A0599
Title: Bayesian semiparametric models for informatively timed, dynamic treatments with incomplete covariate trajectories
Authors: Arman Oganisian - Brown University (United States) [presenting]
Jason Roy - Rutgers University (United States)
Abstract: A Bayesian semiparametric model is developed for assessing the impact of dynamic treatment rules (DTRs) on survival. The motivating data are from a phase III clinical trial in which patients diagnosed with pediatric acute myeloid leukaemia (AML) move through a sequence of four treatment courses. At each course, a decision is made to administer anthracyclines (ACT). Since ACT is cardiotoxic, left ventricular ejection fraction (EF) is sometimes - but not always - measured beforehand to help inform the ACT decision. Inconsistent assessment leads to incomplete information on EF trajectories, a key tailoring variable. Moreover, patients 1) initiate each course at different times depending on the speed of recovery from previous courses, 2) may die or 3) be withdrawn from the study before ever completing the full sequence. The problem is framed in terms of a joint treatment-monitoring DTR that outputs both an EF monitoring decision and an ACT treatment decision. Gamma Process priors are used to flexibly model continuous-time transitions between treatment courses and death under hypothetical DTRs. A g-computation procedure simulates the transition process under hypothetical DTRs and computes posterior marginal survival probabilities.
