Bounding the infectiousness effect in vaccine trials

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Vaccine trials can demonstrate protective effects not only for the vaccinated individual but also for unvaccinated individuals within a population.
• Two primary mechanisms for this indirect protection exist: preventing transmission from an infected vaccinated individual and reducing contagiousness if the vaccinated individual becomes infected.
• The latter mechanism is termed the "infectiousness effect" of a vaccine.

Purpose of the Study:

• To investigate the bias in crude estimators used to quantify the "infectiousness effect" of vaccines.
• To develop a framework for understanding and estimating the causal infectiousness effect under conditions of interference.
• To assess the properties of crude estimators in the presence of selection bias.

Main Methods:

• Utilized causal inference theory, specifically focusing on interference between individuals.
• Employed a principal-stratification framework to dissect different causal pathways.
• Analyzed potential sources of bias, including selection bias due to postvaccination infection status and pathogen virulence.

Main Results:

• Demonstrated that crude estimators for the infectiousness effect are subject to selection bias.
• Showed that, under plausible assumptions, these biased crude estimators are conservative for the true causal infectiousness effect.
• Confirmed that this conservativeness holds even when accounting for selection bias related to comparison groups and pathogen virulence.

Conclusions:

• Crude estimators, despite their bias, provide a conservative lower bound for the causal infectiousness effect of vaccines.
• The principal-stratification framework offers a robust method for causal inference in vaccine trials with interference.
• Findings have implications for accurately assessing the population-level impact of vaccination strategies.