EFFECT OF MODEL PARAMETER VARIABILITY ON THE UNCERTAINTY OF REFRIGERATED MICROBIAL SHELF‐LIFE ESTIMATES

Abstract

ABSTRACTMonte Carlo procedures can be used to evaluate the uncertainty of food safety and quality estimations caused by the variability in model parameters. This study describes shelf‐life predictions based on the growth of Lactobacillus sakei in meat using Ratkowsky‐type models, considering the effect of temperature, water activity (Aw) and modified atmosphere. The shelf life predicted when parameter variability was not considered was 7.0 h for a temperature‐only model (Case 1, T = 4C), 184.6 h for a temperature and Aw model (Case 2, T = 4C, Aw = 0.98), 6.4 h for a temperature and CO2 model (Case 3, T = 4C, CO2 = 2,650 ppm) and 241.6 h for a temperature, Aw and CO2 model (Case 4.1, T = 4C, Aw = 0.98, CO2 = 2,650 ppm), whereas 7.4 ± 3.5, 190.4 ± 34.8, 7.5 ± 2.0 and 266.1 ± 65.8 h, respectively, were the values estimated considering parameter variability. Examining the frequency distribution of the predicted shelf life, as well as imposing a 95% confidence that meat will not spoil before its expiration date, leads to a recommended shelf life of 4, 141, 6 and 176 h for Cases 1–4.1, respectively. If the standard deviation (SD) of all model parameters in Case 4.1 could be lowered by 10, 50 and 90%, the recommended shelf‐life time would increase from 176 to 189, 198 and 202 h, respectively (Case 4.6). The analysis of the impact of lowering the individual SD of the model parameters (Cases 4.2–4.5) showed an even lower impact. This suggests that lowering the uncertainty of microbial shelf‐life predictions is very difficult when multiple factors are considered in the microbial model used for this estimation.PRACTICAL APPLICATIONSNew and practical methodologies are needed to evaluate the uncertainty of estimations associated with the variability in the parameters of food process engineering models to satisfy new requirement in food regulations and high consumer expectations for product safety and quality. For example, processors face an increasing pressure from regulatory agencies and customers to declare the shelf life on the product label. Therefore, processors need to know the time that products will retain the quality desired and the safety expected. However, shelf life depends on many factors often described by statistical distributions and historical records such as temperature during storage and distribution. Therefore, deterministic calculation methods are not useful nor is it possible to test all possible scenarios in laboratory determinations of shelf life. The methodology here presented can be used to generate histograms of expected shelf‐life considering the variability in as many parameters as necessary. These Monte Carlo and predictive microbiology procedures to determine shelf‐life uncertainty can be used to reduce the risk of reaching consumers with unsafe or spoiled products.