The modern U.S. food safety system relies on risk characterization, reduction, and management to provide safe food for consumers. Much of the responsibility for managing risks associated with foodborne illnesses is placed on the food industry, which must incorporate and validate preventive controls whenever a known hazard is reasonably likely to occur. However, because of the biological hazards involved, much of the information needed for a successful preventive control validation originates from independent laboratory-based research groups. The resulting practical challenges include: (1) Science-based evidence for pathogen inactivation developed in laboratory-scale environments often has critically different features from industry-scale applications; (2) Guidelines for predictive modeling- and surrogate-based validations often fail to address how researchers can improve the utility of their research for industrial application; and (3) There is no known standard for how robust a preventive control must be, beyond the expectation that it must be “statistically valid.” Therefore, the overall goal of this dissertation was to improve the utility of food safety research for application in preventive control validations. The specific objectives were to: (1) Quantitatively evaluate current practices for development and application of predictive inactivation models for Salmonella in low-moisture foods and develop a framework for improved practices; and (2) Develop and demonstrate improved statistical tools for the application of Salmonella surrogates in low-moisture preventive control validations. The first step was to define criteria for a robust “statistically valid” standard for evaluating preventive control validations and then evaluate how well current practices are meeting these criteria. Based on an extensive review of the literature, it was demonstrated that the current state-of-practice in predictive microbiology is not yielding models ready for use in preventive controls. This was partially by design, as most instances of inactivation models in relevant literature fill a descriptive rather than predictive function. For models that were clearly intended to inform, or be utilized in, preventive control validations, there were persistent issues of underreporting key model components that would maximize the utility of such models. The issues limiting experimental utility across studies/labs were investigated via a multi-laboratory thermal inactivation study, and a standard template for future thermal inactivation studies was developed to improve research synergy. Subsequently, a widely disseminated, model-based approach to low-moisture food validations, which ignored several core principles of predictive microbiology, was critically evaluated for a baking case study. The results demonstrated significant “fail dangerous” errors resulting from inappropriate application of inactivation models that do not incorporate the critical effect of moisture on Salmonella thermal resistance in low-moisture foods. Non-pathogenic surrogate organisms are commonly used for validating preventive controls in industry; however, much like predictive models, there is no consensus or standard for statistically evaluating a preventive control validation. Therefore, a statistical framework for evaluating reduction performance criteria was developed and tested, including a statistical foundation for translating surrogate-based validation results into likely pathogen outcomes. This framework was demonstrated in a case study encompassing a large dataset comprised of thermal process validations in the almond industry. Overall, this work demonstrated key approaches to improve standards of practice for predictive microbiology and food safety surrogate research, which should improve utility for application to process validations. Additionally, this work demonstrated methods that may help industry improve the design, robustness, and costs of their preventive control validations.
Read
