Nonparametric bounds for vaccine effects in randomized trials

2025-10-29

AI Summary

In randomized vaccine trials, unblinding induces behavioral bias that confounds vaccine efficacy (VE) estimates, conflating immunological effects with behavioral responses. Conventional causal identification approaches rely on the strong assumption of no unmeasured confounding between infection risk and vaccine beliefs—a condition often violated by latent factors such as personality traits. Method: We relax this assumption and propose a nonparametric causal bounding method within the potential outcomes framework, integrating monotonicity constraints with linear programming to derive falsifiable upper and lower bounds for VE under multiple definitions. Contribution/Results: Our approach is validated on fully synthetic data and semi-synthetic data calibrated to real-world COVID-19 vaccine trials. It substantially improves the robustness and credibility of VE estimation under unblinding, offering a novel paradigm for causal inference in settings where behavioral responses compromise treatment assignment integrity.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements

Abstract

Vaccine randomized trials are typically designed to be blinded, ensuring that the estimated vaccine efficacy (VE) reflects the immunological effect of the vaccine. When blinding is broken, however, the estimated VE reflects not only the immunological effect but also behavioral effects stemming from participants' awareness of their treatment status. Recent work has proposed alternative causal estimands to the standard VE to address this issue, but their point identification results require a strong assumption: the absence of unmeasured common causes of infection risk and participants' belief about whether they received the vaccine. Personality traits, for example, may plausibly violate this assumption. We relax this assumption and derive nonparametric causal bounds for different types of VE. We construct these bounds using two approaches: linear programming-based and monotonicity-based methods. We further consider several possible causal structures for vaccine trials and show how the nonparametric bounds differ across these scenarios. Finally, we illustrate the performance of the proposed bounds using fully synthetic data and a semi-synthetic data example based on a COVID-19 vaccine trial.

Problem

• Estimating vaccine efficacy when blinding is compromised in trials
• Relaxing strong assumptions about unmeasured confounding variables
• Deriving nonparametric bounds for vaccine effects under causal structures

Innovation

• Relaxing assumption of no unmeasured confounders
• Deriving nonparametric causal bounds for vaccine efficacy
• Using linear programming and monotonicity methods