Virus Interference and Estimates of Influenza Vaccine Effectiveness from Test-Negative Studies

Abstract

To the Editor: Randomized, placebo-controlled trials provide the most convincing data on the protection against infection and illness conferred by influenza vaccines, but they are not always feasible.1 Despite the presence of technical difficulties including antigenic variations of the virus, nonmeasurable confounders, and typically mild disease presentation with nonspecific symptoms, observational studies can provide ongoing evidence that influenza vaccines retain effectiveness.1 One observational study design for estimating vaccine effectiveness (VE) involves identifying medically attended acute upper respiratory tract infections (URTIs) and attributing them to influenza or other etiologies using laboratory tests.1–8 Vaccination coverage is then compared between influenza-positive patients (cases) and influenza-negative patients (controls), adjusting for potential confounders. An apparent advantage of the use of controls with acute URTI is that it reduces the risk of bias owing to health-care seeking behavior if vaccinated individuals were more likely to seek care when ill.3–5 An assumption required for this “test-negative” case-control study design to be valid is that the risk of illness associated with noninfluenza infections must be independent of receipt of influenza vaccination.3 However, it has been hypothesized that a respiratory virus infection confers immunity against the same and other respiratory viruses for a short time, perhaps a few weeks, associated with the innate immune response to viral infection including interferons that have broad protective effects against a range of viruses.9,10 This biological mechanism, known as temporary nonspecific immunity, has been proposed as the cause of epidemiological “interference” between respiratory virus epidemics in which an epidemic of one virus seems to affect epidemics of other viruses at the ecologic level.11–13 Here, we discuss the implications of virus interference through temporary nonspecific immunity on VE estimates from test-negative studies. In the Table, cases that test positive for influenza are classified by vaccination status. Within the controls, it is possible to separate those that test positive for a noninfluenza respiratory virus from those that test negative for all viruses. Odds ratios (ORs) can be estimated using either or all control groups (OR1, OR2, and OR3 in the Table), and the calculations are illustrated with data from a published study.2 In that study, influenza vaccination seemed to be associated with an increased risk of noninfluenza respiratory virus infections, which is consistent with temporary nonspecific immunity. In a separate placebo-controlled trial of influenza vaccination, we reported that recipients of influenza vaccine had significantly higher risk of noninfluenza respiratory virus infections.14TABLE: Schematic Table of a Test-Negative Case-Control Study Illustrated with Data from a Published StudyTemporary nonspecific immunity would lead to a higher risk of noninfluenza respiratory virus infections among vaccine recipients, ie, in general we might expect B1/D1 > B2/D2 and OR2 < OR1 < OR3. If VE were estimated as 1 − OR, the inclusion of individuals who test positive for other respiratory viruses would tend to increase the VE estimate. It is unclear whether OR3 would provide an unbiased estimate of VE. A negative test for respiratory viruses in a patient with acute URTI does not exclude a respiratory virus etiology because the test may lack perfect sensitivity particularly for lower viral loads, the specimen may have been incorrectly collected from the patient, or for other reasons.2 There are some caveats to this discussion. First, ORs from case-control studies should be adjusted for potential confounders that may relate both to being vaccinated and to experiencing a URTI associated with influenza, although adjustment may not affect the ordering of OR1, OR2, and OR3. Nonspecific immunity cannot easily be directly measured, and the same problem may extend to conditional evaluation methods in transmission studies.15 Second, we have not discussed other potential biases in test-negative studies which are explored in detail elsewhere.3,5 Finally, if influenza incidence is low or temporary nonspecific immunity is weak, there should not be substantial differences among OR1, OR2, and OR3. Selection of the control group can be a weak point of case-control studies,15 and, as discussed here, can be affected by transmission dynamics including interference. More detailed investigations including simulation approaches are warranted to aid in the interpretation of VE estimates from test-negative studies. ACKNOWLEDGMENT We thank Dennis Ip for helpful discussions. Benjamin J. Cowling School of Public Health Li Ka Shing Faculty of Medicine The University of Hong Kong Hong Kong Special Administrative Region, China [email protected] Hiroshi Nishiura School of Public Health Li Ka Shing Faculty of Medicine The University of Hong Kong Hong Kong Special Administrative Region, China PRESTO Japan Science and Technology Agency Saitama, Japan