Evaluation and validation of extensive growth and growth boundary models for mesophilic and psychrotolerant Bacillus cereus in dairy products (Part 2)

Performance was evaluated for two extensive models to predict growth and growth boundaries of mesophilic and psychrotolerant Bacillus cereus in dairy products. Both models incorporated the inhibitory effect of 11 environmental factors and of their interactions. The two models were calibrated and evaluated using data from 66 and 67 new challenge tests, respectively, conducted with various types of well-characterized dairy products. Additionally, the mesophilic model was evaluated using 139 growth responses from literature (growth/no-growth, lag time, and μ_max values) for 24 different B. cereus strains. The psychrotolerant model was evaluated using 109 growth responses from published studies and including data for 26 strains in dairy products. The predictive performance of the evaluated models was compared with four existing models for mesophilic B. cereus and four different models for psychrotolerant B. cereus. The new mesophilic model had good performance and predicted growth responses in new challenge tests, with bias-/accuracy-factor values of 1.13/1.49 and 80% correct, 17% fail-safe, and 3% fail-dangerous growth/no-growth predictions. With literature data for mesophilic B. cereus, predictions were good with bias-/accuracy-factor values of 0.97/1.36 and 91% correct, 9% fail-safe, and 0% fail-dangerous predictions. The evaluated psychrotolerant model also exhibited good performance in predicting growth responses for new challenge tests, with bias-/accuracy-factor values of 1.07/1.38 and 84% correct, 14% fail-safe, and 2% fail-dangerous predictions for growth/no-growth responses. With literature data for psychrotolerant B. cereus, this model did not acceptably predict growth rates at temperatures <10°C. Therefore, the temperature term of the model was expanded at temperatures from 1°C to 10°C. The performance of the updated psychrotolerant model was markedly improved, achieving bias-/accuracy-factor of 1.07/1.80, and 91% correct, 9% fail-safe, and 0% fail-dangerous predictions. The two new and extensive models offer significant advantages over existing models by including the growth inhibiting effects of more environmental factors and their interactions, resulting in un-biased predictions for a wider range of dairy matrices. These validated models can support management of mesophilic and psychrotolerant B. cereus growth in diverse dairy products, contribute to risk assessments and to optimization of combinations of relevant growth-inhibitory factors during product formulation and innovation.