Proceedings of the 2013 annual meeting of the Netherlands Epidemiology Society
Volume 1 Issue S1 Abstract 16
M. Darvishian, University Medical Center Groningen, University of Groningen, the Netherlands
G. Gefenaite, University Medical Center Groningen, University of Groningen, the Netherlands
R.M. Turner, Medical Research Council Biostatistics Unit, Institute of Public Health, Cambridge, UK
P. Pechlivanoglou, University Medical Center Groningen, University of Groningen, the Netherlands
W. Van der Hoek, National Institute for Public Health and the Environment, the Netherlands
E. Van den Heuvel, Department of Epidemiology, University Medical Center Groningen, University of Groningen, the Netherlands
E. Hak, University Medical Center Groningen, University of Groningen, the Netherlands
Currently, existing meta-analyses investigating the real-world influenza vaccine effectiveness among community-dwelling elderly provide conflicting results due to lack of randomized clinical trials and presence of bias in observational studies. This study aimed to assess influenza vaccine effectiveness in non-institutionalized elderly by using a bias-adjusted meta- analysis.
Fourteen cohort studies were included in the analysis. All potential sources of biases were categorized into internal bias and external bias and were identified using self-administered bias-checklists. Magnitude and uncertainty of biases in each study were elicited using expert opinion. Biases for each study were first combined across assessors using median pooling, and then across studies using random-effects meta-analysis.
After bias-adjustment, the effect estimates of different outcomes (laboratory-confirmed influenza, influenza-like illness (ILI), hospitalization for influenza and/or pneumonia and all- cause mortality) did not substantially change with largest absolute difference between unadjusted and bias-adjusted odds ratios for all-cause mortality (OR 0.51; 95% CI: 0.40-0.65 vs. OR 0.64; 95% CI: 0.44-0.92). Influenza vaccine also remained statistically significantly effective in preventing hospitalization for influenza and/or pneumonia. However, the vaccine did not show statistically significant effectiveness against ILI after adjusting for the biases. Confidence intervals widened reflecting uncertainties about the biases. The level of heterogeneity reduced considerably after adjustments with no remaining significant
evidence of variation between studies (the largest reduction for all-cause mortality, I2=91%, P<0.01 vs. I2=4%, P=0.39).
Seasonal influenza vaccine is effective in reducing hospitalization for influenza and/or pneumonia and all-cause mortality even after incorporating large amount of uncertainty due to biases.
Published: 06 Jun, 2013