Sensitivity analysis in quantitative microbial risk assessment.

The occurrence of foodborne disease remains a widespread problem in both the developing and the developed world. A systematic and quantitative evaluation of food safety is important to control the risk of foodborne diseases. World-wide, many initiatives are being taken to develop quantitative risk analysis. However, the quantitative evaluation of food safety in all its aspects is very complex, especially since in many cases specific parameter values are not available. Often many variables have large statistical variability while the quantitative effect of various phenomena is unknown. Therefore, sensitivity analysis can be a useful tool to determine the main risk-determining phenomena, as well as the aspects that mainly determine the inaccuracy in the risk estimate. This paper presents three stages of sensitivity analysis. First, deterministic analysis selects the most relevant determinants for risk. Overlooking of exceptional, but relevant cases is prevented by a second, worst-case analysis. This analysis finds relevant process steps in worst-case situations, and shows the relevance of variations of factors for risk. The third, stochastic analysis, studies the effects of variations of factors for the variability of risk estimates. Care must be taken that the assumptions made as well as the results are clearly communicated. Stochastic risk estimates are, like deterministic ones, just as good (or bad) as the available data, and the stochastic analysis must not be used to mask lack of information. Sensitivity analysis is a valuable tool in quantitative risk assessment by determining critical aspects and effects of variations.

Stepwise quantitative risk assessment as a tool for characterization of microbiological food safety.

This paper describes a system for the microbiological quantitative risk assessment for food products and their production processes. The system applies a stepwise risk assessment, allowing the main problems to be addressed before focusing on less important problems. First, risks are assessed broadly, using order of magnitude estimates. Characteristic numbers are used to quantitatively characterize microbial behaviour during the production process. These numbers help to highlight the major risk-determining phenomena, and to find negligible aspects. Second, the risk-determining phenomena are studied in more detail. Both general and/or specific models can be used for this and varying situations can be simulated to quantitatively describe the risk-determining phenomena. Third, even more detailed studies can be performed where necessary, for instance by using stochastic variables. The system for quantitative risk assessment has been implemented as a decision supporting expert system called SIEFE: Stepwise and Interactive Evaluation of Food safety by an Expert System. SIEFE performs bacterial risk assessments in a structured manner, using various information sources. Because all steps are transparent, every step can easily be scrutinized. In the current study the effectiveness of SIEFE is shown for a cheese spread. With this product, quantitative data concerning the major risk-determining factors were not completely available to carry out a full detailed assessment. However, this did not necessarily hamper adequate risk estimation. Using ranges of values instead helped identifying the quantitatively most important parameters and the magnitude of their impact. This example shows that SIEFE provides quantitative insights into production processes and their risk-determining factors to both risk assessors and decision makers, and highlights critical gaps in knowledge.

A data analysis of the irradiation parameter D10 for bacteria and spores under various conditions.

This paper provides approximate estimates for the irradiation parameter D10 to globally predict the effectiveness of any irradiation process. D10 is often reported to depend on many specific factors, implying that D10 cannot be estimated without exact knowledge of all factors involved. For specific questions these data can of course be useful but only if the conditions reported exactly match the specific question. Alternatively, this study determined the most relevant factors influencing D10, by quantitatively analyzing data from many references. The best first step appeared to be a classification of the data into vegetative bacteria and spores. As expected, spores were found to have significantly higher D10 values (average 2.48 kGy) than vegetative bacteria (average 0.762 kGy). Further analyses of the vegetative bacteria confirmed the expected extreme irradiation resistance of nonpathogenic Deinococcus radiodurans (average 10.4 kGy). Furthermore the analysis identified Enterococcus faecium, Alcaligenes spp., and several members of the Moraxella-Acinetobacter group as having very high resistance at very low temperatures (average 3.65 kGy). After exclusion of high- and low-resistance spores and some specific conditions showing relevant high or low D10 values, the average for spores was estimated to be 2.11 kGy. For vegetative bacteria this average was estimated to be 0.420 kGy. These approximate estimates are not definite, as they depend on the data used in the analyses. It is expected that inclusion of more data will not change the estimates to a great extent. The approximate estimates are therefore useful tools in designing and evaluating irradiation processes.

Growth and inactivation models to be used in quantitative risk assessments.

In past years many models describing growth and inactivation of microorganisms have been developed. This study is a discussion of the growth and inactivation models that can be used in a stepwise procedure for quantitative risk assessment. First, rough risk assessments are performed in which orders of magnitude for microbial processes are estimated by the use of simple models. This method provides an efficient way to find the main determinants of risk. Second, the main determinants of risk are studied more accurately and quantitatively. It is best to compare several models at this level, as no model is expected to be able accurately to predict microbial responses under all circumstances. By comparing various models the main determinants of risk are studied from several points of view, and risks can be assessed on a broad basis. If, however, process variations have a more profound effect on risk than the differences between models, it is most efficient to use the simplest model available. If relevant, the process variations can be stochastically described in the third level of detail. Stochastic description of the process parameters will however not change the conclusion on the usefulness of simple models in quantitative risk assessments. The proposed stepwise procedure that starts simply before going into detail provides a structured method of risk assessment and prevents the researcher from getting caught in too much complexity. This simplicity is necessary because of the complex nature of food safety. The principal aspects are highlighted during the procedure and many factors can be omitted since their quantitative effect is negligible.

An identification procedure for foodborne microbial hazards.

A stepwise and interactive identification procedure for foodborne microbial hazards has been developed in which use is made of several levels of detail ranging from rough hazard identification to comprehensive hazard identification. This approach allows one to tackle the most obvious hazards first, before focusing on less obvious hazards. The interactive character of the identification procedure is based on the use of several knowledge sources. Combination of knowledge sources, expressed in the use of knowledge rules, supports the user in systematically selecting hazards which may pose a real risk to the consumer. Due to the structured method and the clear definitions of the knowledge rules, the procedure is transparent and may be changed if necessary. The hazard identification procedure has been implemented as a computer program, resulting in a decision-supporting identification system. It provides a way to efficiently assess those hazards which may cause harm if not brought under control during processing. The procedure forms a basis for quantitative risk assessments.