Date of Completion
Model diagnostic; Semi-competing risks; Survival models; Time-varying treatment switching; Cure rate model; DIC Decomposition; LPML Decomposition; Componentwise DIC; Markov chain Monte Carlo; Patient-reported outcome;
Elizabeth D. Schifano
Field of Study
Doctor of Philosophy
Treatment switching can occur in clinical trials when patients randomized in one group are allowed to switch to other groups. Joint modeling of longitudinal and survival data has the potential of reducing the bias and providing greater efficiency in the estimate of the treatment effect compared to an estimate based on the survival data alone.
In this dissertation, we propose a class of semiparametric semi-competing risks transition survival models to accommodate two-way time-varying switching. For the first time in the treatment switching literature, we are able to propose a model diagnosis procedure for the proposed approach and apply it to the real data reanalysis.
We also develop a class of trajectory-based joint models for longitudinal and survival data with disease progression. In addition, we derive the decompositions of the deviance information criterion (DIC) and the logarithm of the pseudo marginal likelihood (LPML) and the componentwise DIC to assess the fit of the longitudinal component of the model and the fit of each survival component, separately. We further develop Delta DIC, Delta LPML and Delta componentwise DIC to determine the importance and contribution of the longitudinal data to the fit of the TP and OS data. Moreover, efficient Markov chain Monte Carlo sampling algorithms are developed to carry out posterior computations.
Zhang, Fan, "Statistical Models and Methods for Longitudinal and Survival Data with Treatment Switches" (2018). Doctoral Dissertations. 1926.
Available for download on Thursday, June 29, 2028