<p>The objectives of this project are </p>
<p>1) to build and parameterize a quantitative risk model of the Salmonella prevalence from the farm to the wholesale pork distribution, and </p>
<p>2) to apply the model in evaluating the relative contribution of each Salmonella source, especially lymph nodes, to human food-borne risk. </p>
<p>This project will produce information and analysis which can be presented directly to food safety regulators and industry decision makers about the relative impact of lymph node contamination. It will also produce at least one peer-reviewed scientific journal article. Of longer lasting value will be the quantitative model resulting from this project. Using currently available data it will model impacts of changing pathogen prevalence along the farm to wholesale pork continuum. It will be available for future analysis on other pathogens or other intervention questions. </p>
<p>Project objectives</p>
<p>Objective A. To build and parameterize a quantitative risk model of the Salmonella flow through pork production from the farm to the wholesale pork distribution. </p>
<p>Objective B. To exercise and test the model evaluating the relative contribution of each Salmonella source to human foodborne risk. Procedures to achieve these objectives Build model Human health is the end-point of the quantitative risk model. A forward "farm-to-fork" model will be built based on the flow of pork and Salmonella through the process. </p>
<p>The goal of the model is to simulate the number of pork attributable human cases of Salmonella as a function of changing Salmonella prevalence, abattoir interventions, and further product processing. It will incorporate all available Salmonella data on the input-output relationships using data from the meta-analysis described below and empirical data from the companion proposal; "Evaluating the sources of Salmonella after carcass chilling". As stated in the research summary, the industry is under increasing pressure to stamp out all sources of Salmonella. However, controlling every source is not feasible or beneficial to public health. This project will produce a tool for the entire industry that can quantitatively demonstrate the risk impact of various control or mitigation options. It can be used for other sources that just lymph nodes and other pathogens besides Salmonella.</p>
<p>NON-TECHNICAL SUMMARY: <br/>This project will produce information and analysis which can be presented directly to food safety regulators and industry decision makers about the relative impact of lymph node contamination. It will also produce at least one peer-reviewed scientific journal article. Of longer lasting value will be the quantitative model resulting from this project. Using currently available data it will model impacts of changing pathogen prevalence along the farm to wholesale pork continuum. It will be available for future analysis on other pathogens or other intervention questions. This project will specifically meet the objective of developing a risk assessment model to determine the relationship between the prevalence of Salmonella in various lymph nodes, including deep tissue lymph nodes and the risk to human illness. All meat producing industries are
under extreme pressure to reduce Salmonella illness rates in humans. There seems to be a strongly held opinion in the US that most salmonellosis is due to meat. The public and regulators seem to expect the pork industry to virtually eliminate Salmonella from all sources. This perspective is not "risk-informed" and could lead to unnecessary efforts in non beneficial areas. This project will result in a quantitative risk model which can aid pork producers and packers in decision making for many food safety interventions. Utilization of limited resources is always in the industry and the consumer's best interest. As the Nuclear Regulatory commission stated in 1995, Probabilistic risk assessment should be used to reduce unnecessary conservatism associated with current regulatory requirements and guides. To avoid excessive regulatory burdens stemming from consumer anxiety the industry must
prepare science-based, risk-informed analysis of potential control options. This research is a unique application of probabilistic risk assessment techniques commonly accepted in other fields such as environmental health and engineering. It is important to help decision makers choose among a variety of control options and pathogen sources. The model can be expanded to address pathogens other than Salmonella. This type of work is vital to help interpret and prioritize findings of other scientific disciplines such as microbiology. The recovery of Salmonella in lymph nodes or even on farms does not mean that is a good place to implement controls. For example, a study of the Danish on-farm Salmonella control program found that, contrary to popular opinion, most improvements in public health resulted from the post harvest instead of preharvest investments. Findings from this project will be
presented at selected pork and public health meetings as alternative funds allow. It will result in at least one, maybe two peer-reviewed scientific publications. Specifically, as former Deputy Undersecretary for Food Safety (USDA FSIS), Dr. Hurd will take the result directly to regulators at FSIS. Dr. Dickson serves on the USDA secretary's National Advisory Committee on Meat and Poultry Inspection who deliberates and advises on policies options addressed by this project.
<p>APPROACH: <br/>A meta-analysis (quantitative literature review) will be implemented. This analysis will attempt to identify all quantitative research regarding the levels of Salmonella at all points through the pork production chain. Reliability and generalizability will be evaluated from study design, source of samples and other evidence-based medicine criteria. The primary outcomes are the prevalence or level of Salmonella on farm feces, in lairage pen, on carcasses in different processing locations, etc, and the transmission rate of Salmonella from one process to another. If multiple estimates are available the sources will be weighted as to reliability and generalizability and an overall parameter estimate derived. One of the more difficult data parameters to estimate is attribution; what portion of all human salmonellosis is due to pork contamination. Using a
sophisticated model with some vital and unique assumptions, Denmark estimated that 9%-15% of human salmonellosis is attributable to pork (Hald et al., 2004). Hoffman et al. 2007 estimated about 6% of all salmonellosis cases were due to pork. To address the uncertainty in attribution, the model will calculate the estimated number of human cases, but also estimate relative changes in risk, as we have done in a study of poultry risk (Singer et al., 2007). Data analysis By sensitivity analysis we will quantify how changes in the Salmonella prevalence at various points in the chain would change Salmonella illness risks in the US population. Especially, we want to quantify how changes in the level and prevalence of Salmonella in various lymph nodes would change the risk, and how changes in pre-harvest and post-harvest processes affect the level and prevalence of Salmonella in various lymph
nodes. We will calculate how much (%) DTLN must be reduced to effect a similar 1 point reduction in human illness. After refinement and validation, the model will be used to evaluate various interventions by determining their mitigation elasticity. The relative change in Salmonella prevalence will be compared to the relative change in predicted human cases. These comparisons will be made for various points along the continuum e.g. pigs leaving the farm, after lairage, after the cooler, etc. Quality control plan Model validation will be first conducted to ensure that the model represents and correctly reproduces the behaviors of the real world system. In order to validate, the predicted number of salmonellosis cases from public health surveillance, statistics reported by CDC will be used. Additionally, the model will be validated by exercising it with extreme values of various input
parameters. This method is conducted by setting one of the many parameters as its extreme upper then lower bound. The output of the model (risk) should then change in a similar and relatively predictable fashion. This method can identify incorrect mathematical relationships in the model.