Determination of Quantitative Food Consumption Levels for Use in Microbial Risk Assessments: Cheddar Cheese as an Example

Abstract

Microbial risk assessment (MRA) is becoming increasingly used in the management of food safety because it can be used to quantify risks and help rank intervention strategies. The exposure assessment components of the assessments have become complex with many aspects of the contamination, survival, and growth of a pathogen in a food being taken into consideration. Insufficient consumption data constitutes an important data gap and consequently one of many sources of uncertainty in MRA even though the effects of uncertainty are smaller than those affecting bacterial concentration in foods. Therefore, food consumption data also play an important role in exposure assessment of MRA. In the United States, there are large-scale, nationwide sets of consumption data available for use in MRA, i.e., the National Health and Nutrition Examination Survey (NHANES). Newly released dietary interview data in the NHANES 2001 to 2002 survey show that it has been redesigned and that the data were sufficiently updated from previous versions to have more value for MRAs. We propose a model that can effectively use the new data sets and be incorporated into MRAs, using as an example consumption of Cheddar cheese/American-type cheese. This model included the prevalence of food eaten as well as the amount and frequency. We determined the amount of Cheddar/American cheese consumed per day with probability distribution (e.g., lognormal distribution). These could be further determined by gender, age, pregnancy, and combination food type, which we plan to do in the future. The frequency of the range of serving numbers for Cheddar/American cheese consumed per person per day and prevalence as the proportion of a population (e.g., survey respondents) eating a certain food in a day are also presented. Unlike traditional published mean values, the results of this model provide probability distribution intakes that can be compared with mean and median intakes. This allows values in the upper percentiles to be considered for inclusion in MRAs. We believe this simulation model can be adapted with different variables applicable to different foods, pathogens, and specific health risk population groups.