Distributions of microorganisms in foods and their impact on food safety

The physical distributions of pathogens in foods influence the likelihood that a food product will cause illness, but knowledge about the physical distribution of microorganisms in foods and especially the heterogeneity therein is scarce. This Ph.D. research aims to increase the knowledge of microbial distributions in foods and therewith to provide better insights in their impact on public health and food safety management activities. The research covers both theoretical investigations and practical experiments, focusing on powdered infant formula (PIF) as a suitable model product and the opportunistic pathogen Cronobacter spp. as a relevant microorganism in the practical experiments. The impact of spatial distributions of microorganisms, like homogeneous or more clustered distributions, on public health was investigated. Infrequent high doses were shown to mainly determine the probability of illness and also to dominate the arithmetic mean (mean of the counts) expressing the level of microorganisms present. The distribution of Cronobacter spp. in two industrial batches of PIF (a recalled batch and a reference batch) was quantified in detail. Additionally, batches of PIF on lab scale with well-mixed and localised contaminations of Cronobacter sakazakii were enumerated. In the recalled batch, the sample units were taken in the course of the filling time and the results showed that Cronobacter spp. were heterogeneously distributed. On local-scale, clusters of cells varying between 3 and 560 cells per cluster were present sporadically. Discrete and continuous statistical distributions were compared to model the enumeration data of the industrial and laboratory scale batches. Batches with low counts including zeros were fitted best by the Poisson-Lognormal distribution and Negative Binomial distribution. According to criteria proposed to compare the suitability of statistical distributions to model microbial distributions in foods, these two distributions had already been selected to be the most suitable candidates. Furthermore, the performances of random and systematic sampling ware compared to detect a localised contamination in a batch of food. Our calculations showed that systematic sampling rather than random sampling improved the sampling performance. Moreover, taking many small sample units systematically increased the probability to detect the localised contamination. Another systematic sampling strategy evaluated was stratified random sampling. Using the enumeration data of the recalled batch, stratified random sampling appeared to improve the detection probability of Cronobacter spp. as compared to random sampling. Generally, taking more and smaller sample units, while keeping the total sampling weight constant, improved the performance of the sampling plans. The insights obtained in this thesis are considered to be relevant to a wide variety of dry products and to an extent also to other structured foods. They should be of use to food business operators to improve sampling and testing to verify control of their operation as well as to assess compliance of final products with food safety standards and guidelines before marketing. The results may equally be useful to governmental bodies setting and enforcing food safety standards (such as microbiological criteria) and conducting microbiological risk assessment.