Evaluation of High Pressure Processing Kinetic Models for Microbial Inactivation Using Standard Statistical Tools and Information Theory Criteria, and the Development of Generic Time-Pressure Functions for Process Design

Abstract

Generic kinetic models for microbial inactivation by high pressure processing (HPP) would accelerate the development of commercial applications. The aim of this work was to develop a generic model obtained by fitting peer-reviewed microbial inactivation data (124 kinetic curves) to first-order kinetics (LKM), Weibull (WBLL), and Gompertz (GMPZ) primary and secondary models. Standard statistics (coefficient of determination (R 2), variance, residuals plots, experimental vs. predicted plots) and information theory criteria (Akaike Information Criteria, AIC; Akaike differences, ∆AIC i , Bayesian Information Criteria) determined their goodness of fit. Standard statistics showed no differences between WBLL and GMPZ, whereas information theory criteria identified WBLL as the best model (lowest AIC i value, 61.3 % of cases). LKM performed poorly according to all statistics (e.g., ∆AIC i > 10, 58.1 % of cases). The dispersion of model parameters prevented the derivation of a secondary model for the whole dataset, but clear trends and sufficient data (56 kinetic curves) were found to develop one for milk. A secondary WBLL model (b′ = 0.056–2.230, n = 0.758 − 0.403; 150–600 MPa) was the best alternative (AIC i = 183.8). A GMPZ model yielded similar predictions, but registered ∆AIC i = 19.3 reflecting its larger number of parameters (p = 8). Selecting datasets with pressure holding times of commercial interest (t ≤ 10 min) yielded different parameter estimates for the generic WBLL model (b′ = 0.079–1.859, n = 1.340–0.557; 300–600 MPa). In conclusion, information theory criteria complemented standard statistics, and the simpler WBLL secondary model (p = 4) provided a product-specific time-pressure function of industrial relevance.