Estimation of sample sizes for pooled faecal sampling for detection of Salmonella in pigs

Abstract

Salmonella infection in breeding pigs was the subject of a European survey in 2008. The prevalence of pig-breeding holdings infected with Salmonella was determined by microbiological culture of pooled pen faecal samples. The objective of this study was to estimate the sensitivity of pooled faecal sampling and to calculate the required sample sizes. To do this, individual and pooled faecal samples were collected from a sample of pens from nine farms. Bayesian methods were used to estimate the sensitivity of individual and pooled faecal sampling, and the degree of clustering of Salmonella at the pen level. Sample sizes were then calculated for various values of design prevalence, taking into account the clustering. Pooling was highly efficient compared to individual sampling, e.g. with 18 pooled samples required to detect a 10% prevalence with 95% certainty, compared to 35 individual rectal samples. We recommend that pooled sampling is used for detection of Salmonella in pigs. Results were influenced by the degree of clustering at pen level, and it is important to take this into account both in the estimation of appropriate sample sizes and the estimation of prevalence from pooled sample data.