Staphtox predictor - A dynamic mathematical model to predict formation of Staphylococcus enterotoxin during heating and fermentation of meat products

Abstract

Existing growth models for S. aureus predict growth in relation to temperature, aw/NaCl and pH, and the assessment of probable Staphylococcus enterotoxin (SE) formation is based solely on the number of S. aureus. However, during the production of meat products such as fermented sausages and semi-processed hams, growth of S. aureus is a critical control point in HACCP plans. There is a need to develop a model that evaluates the safety of the product regarding SE formation in relation to the product composition, changes in pH or temperature during the processing and the number of S. aureus in the final product. The objective of the present work is to develop a mathematical model that predicts both the increase in the number of S. aureus and whether SE formation is possible in different meat product processes.A total of 78 experiments were carried out in a meat model system. The experiments covered a range of different temperatures (10–40 °C), pH (4.6–6.0), water phase salt (WPS) (2.2–5.6%) and Sodium nitrite concentrations (0–150 ppm). The meat model system was inoculated with approximately 103 CFU/g of a multi-strain cocktail and incubated at the different temperatures. The cocktail consisted of three strains of S. aureus producing the Staphylococcus enterotoxins A to D (SEA to SED) and a methicillin-resistant strain producing SEG, SEI, SEM, SEN, SEO and SEU. Enumeration of S. aureus was performed several times during the incubation, SE was extracted from samples with >5 log CFU/g, and the SEA-E content was analysed by an ELISA method. Maximum growth rates and lag times calculated from microbiological data, together with temperature, pH, WPS and Sodium nitrite, were used to develop a SE and a growth model. The growth model was developed by training a neural network and the SE model based on logistic regression.The SE and growth models were validated on separate data sets (N = 200 SE model, N = 63 growth model) including both dynamic and static conditions. The SE model predicted all occurrences of toxin formation in the validation data sets. The growth model is a fail-safe model and the prediction errors are comparable to laboratory reproducibility. In conclusion, the models are applicable for predicting the increase in S. aureus and for evaluating if SE formation is likely during processing of meat products. The models are available to producers and other interested parties at www.dmripredict.dk.