Estimation of influenza vaccine effectiveness (VE) is complicated by various degrees of mismatch between circulating and vaccine strains each season. We carried out a cohort study to estimate VE of trivalent (TIV) and pandemic influenza vaccines (PIV) in preventing various respiratory outcomes among general practice (GP) patients in England and Wales between 2008 and 2010. Dates of consultations for influenza-like illness (ILI), acute respiratory tract infection (ARTI), lower respiratory tract infection (LRTI) and nasopharyngeal swabs were obtained from the patient-level electronic records of the 100 practices enrolled in a national GP network. Dates of vaccination with TIV and PIV were also extracted. Confounders including age, time period and consultation frequency were adjusted for through Poisson regression models. In the winter of 2008/9, adjusted VE of TIV in preventing ILI was 22.3% (95% CI 13.5%, 30.2%). During the 2009/10 winter VE for PIV in preventing ILI was 21.0% (5.3%, 34.0%). The VE for PIV in preventing PCR-confirmed influenza A/H1N1 (2009) was 63.7% (-6.1%, 87.6%). TIV during the period of influenza circulation of 2008/9 and PIV in the winter of 2009/10 were effective in preventing GP consultations for ILI. The cohort study design could be used each season to estimate VE; however, residual confounding by indication could still present issues, despite adjustment for propensity to consult.
Copyright © 2011 Elsevier Ltd. All rights reserved.