Estimating seasonal influenza vaccine effectiveness using the Test-Negative Design in Hong Kong

Abstract

Regular monitoring of seasonal influenza vaccine effectiveness (IVE) is essential to confirm adequacy of protection and identify factors affecting IVE in the vaccinated population. Given the practical and logistical difficulties in conducting randomised controlled trial and cohort study regularly and biases in traditional case-control study, test-negative design (TND) has been used for monitoring IVE in different age groups in various settings such as in the I-MOVE multicentre case-control study annually since 2007. Since the winter in 2014-15, IVE study in the community using TND has been started and planned to estimate IVE during winter and summer peak seasons of influenza in 2014-16. While laboratory testing of cases for influenza infection are yet to be completed, preliminary results are reported here. Patients presenting with acute respiratory illness to private out-patient clinics participating in the study within 72 hours of onset of symptoms will be recruited and tested for influenza virus by rapid antigen test and RT-PCR. Subjects tested positive are defined as test-positive cases while those meet with the same inclusion criteria but test negative are test-negative cases. IVE can then be estimated as in traditional case-control study using one minus adjusted odds ratio of vaccination among test-positive versus test-negative cases. During the winter peak from January through April 2015, 359 subjects were recruited through 16 private out-patient clinics. The subjects aged 9 months to 94 years (median 32 years). Symptoms included fever at 37C or above (60.7%), cough (81.8%), runny nose (79.0%), sore throat (70.1%), headache (70.3%), myalgia or arthralgia (67.9%) and phlegm production (65.4%). While 59 (16.3%) subjects were tested positive for influenza A by rapid test, 71 of 117 (60.2%) subjects randomly tested by RT-PCR were positive for influenza A. Recipients of seasonal influenza vaccine for 2014-15 accounted for 14.2% of all subjects. The preliminary estimates of IVE against influenza A were 34.4% by rapid test and 10.4% by RT-PCR. Our study proved that TND was feasible for monitoring IVE across seasons in community setting with reasonable cost. Information including degree of matching between vaccine strains and epidemic strains, previous vaccination history and medical history readily collected during out-patient consultation could be used to adjust IVE for different factors. Precision of estimates might be improved with larger number of subjects to be recruited by more participating clinics.