Effect of Input Data Variability on Estimations of the Equivalent Constant Temperature Time for Microbial Inactivation by HTST and Retort Thermal Processing

Consumer demand for food safety and quality improvements, combined with new regulations, requires determining the processor's confidence level that processes lowering safety risks while retaining quality will meet consumer expectations and regulatory requirements. Monte Carlo calculation procedures incorporate input data variability to obtain the statistical distribution of the output of prediction models. This advantage was used to analyze the survival risk ofMycobacterium aviumsubspeciesparatuberculosis(M. paratuberculosis) andClostridium botulinumspores in high-temperature short-time (HTST) milk and canned mushrooms, respectively. The results showed an estimated 68.4% probability that the 15 sec HTST process would not achieve at least 5 decimal reductions inM. paratuberculosiscounts. Although estimates of the raw milk load of this pathogen are not available to estimate the probability of finding it in pasteurized milk, the wide range of the estimated decimal reductions, reflecting the variability of the experimental data available, should be a concern to dairy processors. Knowledge of theC. botulinuminitial load and decimal thermal time variability was used to estimate an 8.5 min thermal process time at 110 °C for canned mushrooms reducing the risk to 10 ^-9 spores/container with a 95% confidence. This value was substantially higher than the one estimated using average values (6.0 min) with an unacceptable 68.6% probability of missing the desired processing objective. Finally, the benefit of reducing the variability in initial load and decimal thermal time was confirmed, achieving a 26.3% reduction in processing time when standard deviation values were lowered by 90%.