Impact of unequal testing on vaccine effectiveness estimates across two study designs: a simulation study

Observational studies are essential for measuring vaccine effectiveness. Recent research has raised concerns about how a relationship between testing and vaccination may affect estimates of vaccine effectiveness against symptomatic infection (symptomatic VE). Using an agent-based network model and SARS-CoV-2 as an example, we investigated how differences in the likelihood of testing by vaccination could influence estimates of symptomatic VE across two common study designs: retrospective cohort and test-negative designs. First, we measured the influence of unequal testing on symptomatic VE estimates across study designs and sampling periods. Next, we investigated whether the magnitude of bias in VE estimates from unequal testing was shaped by immune escape (vaccine efficacy against susceptibility and against infectiousness) and underlying epidemic potential (probability of transmission). We found that unequal testing led to larger bias in the cohort design than the test-negative design and that bias was largest with lower efficacy against susceptibility. We also found the magnitude of bias was moderated by the study's selected sampling period, efficacy against infectiousness, and probability of transmission, with these moderating effects more pronounced in the test-negative design. Our study illustrates that VE estimates across study designs require careful interpretation, especially in the presence of epidemic and immunological heterogeneity.