Probabilistic Model for Process Control Measure Evaluation, a Practical Case

The objective of this study was to compare the thermal inactivation of Salmonella spp. and its associated surrogate microorganism, Enterococcus faecium, during the drying step of an extrusion process. The log reduction was calculated using two different methods: deterministic and probabilistic. The main stress factor evaluated was the effect of temperature on the two microorganisms.

Cat kibbles, independently inoculated with E. faecium and a cocktail of Salmonella strains, were heat-treated at three different temperatures, which allowed calculation of model parameters: D-values and z-values. The time/temperature profile of the drying process was evaluated and recorded for three different production batches. These time/temperature profiles, combined with the data obtained from the Thermal Death Time (TDT) study, allowed estimation of the log reduction in Salmonella spp. and E. faecium for the three batches, using both the deterministic and probabilistic models. Finally, model estimations for E. faecium were compared with the results of an in-plant process challenge test where the surrogate microorganism had been used. The log reduction of Salmonella spp. was deduced from the log reduction of the surrogate, taking uncertainty into account. The deterministic approach has the advantage of rapidly being understood by non-model-users, but includes variability and uncertainty only through an empirical safety margin (generally retained as 1-log). The probabilistic method provides the probability of a process to achieve the targeted log reduction, including a quantitative evaluation of variability and uncertainty.