Methodological Advances in Using Negative Control Endpoints to Improve Vaccine Effectiveness Estimation

Abstract

This dissertation addresses two problems at the frontier of vaccine effectiveness evaluation that are important to vaccine science and public health: (i) obtaining reliable evidence of waning vaccine effectiveness over time and (ii) improving the precision of vaccine efficacy estimates in randomized vaccine trials. To address these challenges, we develop novel statistical methods that leverage negative control outcomes—outcomes unaffected by vaccination yet correlated with the primary outcome through unmeasured common causes. In Part 1, we examine the unmeasured biases affecting commonly used hazard-based estimators of time-varying vaccine effectiveness and propose statistical frameworks for diagnosing and correcting for unmeasured bias using negative control infection endpoints. As a case study, we apply our methods to the Coronavirus Vaccine Efficacy (COVE) study, which led to qualitative reassessment of mRNA-1273 booster efficacy in this setting. In Part 2, we extend statistical methods for covariate adjustment to incorporate valid negative control endpoints in early-and-late-phase randomized vaccine trials. With particular emphasis on HIV-1 vaccine trials, we show that incorporating negative control immune responses and infection outcomes can enhance the efficiency of vaccine effect estimates when negative control and primary outcomes share unmeasured prognostic factors.