Council of Australian Food Technology Associations
Food Australia vol:59 issue:9 pages:427-432
To minimise and estimate food poisoning risks, it is necessary that the growth ability of pathogenic microorganisms in food products can be predicted. Growth/no growth models are being developed that predict the interface between environmental conditions that allow growth and conditions that do not. In this paper, the differences and limitations of existing growth/no growth model types are illustrated for a case study describing the growth probability of Listen'a monocytogenes under conditions typical for mayonnaise based salads. A data set embeds the effects of pH (5.0-6.0), water activity (0.960-0.990) and acetic acid percentage (0-0.8% (w/w)), narrow intervals between the factor levels and many replicates (usually 20). Thus information was gained not only about presence/absence of growth, but also about probability of growth. Among three existing groups of growth/no growth models considered: (i) probabilistic models, (ii) a deterministic approach separating growth/no growth/uncertainty regions, and (iii) deterministic approaches, only the probabilistic models fully describe the data, since they reflect the incremental increase in growth probability present in the experimental growth/no growth interface. The second group of models represented by the Minimum Convex Polyhedron (MCP) method, can partially describe this incremental increase if an extension proposed by the authors is used. Models developed by the third method are not able to describe any increments in growth probability between 0 (no probability of growth) or 1 (certain growth). An MCP based on total acid data is not capable of describing non-convexity in the interface, in contrast to an MCP based on undissociated acid data.