Implementation of statistical tools to support identification and management of persistent Listeria monocytogenes contamination in smoked fish processing plants.

Listeria monocytogenes persistence in food processing plants is a key source of postprocessing contamination of ready-to-eat foods. Thus, identification and elimination of sites where L. monocytogenes persists (niches) is critical. Two smoked fish processing plants were used as models to develop and implement environmental sampling plans (i) to identify persistent L. monocytogenes subtypes (EcoRI ribotypes) using two statistical approaches and (ii) to identify and eliminate likely L. monocytogenes niches. The first statistic, a binomial test based on ribotype frequencies, was used to evaluate L. monocytogenes ribotype recurrences relative to reference distributions extracted from a public database; the second statistic, a binomial test based on previous positives, was used to measure ribotype occurrences as a risk factor for subsequent isolation of the same ribotype. Both statistics revealed persistent ribotypes in both plants based on data from the initial 4 months of sampling. The statistic based on ribotype frequencies revealed persistence of particular ribotypes at specific sampling sites. Two adaptive sampling strategies guided plant interventions during the study: sampling multiple times before and during processing and vector swabbing (i.e., sampling of additional sites in different directions [vectors] relative to a given site). Among sites sampled for 12 months, a Poisson model regression revealed borderline significant monthly decreases in L. monocytogenes isolates at both plants (P = 0.026 and 0.076). Our data indicate elimination of an L. monocytogenes niche on a food contact surface; niches on nonfood contact surfaces were not eliminated. Although our data illustrate the challenge of identifying and eliminating L. monocytogenes niches, particularly at nonfood contact sites in small and medium plants, the methods for identification of persistence we describe here should broadly facilitate science-based identification of microbial persistence.