Carl Andrew Pearson, W John Edmunds, Thomas J Hladish, Rosalind M Eggo
medRxiv
January 10, 2020
ABSTRACT
Background: Infectious disease outbreaks present unique challenges to study designs for vaccine evaluation. Test-negative case-control (TNCC) studies have been used to estimate vaccine efficacy previously, and have been proposed for Ebola virus disease (EVD) vaccines. However, there are key differences in how cases and controls are recruited during outbreaks that have implications for the reliability of vaccine efficacy estimates from these studies. Methods: We use a modelling approach to quantify TNCC bias for a prophylactic vaccine distributed across varying study and epidemiological scenarios. Our model accounts for vaccine distribution heterogeneity and for the two potential routes of recruitment: self-reporting and contact-tracing. We derive the TNCC estimator for this model and suggest ways to translate outbreak response data into the parameters of the model. Result: We found systematic biases in vaccine estimates from a TNCC study in our model of outbreak conditions. Biases are introduced due to differential recruitment from self-report and contact-tracing, and by clustering of participation in vaccination. We estimate the magnitude of these biases, and highlight options to manage them via restricted recruitment. For the motivating example of EVD, the absolute bias should be less 10%. Conclusion: A TNCC study may generate biased estimates of vaccine efficacy during outbreaks. Bias can be limited via recruitment that either minimizes heterogeneity in vaccination in the recruited population or excludes recruitment of contact-traced individuals. TNCC studies for outbreak infections should record the reason for testing to quantify potential bias in the vaccine efficacy estimate. Perfectly distinguishing the recruitment route may be difficult in practice, so it will be challenging to entirely remove this bias.