Title: A Bayesian nonparametric approach for causal inference with semi-competing risks
Authors: Michael Daniels - University of Florida (United States) [presenting]
Abstract: A Bayesian nonparametric (BNP) model is developed in order to assess the treatment effect in semi-competing risks, where a nonterminal event may be censored by a terminal event, but not vice versa. Semi-competing risks are common in brain cancer trials with death being censored by cerebellar progression. We propose a flexible BNP approach to model the joint distribution of progression and death events, thereby effectively inferring the marginal distributions of progression time and death time, characterizing within-subject dependence structure, predicting the progression and death times given a patient's covariate, and quantifying uncertainties of all estimates. More importantly, we define a causal effect of treatment, which can be estimated from the data and has a nice causal interpretation. We perform extensive simulation studies to evaluate the proposed BNP model. The simulations show that the proposed model can accurately estimate the treatment effect in semi-competing risks setup. We also implement the proposed BNP model on data from a brain cancer Phase II trial.