Use of OC curves in quality control with an example of sampling for mycotoxins

An ‘operating characteristics’ (OC) curve is a simple tool that has been in use in quality control for many years but does not seem to be widely applied in the particulate sampling field. The OC curve provides the probability that a lot of material will be deemed to meet a specification (will be found to have an assay that falls above (or below) a specified level, given the true assay of the lot). In the application considered herein, it provides the probability that a grain shipment will be accepted, given the true value of the assay for the lot. It directly measures the probability of a type II error. To construct the OC curve for a given sampling protocol, it is necessary to know all the relevant components of variance and their distribution, as a function of the level of contamination in the shipment. This may be quite a challenge in many circumstances as the assumption of normality of distributions may be poor when dealing with substances such as mycotoxins. The paper introduces the method of OC curve construction and reviews the method developed by Whitaker for the construction of OC curves for mycotoxins in a wide range of commodities. It is shown that his method excludes a potentially critical component of uncertainty. Further, the discussion concludes that the estimation of the distribution of the missing component of uncertainty is potentially prohibitively expensive and logistically very difficult. The final conclusion is that more intensive sampling methods should be employed for mycotoxins.