type-1 error can be inflated for N-of-1 trials with few treatment blocks under sequential monitoring, but trials with a substantial number of treatment blocks or with a substantial number of periods per block can have nominal rates. For those settings, our proposed framework for sequential monitoring can support clinicians in providing important decisions earlier, on average, for patients engaged in N-of-1 trials.application/pdfen-USCC BYClinical TrialsSequential MonitoringStratified RandomizationSurvival AnalysisBiostatisticsBiostatisticsThesis