A philosophy for the development of kinetic models in predictive microbiology

TL;DR: It is concluded that the experimental strategy proposed offers an efficient method for generating the quantity of data required for the development of reliable kinetic models from which to predict the growth of spoilage and pathogenic organisms of relevance to foods.

Abstract: The development of predictive microbiology is reviewed and specific limitations relating to the generation of kinetic models identified. The issues of variability of response, lag time response, fluctuating environments and their effects, microbial interactions, choice of model for describing bacterial growth curves, and mechanistic versus empirical models are discussed and exemplified using experimental data. A philosophy for the development of reliable predictive kinetic models is developed and the appropriateness of that philosophy examined by simulations and reference to novel and independently published experimental data. Specifically, the use of turbidimetric techniques is advocated for primary model development, methods of calibration to traditional (i.e. viable count) methods demonstrated, and the reliability of that calibration demonstrated. Using that methodology, models for the growth of several strains of Staphylococcus aureus and Listeria monocytogenes are generated. Novel indices of the reliability of models are developed, and used to assess the S. aureus 3b and the L. monocytogenes models, for constant environmental conditions, by comparison to published and novel data. An assessment of the three~parameter (temperature, water activity, pH) square-root model is made using data for the growth of L. monocytogenes. A deliberately minimal experimental design was used to 'test to destruction' the proposed methodology, and revealed potential shortcomings of the lack of replication. It is concluded that the experimental strategy proposed offers an efficient method for generating the quantity of data required for the development of reliable kinetic models from which to predict the growth of spoilage and pathogenic organisms of relevance to foods. Technologies for the transfer of validated, laboratory-generated models to the food industry are demonstrated, and a mechanistic interpretation of the basis of the empirical square-root relationship developed.