An accept-and-reject algorithm to determine performance objectives that comply with a food safety objective
Abstract
Quantitative risk assessment of microbiological hazards in foods (QMRA) can assess the impact of control measures on risk and can help to achieve food safety targets. Generally, the effect of a particular input variable of a QMRA model is assessed by comparing the results corresponding to the default situation to those obtained from scenarios modifying its behaviour. The purpose of this work is to handle and propose alternative methods to scenario testing for determining PO (performance objective) or process criterion from a given FSO (food safety objective). These methods were applied in the framework of a QMRA model that considered the fate of a hazard from raw material to the consumption stage. A second order Monte-Carlo simulation approach separately assessing the uncertainty and variability on the final exposure, we applied an accept-and-reject algorithm to measure the importance of each variable of the model and to determine, with its uncertainty, the probability of compliance with an FSO according to the range of values that these variables can take. Within a first order Monte-Carlo simulation approach, we applied the Saltelli sensitivity analysis method to select the most influential variables on the compliance with the FSO. Then, an accept-and- reject algorithm was applied for the most influential variables. With both approaches, we managed to identify influential variables and were able to determine which ranges of values should be met to respect the FSO. Complications originated from correlated variables or QMRA models with low probability to reach the FSO were also tackled. It is concluded that accept and-reject algorithms are simple methods to apply and that they allow extrapolation of the classical point estimate scenario analysis to the entire range of values taken by input variables.
Origin | Publisher files allowed on an open archive |
---|