Testing an ecological model for transmission of Salmonella enterica in swine production ecosystems using genotyping data.

An ecological model for transmission of Salmonella enterica in swine production ecosystems was developed, identifying host species, environmental reservoirs, and temporal, spatial, and functional (i.e., stage of production) dimensions. It was hypothesized that transmission was most likely within spatial and functional compartments, between hosts of the same species and abiotic compartments of the same type. Eighteen swine production systems in Illinois, USA, were sampled in four collection cycles (1998, 1999, 2000, 2003). There were 11,873 samples collected, including feces from swine and other mammals and birds, and samples from insects, pen floors, boots, feed, and water. The 460 Salmonella isolates obtained were genotyped using repetitive sequence PCR with three primers-REP, BOX, and ERIC. All isolates from 2000 and 2003 were serotyped, as well as a subsample from 1998 and 1998. Genetic relatedness was estimated from the similarity of fragmentation patterns after gel electrophoresis of PCR products. Cluster analysis identified genetically related isolates. Linking of isolates in tight clusters (similarity >or=85%) was viewed as evidence for transmission. Five farms had a sufficient number of tight clusters for hypothesis testing. The factors most differentiating isolates genetically were farm of origin and time of sampling. Isolates were also differentiated genetically by site, building, room, and pen. There was no consistent association of genotype with stage of production or host/environment reservoir. Serotype analysis confirmed that Salmonella lineages were differentiated by visit and site. Thus, Salmonella transmission was primarily over short distances, i.e., within the same pen or room, with some transmission between rooms and buildings on the same site, but with limited transmission between sites. Transmission was observed across a variety of ecological niches represented by different host species and environmental reservoirs. Genetic differences over time reflected multiple introductions into the ecosystem of different Salmonella genotypes, as well as evolutionary changes within lineages. Intervention strategies to reduce Salmonella prevalence within swine production ecosystems would be best targeted at maintaining spatial barriers to transmission, whereas intervention targeted at specific biological hosts or environmental reservoirs is less likely to be effective.

Influence of Salmonella in pigs preharvest and during pork processing on human health costs and risks from pork.

Salmonellosis in humans is a costly disease traditionally assumed to be associated with exposure to contaminated food. We have developed a farm-to-fork model that allows estimation of the human health costs and risks associated with Salmonella in pork. This analysis focuses on the stages of the pork production chain up to the point of producing a chilled pork carcass. The model predicts the number of human cases of salmonellosis associated with pork (mean, 99,430; 90% confidence interval, 20,970 to 245,560) and the corresponding social costs (mean, $81.53 million; 90% confidence interval, $18.75 million to $197.44 million). Sensitivity and scenario analyses suggest that changes in Salmonella status during processing are more important for human health risk and have a higher benefit:cost ratio when compared with on-farm strategies for Salmonella control. Specifically, benefit:cost ratios are less than 1 (indicating they are not likely to be profitable from a social economic perspective) for the on-farm strategies of vaccination and meal feeding, whereas rinsing carcasses at various temperatures with and without sanitizer all have benefit:cost ratios greater than 1 (indicating they are profitable from a social economic perspective). This type of modeling is useful for evaluation of the relative cost effectiveness of interventions at different points in the food chain when allocating limited food safety dollars and is best used for examining trends and alternative strategies rather than for providing definitive dollar value estimates of risk. The dollar value estimates must be considered in the context of the wide confidence intervals.

Comparison of pulsed field gel electrophoresis and repetitive sequence polymerase chain reaction as genotyping methods for detection of genetic diversity and inferring transmission of Salmonella.

Pulsed field gel electrophoresis (PFGE) using restriction enzymes AvrII, SpeI, and XbaI, and repetitive sequence polymerase chain reaction (Rep-PCR) using BOX, ERIC, and REP primers, were compared with respect to their ability to detect genetic differences among 68 Salmonella isolates from nine Illinois swine farms. Both genotyping methods had high reproducibility of fragment numbers (reliability>0.9) and sizes (reliability>0.85), and sizes [Formula: see text], and produced approximately the same number of DNA fragments, but Rep-PCR fragment profiles had considerably greater variation. Genetic distances between isolates were calculated from fragment size matching. There was good agreement between the genetic distance matrices for the composite (3-enzyme and 3-primer) methods (Mantel's r=0.83). PFGE detected slightly greater variation in genetic distances among isolates, but failed to differentiate seven pairs of isolates, three of which were sampled at least 1 month apart and therefore unlikely to be truly identical genetically. In contrast, Rep-PCR identified no isolates as genetically identical. In cluster analyses based on genetic distances, there were moderate differences between PFGE and Rep-PCR (about 2/3 agreement in tight cluster membership). Both PFGE and Rep-PCR were able to differentiate isolates of the same serotype. However, some serotypes (Agona, Anatum, Derby, Infantis, Worthington) were distributed across clusters. There was less agreement between individual primer/enzyme and composite results for Rep-PCR than for PFGE. This greater independence of results for individual primers for Rep-PCR accounted in part for the greater discriminative ability of the composite method. Both composite methods indicated that most Salmonella transmission occurred within a farm and that there was no preference for transmission between specific ecological compartments. Given the equally high reliability of both genotyping methods, the greater discriminative ability of Rep-PCR recommends it as the preferred method for precise detection of transmission links.

Models of antimicrobial resistance and foodborne illness: examining assumptions and practical applications.

Antimicrobial resistance is an issue of increasing global concern. Several investigators have suggested that antibiotic use in food-producing animals is a major contributor to the increasing incidence of antimicrobial-resistant organisms causing illness in humans (F. J. Angulo, K. R. Johnson, R. V. Tauxe, and M. L. Cohen, Microb. Drug Res. 6:77-83, 2000; P. D. Fey, T. J. Safranek, M. E. Rupp, E. F. Dunne, R. Efrain, P. C. Iwen, P. A. Bradford, F. J. Angulo, and S. H. Hinrichs, N. Engl. J. Med. 342:1242-1249, 2000; S. A. McEwen and P. J. Fedorka-Cray, Commun. Infect. Dis. 34(Suppl. 3):S93-S106, 2002; D. L. Smith, A. D. Harris, J. A. Johnson, E. K. Silbergeld, and J. G. Morris, Jr., Proc. Natl. Acad. Sci. USA 99:6434-6439, 2002; D. G. White, S. Zhao, R. Sudler, S. Ayers, S. Friedman, S. Chen, P. F. McDermott, D. D. Wagner, and J. Meng, N. Engl. J. Med. 345:1147-1154, 2001; W. Witte, Science 279:996, 1998). In this paper, we discuss this and other assumptions relevant to a quantitative risk assessment model for salmonellosis in humans. We also discuss other important aspects of modeling food safety and food-associated antimicrobial resistance risk to humans. We suggest that the role of food-producing animals in the origin and transmission of antimicrobial resistance and "foodborne" pathogens has been overestimated and overemphasized in the scientific literature; consequently, nonfoodborne transmission, including pet-associated human cases, has been underemphasized. Much evidence exists for the potential contribution to infectious disease that may be of human or pet origin (that may contact humans through food but not be of a food origin). Risk analyses that do not acknowledge the potential for these sources of cross-contamination will understate the contribution that origin has in the realm of foodborne and food-associated diseases (e.g., Salmonella) and the resulting uncertainty levels in the food system, thus leading to biased inferences. We emphasize the importance of evaluating both the foodborne and nonfoodborne transmission risk for salmonellosis and outline the basics of an analytical modeling approach in food safety with examples to illustrate strengths and limitations in the modeling. Examples illustrate, on a simplistic level, how varying assumptions and other inputs can influence the output of food-associated quantitative risk models.

Distribution of Salmonella in swine production ecosystems.

The objective of this 2-year field survey was to sample multiple ecological compartments within swine production systems to identify potential sources of Salmonella infection for swine. Twelve single-site production systems within Illinois were identified by slaughter sampling to have detectable Salmonella in swine and therefore selected for study. There were four visits to each farm during a 5-month period. Fecal samples were obtained from swine and other wild and domestic mammals. Arthropods and environmental samples of feed, water, pen floors, boots, and bird feces were also collected. All 8,066 samples obtained were cultured to detect Salmonella. Salmonella was detected on 11 of the 12 farms. There were 206 positive cultures, including samples from swine (83), pen floors (54), boots (32), flies (16), mice (9), cats (3), and birds (3). Swine shedding Salmonella in feces were detected on 9 of the 12 farms. The more Salmonella-abundant ecological compartments were cats (12% of samples positive), boots (11%), bird feces (8%), flies (6%), and mice (5%); 2.1% of 4,024 swine samples were positive. All 221 feed samples were negative for Salmonella. There was a correlation between a farm having a high prevalence of shedding Salmonella in pigs and a high abundance on pen floors, flies, and boots. The most common serotypes detected were Derby, Agona, Worthington, and Uganda, which were distributed throughout the ecosystem, suggesting widespread transmission across ecological compartments. The ubiquitous distribution of Salmonella suggests that an effective control strategy must target multiple compartments of the swine production ecosystem.