Serological cross-sectional studies on salmonella incidence in eight European countries: no correlation with incidence of reported cases

BACKGROUND Published incidence rates of human salmonella infections are mostly based on numbers of stool culture-confirmed cases reported to public health surveillance. These cases constitute only a small fraction of all cases occurring in the community. The extent of underascertainment is influenced by health care seeking behaviour and sensitivity of surveillance systems, so that reported incidence rates from different countries are not comparable. We performed serological cross-sectional studies to compare infection risks in eight European countries independent of underascertainment. METHODS A total of 6,393 sera from adults in Denmark, Finland, France, Italy, Poland, Romania, Sweden, and The Netherlands were analysed, mostly from existing serum banks collected in the years 2003 to 2008. Immunoglobulin A (IgA), IgM, and IgG against salmonella lipopolysaccharides were measured by in-house mixed ELISA. We converted antibody concentrations to estimates of infection incidence ('sero-incidence') using a Bayesian backcalculation model, based on previously studied antibody decay profiles in persons with culture-confirmed salmonella infections. We compared sero-incidence with incidence of cases reported through routine public health surveillance and with published incidence estimates derived from infection risks in Swedish travellers to those countries. RESULTS Sero-incidence of salmonella infections ranged from 56 (95% credible interval 8-151) infections per 1,000 person-years in Finland to 547 (343-813) in Poland. Depending on country, sero-incidence was approximately 100 to 2,000 times higher than incidence of culture-confirmed cases reported through routine surveillance, with a trend for an inverse correlation. Sero-incidence was significantly correlated with incidence estimated from infection risks in Swedish travellers. CONCLUSIONS Sero-incidence estimation is a new method to estimate and compare the incidence of salmonella infections in human populations independent of surveillance artefacts. Our results confirm that comparison of reported incidence between countries can be grossly misleading, even within the European Union. Because sero-incidence includes asymptomatic infections, it is not a direct measure of burden of illness. But, pending further validation of this novel method, it may be a promising and cost-effective way to assess infection risks and to evaluate the effectiveness of salmonella control programmes across countries or over time.