Meta-analysis of estimates of influenza vaccine effectiveness using alternative comparison groups in the context of potential virus interference

Abstract

BACKGROUND: The test-negative design (TND) has been increasingly used to estimate influenza vaccine effectiveness (VE). Patients who meet with certain case definitions are swabbed and tested for influenza by laboratory methods. In some studies, patients are also tested for other respiratory viruses. A hypothesis was previously proposed that virus interference might lead to a correlation between receipt of influenza vaccination and increased risk of infection with other respiratory viruses. Among patients testing negative for influenza virus, VE can be estimated by comparing cases with (1) influenza-negative controls, VE(FLU-), (2) controls with positive detection for other respiratory viruses, VE(ORV+) , or (3) controls negative for all viruses tested, VE(PAN-). The aim of our study is to assess if the choice of alternative control groups might affect VE estimates in the TND. METHOD: We searched publications in Medline, PubMed and Web of Science using a combination of keywords. Articles evaluating VE by the TND against any type of influenza virus among any population were considered. Studies reporting VE estimates using more than one control group were included. Interim studies or re-analysis were excluded. For each eligible study, VE estimates by all alternative control groups against any type of influenza virus for any age group were further extracted. Influenza vaccination coverage among alternative control groups were compared by paired t-test. We further calculated VE differences (ΔVE) between alternative control groups for each available pair using bootstrap method to estimate confidence intervals. RESULTS: In total 9 articles were identified reporting VE estimates by both influenza negative and other respiratory virus positive controls, while 5/9 also reported VE estimates by pan-negative controls. These included VE estimates in 6 countries for all ages from 2004/05 to 2012/13. The proportion of patients testing positive for other respiratory viruses ranged from 33.7% to 74.6% among patients who tested negative for influenza. Differences in vaccination coverage between influenza-negative and other respiratory virus-positive group ranged from -7.8% to 13.3%. There was no difference in vaccination coverage between these two control groups (p=0.745). A total of 38 VE(FLU-), 38 VE(ORV+), and 10 VE(PAN-) estimates were extracted for further comparison and there were no statistically significant differences in ΔVE. CONCLUSION: Based on 9 studies estimating VE using alternative control groups by the TND, we did not find any evidence of differences in VE based on choice of comparison groups.