[Estimation of influenza vaccine effectiveness using routine surveillance data]

The continuous antigenic drift of influenza viruses requires annual adaptation of the vaccine. Protection depends largely on the match of the variants represented in the vaccine with the viruses actually known to be in circulation and may differ considerably from season to season. Therefore studies to assess the efficacy and effectiveness of the vaccine are conducted rather sporadically on an annual basis and it would be desirable to make use of routinely available data from surveillance programs. We compared two different approaches: (1) the "screening method" where cases are identified from laboratory data and controls are taken from data on vaccination rates and (2) a second method that uses the same cases, but controls were influenza-negative individuals with influenza-like illness (also identified from laboratory data). The sensitivity of the methods to confounders that were considered as relevant was tested with a simulation. Both methods were applied to the data of the German influenza surveillance data of the season 2004/2005. The estimated effectiveness over all age groups was rather low with both methods, but comparable with other estimations from the literature. We observed differences in certain age groups between the methods as well as large differences between particular age groups within one method. Possible explanations are random variations due to low numbers in age strata and other influences not yet considered. Therefore the estimations should be interpreted with care; however, relative comparisons among seasons may still be meaningful.