[Variant component analysis of bacterial count in milk. 2. Consequences for the preparation of microbiological...

Results of the repeated testing of pasturized and raw milk to determine the bacterial count deviation in parallel samples were described in the first communication. Using the same data, the second communication demonstrates the influence of the observed bacterial count variation on the decision-making process of the simultaneously performed microbiological sampling plan. Analysis of the 3-class sampling plans under applied conditions showed that little attention was paid to the real variation of the criterion and in some plans, too much weight was placed on the possible outliers. Due to theoretical considerations, the difference between m and M in sampling plans with n = 5 and c(m) = 2 should at least amount to 1.85 times the standard deviation. This premise corresponds to the half of a tenth power in milk with homogenous bacterial population. It seems absolutely unwarranted to interrupt the examination of raw milk when the first random sample with n = 1 is lower than the GMP-limit. Examples show that in borderline cases where the bacterial counts are close to the cut-off value, the application of this strategy leads to the acceptance of numerous rejectable batches. Because of the few coliform counts in fresh, raw milk and especially in pasteurized milk, examination procedures with a detection limit of 1 cfu/ml are not advised and the quality assessment of a lot based on single random samples is not possible.