Causal inference in HIV vaccine trials: comparing outcomes in a subset chosen after randomization
In many experiments researchers would like to compare between treatments an outcome that only exists in a subset of participants selected after randomization. For example, in preventative HIV vaccine efficacy trials it is of interest to determine whether randomization to vaccine affects post-infection outcomes such as HIV viral load or the time from infection diagnosis to AIDS. This dissertation addresses some of the challenges of making causal comparisons conditioning on an event that occurs after randomization. Following the approach of Gilbert, Bosch, and Hudgens (2003) (GBH), I propose sensitivity analysis methods to estimate the average causal effect of treatment assignment on a post-infection outcome among those who would be infected whether randomized to vaccine or placebo. My key assumption is that subjects randomized to the vaccine arm who become infected would also have become infected if randomized to the placebo arm. It is not known which of those subjects infected in the placebo arm would have been infected if assigned vaccine, but this can be modeled using baseline covariates, the observed outcome variable, and a specified sensitivity parameter. In this dissertation I first construct a general likelihood. Using this likelihood, I then show that the method proposed by GBH yields a semiparametric maximum likelihood estimate of the average causal effect. Based on the likelihood, I extend GBH by including baseline covariates; allowing discrete, truncated, and censored outcomes; and permitting more general selection bias functions. In particular, I study two different modeling approaches for estimating the average causal effect conditional on base-line covariates when the outcome of interest is continuous. In addition, I propose and study the behavior of semiparametric estimators of the causal effect of vaccination on an independently censored time to event outcome, deriving asymptotic properties and evaluating small sample behavior through simulations. I apply these methods to the first Phase III preventative HIV vaccine trial (VaxGen's trial of AIDSVAX B/B).
- Biostatistics