Submitted to: Journal of the Association of Official Analytical Chemists
Publication Type: Peer reviewed journal
Publication Acceptance Date: 12/13/2004
Publication Date: 2/2/2005
Citation: Vargas, E.A., Whitaker, T.B., Santos, E.A., Slate, A.B., Lima, F.B., Franca, R.C. 2005. Testing green coffee for ochratoxin a, part ii: observed distribution of ochratoxin a test results. Journal of the Association of Official Analytical Chemists. 88:780-787. Interpretive Summary: Ochratoxin A (OTA) is a carcinogenic and toxic compound produced by molds found in coffee, wine, grains, and other agricultural commodities. The Food and Drug Administration is considering the establishment of an advisory or legal limit that would control the maximum quantity of ochratoxin allowed in coffee imported into the United States. As a result, coffee beans are inspected by processors, exporters, importers, and food manufacturers to detect and remove contaminated lots from the food chain. It is difficult to accurately determine the ochratoxin level of large shipments because of the errors associated with the ochratoxin test procedure used to quantify OTA in a bulk shipment and as a result the inspection program will misclassify some lots. Some of the good lots test bad and some of the bad lots test good. A method is being developed to predict the rate of misclassifications, which requires an understanding of the distribution of OTA test results taken from a contaminated lot. The distribution associated with measuring OTA in coffee beans over a wide range of lot concentrations was determined. Once the distribution of OTA test results is know, methods can be developed to reduce the testing errors which will reduce the number of lots misclassified. This will reduce both health risks to the consumer and economic loss to importers, processors, and manufacturers.
Technical Abstract: The suitability of four theoretical distributions (normal, lognormal, negative binomial and gamma) to predict the observed distribution of ochratoxin A (OTA) in green coffee was investigated. One symmetrical and three positively skewed theoretical distributions were each fitted to 25 empirical distributions of OTA test results for green coffee beans. Parameters of each theoretical distribution were calculated using Methods of Moments. The three skewed theoretical distributions provided acceptable fits to each of the 25 observed distributions. Because of its simplicity, the lognormal distribution was selected to model OTA test results for green coffee. Using variance equations determined in previous studies, mathematical expressions were developed to calculate the parameters of the log normal distribution for a given OTA lot concentration and test procedure. Observed acceptance probabilities were compared to an operating characteristic curve predicted from the lognormal distribution, and all 25 observed acceptance probabilities were found to lie within 95% confidence band associated with the predicted operating characteristic curve. The parameters of compound gamma distribution were used to calculate the fraction of OTA contamination kernels within a contaminated lot. The percent-contaminated kernels were estimated to be a function of the lot concentration and increased with lot concentration. It is estimated that at a lot concentration of 5 ug/kg, approximately 6 beans per 10,000 beans are contaminated.