Bayesian Semiparametric Model for Sequential AML Treatment Decisions with Informative Timing

We develop a Bayesian semi-parametric model for the impact of dynamic treatment rules on survival among patients diagnosed with pediatric acute myeloid leukemia (AML). The data are from a phase III clinical trial in which patients move through a sequence of four treatment courses. At each course, they undergo backbone chemotherapy that may or may not include an anthracycline-based (ACT) agent. While ACT is thought to more aggressively suppress AML, it is also cardiotoxic. Thus, treating overzealously with either may reduce survival. Our task is to estimate the potential survival probability under hypothetical dynamic treatment strategies, but there are several impediments. First, since ACT is not randomized, its effect on survival is confounded over time. Second, subjects initiate the next course depending on when they recover from the previous course - thus, timing is variable and potentially informs subsequent treatment decisions and survival. Third, patients may die or drop out before ever completing the full sequence. We develop a generative Bayesian semi-parametric model based on Gamma Process priors to address these complexities. At each treatment course, the model captures subjects' transition to subsequent treatment or death in continuous time. A g-computation procedure is used to compute a posterior over potential survival probability that is adjusted for time-varying confounding. We estimate the efficacy of AML treatment rules that dynamically assign ACT based on evolving cardiac function.