Quantitative Methods for Hazard Characterization of Food-Borne Pathogens

Abstract

Over the last decade, heightened awareness about the consequences of foodborne illnesses has fomented the application of quantitative risk assessment to food safety issues. A four-step paradigm - hazard identification, exposure assessment, hazard characterization, and risk characterization - is commonly followed. Hazard characterization is meant to consider the multifaceted interaction between pathogen, host, and food matrix, but it is reduced in practice to a mathematical function linking an exposure dose to an infection/illness probability. Such an approach has obviously a limited capacity to weigh into a risk assessment potential sources of uncertainty (lack of knowledge) and variability (population heterogeneity).

This study develops an analytical framework that makes possible the quantitative consideration of selected uncertainty and variability elements in microbial hazard characterization. Firstly, the bootstrap method is applied to quantify the sampling error associated with fitting dose-response functions to data from volunteer feeding trials with Campylobacter jejuni and Shigella dysenteriae. The results show that the relevance of theoretical considerations regarding the form of the dose-response function or the resampling scheme depends on the dose range of interest. Further, an epidemiological analysis of FoodNet surveillance data is carried out to estimate the effect of the covariates age and gender on the occurrence of foodborne infection. Specifically, Poisson regression analysis is applied to model Campylobacter, Salmonella, and Shigella rates. While gender does not cause significant differences, the analysis characterizes the risk associated with specific age groups. Setting young adults as the reference group, the highest relative risks of Campylobacter and Salmonella infection are in infants (1.87 and 9.15, respectively), while teenagers and the elderly are associated with the lowest relative risks (0.43/0.51 and 0.71/0.70, respectively). Results for Shigella are less reliable due to questionable model fit. The final part of the study integrates the previous findings into a probabilistic risk assessment that, with risk management in mind, keeps uncertainty and variability separate. Using two-dimensional Monte Carlo simulation and stratification into homogeneous population subgroups, the risk of foodborne Campylobacter infection is calculated for eight different age groups taking into consideration the sampling error attendant to the dose-response function. While age variability turns out to have only a minimal relevance, uncertainty associated with the dose-response function has a major impact on the results.

Overall, this study shows that biological plausibility and epidemiological evidence do not necessarily translate into risk assessment relevance. It is advanced that, in microbial hazard characterization, sources of variability have to be explicitly modeled only to the extent that the magnitude of their effect is large or that they are fundamental to the needs of the risk manager. In contrast, characterization of key sources of uncertainties - in particular of the sampling error associated with the dose-response function - and their consistent propagation throughout a microbial risk assessment appear to be of great importance.