Statistically based acceptance sampling plans have been integrated into mic
robiological criteria for food in international trade for some time. Especi
ally two- and three-class sampling plans for attributes are used, variables
plans are applied less. A lot of examples can be found in the so-called IC
MSF (International Commission on Microbiological Specifications for Foods)
"sampling book" or in EU guidelines. However, quite often the required conf
idence and the assumptions on tolerable lot qualities on which the choice a
nd design of sampling plans are based can't be deduced from such recommenda
tions and prescriptions.
These problems have led to critical discussion of the performance of two- a
nd three-class sampling plans used in microbiological criteria, showing com
plex dependencies on microbiological limits and on the expected variability
of sampling results. As a result means should be considered to use knowled
ge gained from process analyses in food production in the design of samplin
g plans.
In this context the performance of acceptance sampling plans as prescribed
in microbiological criteria for foods will be discussed as well as their ro
le in recently developed food safety concepts.