AUTHOR=Koyama Kento , Aspridou Zafiro , Koseki Shige , Koutsoumanis Konstantinos 

TITLE=Describing Uncertainty in Salmonella Thermal Inactivation Using Bayesian Statistical Modeling

JOURNAL=Frontiers in Microbiology

VOLUME=Volume 10 - 2019

YEAR=2019

URL=https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2019.02239

DOI=10.3389/fmicb.2019.02239

ISSN=1664-302X

ABSTRACT=Uncertainty analysis is the process of identifying limitations in scientific knowledge and evaluating their implications for scientific conclusions. Risk assessors should examine in a systematic way every part of their assessment in order to identify all uncertainties, including those related with the inputs to the assessment and the methods used in the assessment. In the context of microbial risk assessment, the uncertainty in the predicted microbial behaviour can be an important component of the overall uncertainty. Conventional deterministic modelling approaches which provide point estimates of the pathogen’s levels cannot quantify the uncertainty around the predictions. The objective of this study was to use Bayesian statistical modelling for describing uncertainty in predicted microbial thermal inactivation of Salmonella enterica Typhimurium DT104. A set of Salmonella enterica Typhimurium DT104 thermal inactivation data in broth with aw adjusted to 0.75 at 9 different temperature conditions was obtained from the ComBase database (www.combase.cc). Data at 8 temperature conditions were used for model development and one temperature for model validation. A log-linear microbial inactivation was used as a primary model while for secondary modelling, a linear relation between the logarithm of inactivation rate and temperature was assumed. For comparison, data were fitted with a two-step and a global Bayesian regression. Posterior distribution of model’s parameters were used to predict Salmonella thermal inactivation using R, Stan and rstan packages of R software. The model described successfully uncertainty in predicted thermal inactivation. The global regression approach resulted in less uncertain predictions compared to the two-step regression. Combination of the joint posterior distributions allowed to express variables such as cell density with time, total reduction time and inactivation rate as probability distributions at different time and temperature conditions. The validation of the model also showed that most observed data were within the 95% prediction intervals of the model. The model developed using Bayesian regression enabled to describe the uncertainty in predicted thermal inactivation of Salmonella. The model provides prediction in the form of probability distributions and can be used to quantify uncertainty related to model fitting in risk assessment studies.