Test-Negative Designs With Multiple Testing Sources

Test-negative designs, a form of case-cohort studies, have been commonly used to assess infectious disease interventions. Early examples of the design included the evaluation of seasonal influenza vaccines in the field. Recently, they have also been widely used to evaluate the efficacy of COVID-19 vaccines in preventing symptomatic disease for different variants [32]. The design hinges on individuals being tested for the disease of interest; upon recruitment, such individuals are subjected to a definitive test for the presence of the disease of interest (test-positives) or not (test-negatives), along with the determination of whether the individual has been exposed to the intervention under study (e.g., vaccination). In most early TND studies, individuals were tested because they were suffering from symptoms consistent with the disease in question, and the TND was a tool to reduce confounding due to healthcare-seeking behavior. However, in many cases, such as COVID-19 and Ebola, testing results were available at healthcare facilities for individuals who presented for a variety of reasons in addition to symptoms (e.g., case contact tracing, etc.). Aggregating samples from symptomatic and asymptomatic test results leads to bias in the assessment of the efficacy of the intervention. Here we consider these issues in the context of a specific version of the 'multiple reasons for testing problem,' motivated by a vaccine trial designed to assess a new Ebola viral disease vaccine (EVD) [22]. Some participants are recruited in the usual TND fashion as they present for care suffering from symptoms consistent with an Ebola diagnosis (and are thus tested); in addition, however, any test-positive identified in this fashion leads to immediate testing for Ebola for all close contacts of the test-positive who are likely asymptomatic at that point. We examine a simple approach to estimate the common efficacy of the vaccine intervention based on these two sources of test positives and test negatives, complemented by an assessment of whether efficacy is the same for both sources. The EVD trial was not completed for the fortunate reason that the prevailing disease outbreak ended; nevertheless, the approach here will be important if this trial is ever recommenced or similar trials are conducted in the future.