Sampling plan designs for gluten estimation in oat flour by discrete and composite sampling

Abstract

Heterogeneous gluten distribution in certain food commodities, such as oat flour, warrants a sound sampling plan to reduce measurement variability and the risk associated with lot misclassification. Ten lots of oat flour, 45 kg each, with varying gluten content were produced in a pilot-scale hammer mill to evaluate various sampling plans for gluten in oat flour. Thirty-two samples from each lot were collected during milling to study the effect of discrete (individual samples taken from the lot) and composite (samples taken from a bulk sample prepared by compositing several individual samples) sampling methods. A 5 g test sample was manually taken from each of the 32 samples, extracted with a cocktail extraction buffer, and gluten content was estimated using ELISA by analyzing duplicate aliquots of the extract. The total variance (Vt) from gluten measurements was partitioned into variance between samples (Vs) and aliquots analyzed (Va). Regression analysis revealed a linear relationship between the log of all three variances and the log of gluten concentration, which can be explained by the power equation for discrete and composite sampling methods. The Vt and Vs for discrete sampling tended to be higher than those for composite sampling at a given gluten concentration. A log-normal distribution was found suitable to characterize the distribution of measured gluten in oat flour samples from a test procedure. Operating characteristic (OC) curves plotted to evaluate various sampling plans for gluten in oat flour showed reduced risk of misclassification for composite sampling, as compared to the discrete sampling method for a given sampling plan. Examples are shown for the change in OC curves depending on the sampling method, sample size, number of samples analyzed, and accept/reject limit.