Primary Submission Category: Bayesian Causal Inference
Identified vaccine efficacy for binary post-infection outcomes under misclassification without monotonicity
Authors: Rob Trangucci, Yang Chen, Jon Zelner,
Presenting Author: Rob Trangucci*
In order to meet regulatory approval, pharmaceutical companies often demonstrate that new vaccines reduce the total risk of a post-infection outcome like transmission, symptomatic illness, or death in randomized, placebo-controlled trials.
One can use principal stratification to partition this causal effect into vaccine efficacy against infection, and the principal effect of vaccine efficacy on post-infection outcomes in always-infected patients.
Unfortunately, even under strong assumptions, these principal effects are generally unidentifiable.
We develop a novel method to nonparametrically point identify these principal effects while eliminating the typical monotonicity assumption and allowing for measurement error in both infection and post-infection outcomes.
Furthermore, we show that our results readily extend to multiple treatments.
Our method takes advantage of the geographic heterogeneity of disease incidence, and well-measured biologically-relevant categorical pretreatment covariates, each a feature of many vaccine trials.
We show that a Bayesian version of our method can be applied to a variety of clinical trial settings where vaccine efficacy against infection and a post-infection outcome can be jointly inferred, and investigate the sensitivity to prior specification using simulation studies.
Finally, we apply this method to an influenza vaccine trial to yield new insights into vaccine efficacy against symptomatic illness.