Design of a sampling plan to detect ochratoxin A in green coffee

Abstract

The establishment of maximum limits for ochratoxin A (OTA) in coffee by importing countries requires that coffee-producing countries develop scientifically based sampling plans to assess OTA contents in lots of green coffee before coffee enters the market thus reducing consumer exposure to OTA, minimizing the number of lots rejected, and reducing financial loss for producing countries. A study was carried out to design an official sampling plan to determine OTA in green coffee produced in Brazil. Twenty-five lots of green coffee (type 7 - approximately 160 defects) were sampled according to an experimental protocol where 16 test samples were taken from each lot (total of 16 kg) resulting in a total of 800 OTA analyses. The total, sampling, sample preparation, and analytical variances were 10.75 (CV = 65.6%), 7.80 (CV = 55.8%), 2.84 (CV = 33.7%), and 0.11 (CV = 6.6%), respectively, assuming a regulatory limit of 5 microg kg(-1) OTA and using a 1 kg sample, Romer RAS mill, 25 g sub-samples, and high performance liquid chromatography. The observed OTA distribution among the 16 OTA sample results was compared to several theoretical distributions. The 2 parameter-log normal distribution was selected to model OTA test results for green coffee as it gave the best fit across all 25 lot distributions. Specific computer software was developed using the variance and distribution information to predict the probability of accepting or rejecting coffee lots at specific OTA concentrations. The acceptation probability was used to compute an operating characteristic (OC) curve specific to a sampling plan design. The OC curve was used to predict the rejection of good lots (sellers' or exporters' risk) and the acceptance of bad lots (buyers' or importers' risk).

Comparison of the efficiency between two sampling plans for aflatoxins analysis in maize

“…Despite the fact that similar models have been used to precisely simulate the distribution of the results from mycotoxins quantification in seeds and feeds (Sharkey et al , 1994; Giesbrecht et al , 1998; Johansson et al , 2000b; Vargas et al , 2006), lognormal distribution was chosen because it obtained the major number of lots with acceptable distribution and highest average probability. Thus, this distribution was used to calculate the OC curve and evaluate the performance of both sampling plans for aflatoxins quantification in maize.…”

“…However, the heterogeneous distribution of mycotoxins in cereals leads to variations on the final results, so called errors, not being possible to establish the real mycotoxins concentration in a determined lot (Whitaker 2004; Miraglia et al , 2005; Vargas et al , 2006). Thus, the efficiency of a sampling plan can be estimated by the evaluation of the variance of the results generated by the quantification procedures applied.…”

“…The frequency of occurrence of those two kinds of errors depends on the sampling plan and is given by the design of a characteristic operation curve (OC) (Johansson et al , 2000c; Miraglia et al , 2005; Vargas et al , 2006). …”

Sampling Techniques for the Determination of Mycotoxins in Food Matrices

“…102 The most diffuse extraction method for OTA determination in green and roasted coffee is the use of a mixture of methanol and 3% aqueous sodium bicarbonate solution (50:50), using a laboratory shaker or homogenizer, followed by IAC cleanup for HPLC-FLD or LC-MS/MS determination. 100,101 Recoveries above 80% are usually obtained, exploiting the high solubility of the dissociated form of the toxin at alkaline pH.…”

Comparing the performance of sampling plans that use a single regulatory limit based upon total aflatoxins to sampling plans that use dual limits based upon B1 and total aflatoxins

“…The U.S. almond industry was concerned that using a AFB 1 limit in addition to a limit based on AFT would increase the economic burden on the industry by increasing the number of lots rejected from the export market when compared to using a similar sampling plan with a single limit based upon AFT. Statistical methods developed by Whitaker and colleagues (Vargas et al, 2006;Whitaker et al, 2006a,b) to evaluate the performance of mycotoxin sampling plan designs (which determine the chances of rejecting good lots and accepting bad lots) had only been determined for a single limit, not dual limits. To determine the effect that dual limits have on the probability of accepting and rejecting lots compared to a single limit, studies were designed with the objectives of (a) measuring distribution of AFB 1 /AFT ratio values using samples test results associated with testing U.S. almond lots exported to the EU; (b) using Monte Carlo methods to develop a model to compute the effects of using dual limits based upon AFB1 and AFT on the probability of accepting and rejecting almond lots; and (c) comparing the probability of accepting and rejecting almond lots using the current RTE and DFP Codex aflatoxin sampling plans for tree nuts when using single versus dual limits.…”