UNCERTAINTY OF MICROBIAL SHELF-LIFE ESTIMATIONS FOR REFRIGERATED FOODS DUE TO THE EXPERIMENTAL VARIABILITY OF THE MODEL PARAMETERS
The uncertainty of shelf-life estimations for refrigerated foods exposed to changing temperature was quantified by considering the contribution of the experimental variability of the model parameters. Assuming the real distribution of the parameters can be replaced by an unknown empirical distribution, this uncertainty was analyzed by a bootstrap methodology. Independent sets of heat transfer and microbial growth parameters were chosen randomly from experimental values. Shelf-life values were estimated for each set, with the variability evaluated using an SD value. The procedure was repeated 10 times to obtain 10 shelf-life SDs. The variation coefficient of these SDs is an absolute measurement of shelf-life variability. At the recommended variation coefficient of 0.1, a sample size of 25 was found acceptable. The shelf-life variability estimated was similar (±1 day) for surimi in cardboard and expanded polystyrene containers with different shelf life and temperature profiles. The method is applicable to other predictions of food stability. PRACTICAL APPLICATIONS
The effect of the experimental variability of model parameters on shelf-life estimations confirmed the need to assess the uncertainty of model predictions. Another beneficial aspect of the approach presented is that it encourages an integrated approach to food safety and shelf life. Decisions on raw materials and modifications to processing operations affect the initial microbial load of a product. Improvement of storage, distribution and retail facilities reduces temperature abuse. Product size, geometry and packaging choice are also aspects that should be considered. The impact on shelf life of all these factors can be examined by the integrated microbial and heat-transfer models of the type here presented. The increasing availability of microbial growth parameters and electronic recorders to monitor temperature during processing, storage and distribution facilitate their implementation. The impact of the experimental variability of model parameters is important. For example, this study suggested the need for microbial determinations with lower experimental variability.
Document Type: Research Article
Affiliations: Departamento de Procesos QuímicosBiotecnológicos y AmbientalesUniversidad Técnica Federico Santa MariaValparaíso, Chile
Publication date: 2010-02-01