Wageningen Academic P u b l i s h e r s World Mycotoxin Journal, May 2010; 3 (2): 157-168 Number of export almond lots rejected in the EU due to USA sampling plans and aflatoxin contamination levels among lots tested T. Whitaker1, A. Slate1, J. Adams2 and T. Birmingham2 !N.C. State University, Biological and Agricultural Engineering Department, Box 7625, Raleigh, NC 27695-7625, USA; Board of California, 1150 9th Street, Suite 1500, Modesto, CA 95354, USA; [email protected] 2Almond Received: 2 March 2010 / Accepted: 26 April 2010 © 2010 Wageningen Academic Publishers Abstract In March 2010, the European Union (EU) modified their aflatoxin limits and sampling plans for almonds, pistachios, and hazelnuts to closely resemble that developed by Codex in 2008. To assist the USA almond industry evaluate how to modify their aflatoxin sampling plan to minimise USA lots rejected by the EU at import, studies were conducted to: (a) develop a model to predict the number of USA almond lots rejected at origin and at destination by various Codex style aflatoxin-sampling plans used by the USA where accept/reject limits vary from EU limits of 8 ng/g aflatoxin Bj / 10 ng/g total aflatoxins (AFT) to 8, 6, 4, 2, and 0.5 ng/g AFT; (b) determine the effect of the level of aflatoxin contamination among lots tested in the USA on the number of lots accepted and rejected at origin and at destination; and (c) develop a method based upon lots rejected in the USA to predict the lots rejected several months later in the EU. For a given USA crop contamination level, as the accept/reject limits of the USA sampling plan decreases from 8 to 0.5 ng/g AFT, the number of lots rejected in the USA increases, the average aflatoxin concentration among all lots accepted for export decreases, the number of lots rejected in the EU decreases, and the average aflatoxin concentration among lots accepted in the EU decreases. For a given USA aflatoxin sampling plan used at origin, an increase in the USA crop aflatoxin contamination level increases the number of lots rejected at origin, increases the average aflatoxin concentration in USA lots accepted for export, increases the number of lots rejected in the EU, and increases the average aflatoxin concentration among lots accepted in the EU. Graphical techniques were developed to use the percentage of lots rejected in the USA to predict the percentage of USA lots rejected by the EU. Keywords: aflatoxin, almonds, treenuts, maximum levels, Codex, export sampling plans 1. Introduction Because aflatoxin is considered toxic and carcinogenic, about 100 countries have established maximum levels or regulatory limits for aflatoxin in food and feed products (FAO, 2003). Maximum levels for aflatoxin not only vary in magnitude, but vary with the type of aflatoxin being controlled. Maximum limits can be based upon aflatoxin Bj (AFBj), total aflatoxins (AFT = AFBj + AFB2 + AFGj + AFG2), or a combination of AFBt and AFT. When a regulatory agency establishes regulatory limits for aflatoxin, there is a ripple effect throughout the export market that affects both the importer at destination and the exporter at origin. The exporter usually samples bulk lots at origin to ascertain if each lot's aflatoxin concentration is below the importing country's regulatory limit. If the exporter's sample test results indicate the lot is above the importing country's regulatory limit, the lot will not be shipped to the importer. The importer (or import control authorities) will sample imported lots at destination to ascertain if the aflatoxin concentration of the imported lot is below the importing country's regulatory limit. If the importer's sample test results indicate that the aflatoxin level in the lot is above the regulatory limit, the lot is diverted from the food chain by the importer, and subjected to additional measures. It is possible that the imported product is shipped back to the exporter or re-processed at destination using proven sorting techniques to reduce contamination levels; ISSN 1875-0710 print, ISSN 1875-0796 online, DO/ 10.3920/WMJ2010.1209 157 IVWiMeretal. in extreme situations, the imported product could be destroyed at destination. Lots that fail to test below the regulatory limit at destination pose an economic burden to both the exporter and the importer. The exporter must pay the costs associated with the alternatives mentioned above and the importer suffers from problems associated with a loss in the consistency of supply. Because there is variability associated with the sampling, sample preparation, and analytical steps of the aflatoxin test procedure (Whitaker et al., 2006), it is not possible to determine the true aflatoxin level in a lot with 100% certainty by measuring the aflatoxin concentration in samples taken from the lot. As a result, some lots will be misclassified by both the exporter and importer when using sample test results to determine if the true lot aflatoxin concentration is above or below a defined regulatory limit. Two risks are associated with any sampling plan design: (a) there is a chance that some good lots (lots with aflatoxin levels below a regulatory limit) will be rejected by the sampling plan (false positive) and (b) there is a chance that some bad lots (lots with aflatoxin levels above a regulatory limit) will be accepted by the sampling plan (false negative). For sampling plans conducted at origin by the USA exporter, a false positive (good lot rejected) can be called the exporter's risk and a false negative (bad lot accepted) can be called the importer's risk. As a consequence of the random variation associated with the aflatoxin test procedure (sampling, sample preparation, and analysis), sampling product at origin doesn't guarantee that all lots will be accepted by the importer when re-sampled and tested for aflatoxin at destination (Whitaker et al., 2006). It is inevitable that some lots tested at destination will still be rejected even if the exporter and importer are conducting their aflatoxin sampling programs according to standard protocols and without mistakes. Unfortunately, these rejections are sometimes incorrectly perceived as a failure of the aflatoxin sampling program rather than the result of random variability associated with sampling, sample preparation, and analysis. According to 2008 industry statistics (Almond Board of California, 2009), USA almond growers produce over 80% of the world's almond production. Of the total USA almond production, approximately 70% went into the export market. Western Europe is the largest importer of USA almonds receiving 46% of the total USA export shipments. This constitutes almost 11,000 lots exported to the EU during the 2008-2009 crop year (Note: a crop year spans two calendar years, beginning August 1, 2008 approximately from the beginning of harvest, through July 31,2009). Each export lot destined for the EU is sampled and tested for aflatoxin at origin to ascertain if the lot aflatoxin concentration will comply with the EU ready-to-eat (RTE) limits of 2 ng/g AFBj and 4 ng/g AFT. 158 In early 2010, the European Commission (EC) adopted legislation changing their aflatoxin regulatory limits (EC, 2010a) and sampling plans (EC, 2010b) for treenuts to more closely conform to that developed by the Codex Committee on Contaminants in Foods (CCCF) and adopted by the Codex Alimentarius Commission (CAC) in July 2008 (CCCF, 2008). The Codex aflatoxin sampling plan for RTE treenuts (almonds, pistachios, and hazelnuts) requires that two 10 kg samples both test less than 10 ng/g AFT (2x 10 kg < 10 ng/g AFT) to accept the lot. The EU adopted the Codex plan, but added an AFB1 limit of 8 ng/g. As a result, an RTE almond lot requires two 10 kg samples to each test less than both limits (8 ng/g AFBt and 10 ng/g AFT) for the lot to be accepted into the food chain. For consistency, the USA almond industry decided to modify their aflatoxin sampling plan at origin in order to ascertain if lots exported to the EU will meet the new EU aflatoxin regulatory limits of 8 ng/g AFBj and 10 ng/g AFT for RTE almonds. To assist the USA almond industry in the design of a new aflatoxin sampling plan that would meet or exceed the performance of the EU sampling plan and minimise lots rejected by the EU at destination, a study was developed with the following objectives: (a) develop a spreadsheet model to predict the number of USA almond lots accepted and rejected both at origin in the USA and at destination in the EU by various Codex-style (2x10 kg) aflatoxinsampling plan designs used by the USA almond industry; (b) determine the effect of the level of aflatoxin contamination among all lots tested in the USA during a crop year on the number of lots accepted and rejected at origin and at destination; and (c) develop a method that uses sample test results from the USA sampling program that USA almond exporters can use in 'real-time' to predict the percentage of lots that will be rejected several months later at destination in the EU. The study results can further provide exporters, importers, and regulatory authorities with a better understanding of how sampling plan designs and crop contamination levels influence the number of lots rejected at origin and at destination when both the exporter and the importer are using standard sampling, sample preparation, and analytical protocols to identify and reduce the number of contaminated lots in the export market. 2. Methods A flow chart describing the spreadsheet model that simulates the aflatoxin testing of RTE almond lots in the USA and in the EU during a given crop year is shown in Figure 1. The simulation model not only computes the total number of lots accepted and rejected at origin in the USA and at destination in the EU during a given crop year, but keeps track of the distribution among lots according to their aflatoxin concentration that are tested, accepted, and World Mycotoxin Journal 3 (2) Almond lots accepted and rejected in the export market Aflatoxin distribution among lots tested at origin FLReu(C). Equations developed for the spreadsheet model are given below. The number of USA lots accepted by the USA sampling plan at a given aflatoxin concentration C: B USA aflatoxin sampling plan ) = TLusFLus(C)PAas(C) Lots rejected at origin (1) for C = 0,1,2,.... max. ng/g AFT. Lots accepted for export to EU The number of lots rejected by the USA sampling plan at a given aflatoxin concentration C: EU aflatoxin sampling plan = TLUSFLUS(C)PRUS(C) Lots rejected at destination Lots accepted at destination Figure 1. Flow diagram of spreadsheet model showing the lots tested, accepted, and rejected in the USA (origin) and in the EU (destination). for C = 0,1, 2,..., max. ng/g AFT. Total number of lots accepted in a crop year by the USA sampling plan and exported to the EU: Omax (3) c=o Total number of lots rejected in a crop year by the USA sampling plan and diverted from export: rejected in the USA and in the EU. From the lot distribution information, the amount of aflatoxin removed from or remaining in the export market can also be computed. The simulation starts with a known aflatoxin distribution among USA lots in a given crop year, FLUS(C), which is the fraction of total lots, TLus, at a given aflatoxin concentration, C, (called the USA lot distribution). Each lot is sampled and tested for aflatoxin at origin in the USA during a given crop year. There is a chance that a given lot in the USA lot distribution with aflatoxin concentration C will be accepted with probability PA^C) or rejected with probability PR^C) = (1 - PA,,, (C)) by the USA sampling plan. The accept and reject probabilities are unique for the design of the aflatoxin sampling plan (sample size, etc.), regulatory limit, and the lot aflatoxin concentration C. As a result, the USA lot distribution, FLUS(C), is partitioned into a USA accepted lot distribution, FL\(C), and a USA rejected lot distribution, FLRUS(C). (2) C=max C=0 The number of lots accepted at destination by the EU sampling plan at a given aflatoxin concentration C: LAeu(Q=TLAusFLAus(C)PAeu(C) (5) for C=0,1, 2,..., max ng/g AFT. The number of lots rejected at destination by the EU sampling plan at a given aflatoxin concentration C: LReu(C) = TLAUS FLAUS(C) PReu(C) (6) for C=0,1,2,..., max ng/g AFT. Total number of lots accepted in a crop year by the EU sampling plan at destination in the EU: C=max The distribution of lots accepted in the USA, FLA^C), is exported to the EU (called the USA export lot distribution). Once the USA export lot distribution arrives at destination in the EU, all or a portion of the export lot distribution is sampled and tested for aflatoxin by EU authorities. Each lot in the USA export lot distribution, FLA^C), will be accepted with probability PAeu(C) or rejected with probability PReu(C) = (1 - PAeu (C)) by the EU sampling plan. As a result, the USA export lot distribution is partitioned by the EU sampling plan into an EU accepted lot distribution, FLAeu(Q, and an EU rejected lot distribution, World Mycotoxin Journal 3 (2) TLAeu = ZLAeu(Q c=o (7) Total number of lots rejected in a crop year by the EU sampling plan at destination in the EU: C=max c=o The average aflatoxin concentration among lots tested, accepted, and rejected in a crop year after each testing point in the USA and in the EU can be calculated from the above equations. The average aflatoxin concentration among all lots before testing in the USA (ATL^) is: 159 I Whitaker Q[a\. C=max = £(FLUS(C)C) c=o (9) The average aflatoxin concentration among all USA lots accepted (ALAJ and rejected (ALRUS) in the USA for export to the EU are: C=max (10) 00 C=max C=0 (11) The average aflatoxin concentration among all USA lots exported to the EU and tested for aflatoxin in the EU is also described by Equation 10. The following information is needed for the development of the spreadsheet model shown in Figure 1: (a) USA lot distribution to be tested at origin (Box A), (b) the accept and reject probabilities associated with a specific USA aflatoxin sampling plan design used at origin (Box B), and (c) the accept and reject probabilities associated with the EU aflatoxin sampling plan design used at destination (Box E). Once the information in Boxes A, B and E are known, the remaining information in Boxes C, D, F and G can be computed using Equations 1 to 11. Development of information for the spreadsheet model (Boxes A, B and E) is discussed below. Aflatoxin distribution among USA lots to be tested for export The desire was to find at least four USA lot distributions where the average aflatoxin concentration among all lots in each crop (called the crop contamination level) varied from some low value to some high value to demonstrate the effect of crop contamination level on the number of lots accepted and rejected at origin in the USA and at destination in the EU. The California almond industry began testing all export lots destined for the EU using USDA-approved laboratories in September 2007. By the summer of 2009, lots from three crop years, 2006,2007, and 2008 had been sampled and tested for aflatoxin. As mentioned earlier, a crop year spans two calendar years. Because sampling started in late summer of 2007, most of the 2006 crop had been marketed and only the remaining 1,519 lots from the 2006 crop were sampled and tested for aflatoxin prior to export to the EU. The entire 2007 crop of 15,021 lots was sampled and as of December 2009 a total of 11,078 lots of the 2008 crop have been sampled and tested for aflatoxin prior to export to the EU. For a given crop year, the sample test results from the USA almond industry sampling program (called Voluntary Aflatoxin Sampling Plan or VASP by the California almond industry) were used to estimate the USA lot distribution 160 identified for export to the EU. Three 5 kg laboratory samples are taken from each lot destined for the EU and tested for aflatoxin. The average of the three sample test results was assumed to equal the lot aflatoxin concentration in total aflatoxins. Therefore, the USA lot distribution was assumed to equal the distribution among the averages of the three sample test results for each lot. Table 1 shows the number of lots sampled by crop year, the average AFT level among all lots tested in each crop year, variance among all lot concentrations, and the percentage of lots testing zero AFT or non-detect. From Table 1, it can be seen that the aflatoxin contamination in the 2006 crop was higher than the aflatoxin contamination in the 2007 and 2008 crops (based upon the average aflatoxin concentration among all lots tested). To get a range in crop contamination levels, it was decided to use the 2008 (0.42 ng/g AFT) and 2006 (1.26 ng/g AFT) sample averages to construct two USA lot distributions. It was also decided to construct two additional hypothetical USA lot distributions where the average aflatoxin concentrations were at least two times and four times the average aflatoxin concentration of the 2006 lot distribution or 3.0 and 6.0 ng/g AFT. The two hypothetical lot distributions were generated using the compound gamma distribution (Giesbrecht and Whitaker, 1998; Whitaker et al,, 1996) where the parameters for the compound gamma distribution were determined from the mean and variance among lot aflatoxin concentrations shown in Table 1. The compound gamma distribution was chosen to generate the hypothetical lot distributions after several skewed distributions such as the negative binomial and lognormal distributions were investigated (Giesbrecht and Whitaker, 1998; Whitaker et al, 1996). Using four USA lot distributions with a wide range in aflatoxin concentration (0.42 to 6.0 ng/g AFT) should provide a reasonable demonstration of the effect of aflatoxin Table 1. Number of lots tested before shipment to the European Union, average aflatoxin concentration among all lots tested, and the percent lots that tested zero (non-detect) aflatoxin for the 2006,2007 and 2008 almond crops produced in the USA. Crop year Number Average Variance of lots aflatoxin among lot sampled concentration concentrations (ng/gAFTp Lots at zero aflatoxin concentration (%) 2006 1,519 1.26 64.8 77.0 2007 15,021 0.47 21.5 90.4 20081 11,078 0.42 22.8 91.3 1 Lots from the 2008 crap year are still being marketed. 2 Average aflatoxin concentration (total ng/g AFT) among all lots sampled for export. World Mycotoxin Journal 3 (2) Almond lots accepted and rejected in the export market contamination in the USA lot distribution on lots accepted and rejected at origin in the USA and at destination in the EU. The four USA lot distributions, calculated with the compound gamma distribution, are shown as cumulative distributions in Figure 2. The USA lot distribution information is used in Box A in Figure 1. Accept and reject probabilities of USA and EU aflatoxin sampling plans for RTE shelled almonds Since the EU developed a sampling plan similar to the Codex plan (two 10 kg laboratory samples taken from a lot must each test less than 10 ng/g AFT (2x10 kg < 10 ng/g AFT) for the lot to be accepted), but added a AFBj accept/ reject limit of 8 ng/g for RTE almonds (2x10 kg <, 8 ng/g AFB1 / 10 ng/g AFT), a single EU aflatoxin sampling plan design was evaluated in the spreadsheet model for use at destination (2x10 kg < 8 ng/g AFBj / 10 ng/g AFT). A Monte Carlo method was developed (Whitaker et al., 2010) to predict the accept and reject probabilities associated with the EU sampling plan using dual limits. A plot of the accept probability, PAeu(C), associated with the EU sampling plan versus lot concentration C (called an operating characteristic or OC curve) is shown in Figure 3. The accept probabilities associated with the EU sampling plan are used in Box E (Figure 1). The design of the Codex sampling plan for treenuts influenced the 2010 EU aflatoxin sampling plan design that in turn influences the aflatoxin sampling plan designs considered by the USA almond industry to test USA lots exported to the EU. Because the EU has decided to use a AFBj accept/reject limit of 8 ng/g in addition to the 10 ng/g AFT for RTE almonds, it was decided to evaluate five USA aflatoxin sampling plans similar to the Codex plan where two 10 kg samples must each test less than a single accept/ reject limit of 8, 6, 4, 2, and 0.5 ng/g AFT. A sixth USA sampling plan design that was identical to the EU sampling plan design with these dual limits was also evaluated. The accept and reject probabilities for the five USA sampling plan designs with single limits were computed using the method used by CCCF to design harmonised aflatoxin sampling plans for almonds, pistachios, and hazelnuts (CCCF, 2008). The CCCF method of computing accept and reject probabilities for almonds was based upon variance and distribution data among sample test results published by Whitaker et al. (2006). The six OC curves shown in Figure 3 describe the accept and reject probabilities associated with the six USA aflatoxin-sampling plan designs considered for use by the USA almond industry at origin. The accept and reject probabilities associated with each of the six USA sampling plans are used in Box B in Figure 1. aflatoxin concentration 2006 crop avg. aflatoxin concentration .26 ng/g AFT Hypothetical avg afla oxin concentration • 3.00 ng/g AFT Hypothetical avg afla oxin concentration 6.00 ng/g AFT o 50 > i | . i . , | . | i i | i i i | . ,i Lot aflatoxin concentration (ng/g AFT) Figure 2. Four USA almond lot distributions, according to the average aflatoxin crop contamination level among all lots, that are tested prior to shipment to the European Union. World Mycotoxin Journal 3 (2) 161 T.Whitakere\a\. OC curves for VASP sampling plans Shelled almonds Laboratory sample size shown Dry grind, 50 g test portion Analysis, 1 aliquot, 22% reproducibility RSD Maximum level shown (ng/g) 2x10kg<0.5T 0 I ' ' ' ' I ' ' ' ' 10 15 20 25 Lot aflatoxin concentration (ng/g AFT) Figure 3. Operating characteristic (OC) curves describing the probability of accepting almond lots by sampling plans used in the USA and in the European Union. All sampling plans are a modification of the Codex aflatoxin sampling plan for treenuts where two 10 kg samples must each test less than the accept/reject limit shown. Export market assumptions for model development Several basic assumptions had to be made about sampling plans and export market conditions before simulations were conducted: a. Because the number of USA lots tested and accepted by a USA sampling plan for export to the EU will vary from one crop year to another, it was decided to make the total number of USA lots tested in a crop year (TL^) vary with crop contamination levels and sampling plan designs such that 11,000 lots were always accepted (TLAUS) for export to the EU in all 24 simulations (six USA sampling plan designs by four crop contamination levels). b. When USA lots are imported into the EU, the almond industry estimates that approximately 5% of the imported lots are sampled by EU authorities under random control (EC, 2007). For each of the 24 simulations, three rates of EU inspection were evaluated (2, 5, and 10%) for the 11,000 USA export lots sampled at destination in the EU. It is assumed that the aflatoxin distribution among the lots identified by the EU for testing at import is the same as the aflatoxin distribution among the 11,000 lots imported into the EU. c. No biases or mistakes are associated with the sample selection, sample preparation, or analytical methods used by USA and EU authorities to test bulk lots for aflatoxin. Since the number of lots tested in the USA may vary with crop year, and the number of USA lots tested in the EU can vary due to regulatory decisions, the number of lots accepted and rejected in the USA and in the EU was converted to percentage of lots tested in the USA and in the EU. The percentages can be converted back into number of lots accepted and rejected in the USA and in the EU once the actual number of lots tested in the USA and in the EU is known. 3. Results and discussion A total of 24 spreadsheet simulations (six USA sampling plan designs times four USA crop contamination levels) were run to solve for the number (percentage) of lots accepted and rejected in the USA and in the EU and the average aflatoxin concentration among lots accepted and rejected in the USA and in the EU. Because of the size and complexity of the output of a spreadsheet simulation (approximately 10 columns by 70 rows), only a summary of each of the 24 simulation results is shown in Tables 2 and 3. These two tables contain the same simulation results, but World Mycotoxln Journal 3 (2) 550 lots (m Table 2. Effect of four different crop contamination levels and six differentLots USA sampling rejected 2% plan designs on the number of lots accepted and rejected in the USA and in the European Union. The number of lots accepted and rejected at destination in the EU is computed220 forlots 2,5, (ntand 10% sampling rate with the new 2010 EU sampling plan (2x10 kg < 8 ng/g AFB^IO ng/g AFT). Results sorted first by crop contamination level. Crop avg. USA sampling plan USA (11 ,000 lots exported) EU (11,000 lots imported), sampling plan: 2x10 kg & 8 ng/g AFB1 / 10 ng/g AFT auaioxm ceinc. (ng/g AFT)1 j | Average afla. o rejects (ng/ "3 1 •3. 0.42 1.26 3.00 6.00 0.42 1.26 3.00 6.00 0.42 1.26 3.00 6.00 0.42 1.26 3.00 6.00 0.42 1.26 3.00 6.00 0.42 1.26 3.00 6.00 1 Average total USA export 2x10 £ 8 AFB,/10AR USA export 2x10 £ 8 AF^/10 AFT USA export 2x10 £ 8 AFB,/10 AFT USA export 2x 1 0 £ 8 AFB/1 0 AFT USA export 2x10 £8 AFT USA export 2x10 £8 AFT USA export 2x10 £8 AFT USA export 2x10 £8 AFT USA export 2x10 £6 AFT USA export 2x10 £6 AFT USA export 2x10 £6 AFT USA export 2x10 £6 AFT USA export 2x10 £ 4 AR USA export 2x10 £ 4 AR USA export 2x10 £4 AFT USA export 2x10 £4 AFT USA export 2x10 £ 2 AR USA export 2x10 £2 AFT USA export 2x10 £ 2 AR USA export 2x10 £2 AFT USA export 2x10 £0.5 AFT USA export 2x10 £0.5 AFT USA export 2x10 £0.5 AFT USA export 2x10 £0.5 AFT 11,149 11,446 12,024 12,942 11,153 11,461 12,061 13,012 11,178 11,529 12,203 13,254 11,225 11,644 12,417 13,598 11,374 11,935 12,854 14,206 11,983 12,943 14,029 15,498 ! 149 446 1,024 1,942 153 461 1,061 2,012 178 529 1,203 2,254 225 644 1,417 2,598 374 935 1,854 3,206 983 1,943 3,029 4,498 J 1 Average afla. o accepts (ng/ 1.36 4.06 9.31 17.66 1.39 4.19 9.64 18.29 1.62 4.81 10.93 20.49 2.05 5.85 12.88 23.61 3.40 8.50 16.86 29.15 8.94 17.67 27.54 40.89 10 || C3 5' 23.37 23,82 25.30 27.49 23.16 23.47 24.86 26.98 20.80 21.32 22.80 24.96 17.18 18,35 20.20 22.51 10.90 13.32 16.21 19.05 4.42 6.79 10.43 14.17 1 3 ta 0 11 0.15 0.38 0.77 1.44 0.15 0.36 0.73 1.35 0.13 0.32 0.63 1.17 0.12 0.27 0.52 0.95 0.10 0.21 0.39 0.69 0.08 0.16 0.25 0.41 3 0.51 1.54 3.56 6.94 0.48 1.46 3.38 6.59 0.42 1.26 2.92 5.74 0.34 1.03 2.38 4.71 0.24 0.73 1.69 3.37 0.13 0.40 0.91 1.85 I I 1.1 3.4 7.8 15.3 1.1 3.2 7.4 14.5 0.9 2.8 6.4 12.6 0.7 2.3 5.2 10.4 0.5 1.6 3.7 7.4 0.3 0.9 2.0 4.1 I I 2.8 8.5 19.6 38.2 2.7 8.0 18.6 36.3 2.3 6.9 16.1 31.6 1.9 5.6 13.1 25.9 1,3 4.0 9.3 18.5 0.7 2.2 5.0 10.2 o 2* •o Tj .3. §• P 5.6 16.9 39.2 76.4 5.3 16.1 37.2 72.5 4.6 13.9 32.2 63.1 3.7 11.3 26.2 51.8 2.7 8.0 18.5 37.1 1.5 4.4 10.1 20.3 •Z? 5" 0.11 0.25 0.46 0.81 0.11 0.24 0.44 0.78 0.10 0.22 0.41 0.71 0.09 0.20 0.36 0.61 0.08 0.18 0.29 0.49 0.07 0.14 0.22 0.34 ~3 3' 8.59 8.89 9.34 9.95 8.19 8.49 8.93 9.51 7.43 7.74 8.16 8.70 6.50 6.83 7.23 7.72 5.21 5.57 5.95 6.38 3.41 3.81 4.18 4.54 CL s- 53" I -i CD" 3" 3s CD I a i aflatoxin concentration: 2008 Crop - 0.42 ng/g AFT; 2006 Crop -1.26 ng/g AFT; hypothetical crops - 3.00 and 6.00 ng/g AFT. CD 550 lots (nu Lots rejected 2% I 220 tots (nu The number of lots accepted and rejected at destination in the EU is computed for 2,5, and 10% sampling rate with the new 2010 EU sampling plan (2x10 kg 5 8 ng/g AFB.,/10 ng/g AFT). Results sorted first by sampling plan design. Crop avg. USA sampling plan USA (11,000 lots exported) aflatoxin cone. Average afla. cc EU (11,000 lots imported), sampling plan: 2x10 kg £ 8 ng/g AFB1 / 10 ng/g AFT rejects (ng/s (ng/gAFT)' | i t i a 0.42 0.42 0.42 0.42 0.42 0.42 1.26 1.26 1.26 1.26 1.26 1.26 3.00 3.00 3.00 3.00 3,00 3.00 6.00 6.00 6.00 6.00 6.00 6.00 Average afla. cc accepts (ng/j WM 3 9.64 1,061 12,061 9.31 1,024 12,024 17.67 1,943 0.21 8.50 0.27 5.85 149 153 178 225 374 983 446 461 529 644 935 11,149 11,461 USA export 2x10 5 8 AFT 11,983 11,446 USA export 2x10 5 0.5 AFT USA export 2x10 £ 8 AFB^IO.SAFT 11,374 USA export 2x10 5 2AFT 11,225 USA export 2x10 54 AFT 11,178 USA export 2x10s6 AFT 11,153 USA export 2x10 s 8 AFB,/10 AFT USA export 2x10 S 8 AFT 13,598 USA export 2x10 s 4 AFT 13,012 13,254 USA export 2x10 5 8 AFT USA export 2x10 56 AFT 12,942 USA export 2x10 s 8 AFB^IO.S AFT 12,854 14,029 USA export 2x1052AFT USA export 2x10 50.5 AFT 12,417 USA export 2x10 s 4 AFT 12,203 USA export 2x10 5 6 AFT USA export 2x10 s 8 AFB.,/10.5 AFT USA export 2x1058AFT 12,943 USA export 2x10sO,5 AFT 11,529 11,644 11,935 USA export 2x10 5 4 AFT USA export 2x10 5 2 AFT 0.15 1.39 0.15 1.36 1.62 2.05 3.40 11 0.08 8.94 10.90 0.10 16.86 1,854 a 0.51 23.37 4.42 23.82 4.19 27.54 3,029 18.29 20.49 23.61 29.15 40.89 3 0.48 23.16 20.80 0.13 Lots rejected (nu 17.18 0.12 23.47 0.36 0.38 4.06 4.81 17.66 1,942 2,012 2,254 2,598 3,206 4,498 —J fe p •3 3' 10.93 1,203 14,206 USA export 2x10 56 AFT t 12.88 1,417 15,498 USA export 2x10 5 0.5 AFT USA export 2x10 52 AFT 0.42 0.34 0.24 0.13 1.54 1.26 21.32 0.32 1.46 18.35 13.32 0.16 6.79 25.30 0.77 0.73 0.63 0.52 1.03 0.73 0.40 3.38 24.86 16.21 0.39 26.98 24.96 1.17 27.49 1.44 1.35 3.56 2.92 22.80 20.20 10.43 0.25 22.51 0.95 14.17 0.41 19.05 0.69 2.38 1.69 0.91 6.94 6.59 5.74 4.71 3.37 1.85 !! 1.1 1.1 0.9 0.7 0.5 0.3 3.4 3.2 2.8 2.3 1.6 0.9 7.8 7.4 6.4 5.2 3.7 2.0 15.3 14.5 12.6 10.4 7.4 4.1 ^- ~ ET PL f& o S §? I I 2.8 2.7 2.3 1.9 1.3 0.7 8.5 8.0 6.9 5.6 4.0 2.2 19.6 18.6 -3 3 5.6 5.3 4.6 3.7 2.7 1.5 16.9 16.1 13.9 11.3 8.0 4.4 39.2 37.2 26.2 13.1 32.2 16.1 76.4 38.2 10.1 9.3 5.0 36.3 31.6 18.5 72.5 63.1 37.1 18.5 51.8 25.9 10.2 20.3 0.11 0.11 0.10 0.09 0.08 0.07 0.25 0.24 0.22 0.20 0.18 0.14 0.46 0.44 0.41 0.36 0.29 0.22 0.81 0.78 0.71 0.61 0.49 0.34 It 8.59 8.19 7.43 6.50 5.21 3.41 8.89 8.49 7.74 6.83 5.57 3.81 9.34 8.93 8.16 7,23 5.95 4.18 9,95 9.51 8.70 7.72 6.38 4.54 ' Average total atetoxin concentration: 2008 Crop - 0,42 ng/g AFT; 2006 Crop -1.26 ng/g AFT; hypothetical crops - 3.00 and 6.00 ng/g AFT. Almond lots accepted and rejected in the export market Table 2 results are sorted first by crop contamination level and then by sampling plan design and Table 3 simulation results are sorted first by sampling plan design and then by crop contamination level. The entire spreadsheet output for each of the 24 simulations can be obtained upon request using the email address of one of the authors ([email protected]). The first simulation in Table 2 is used as an example to describe simulation results when the USA almond industry uses a sampling plan identical to the EU aflatoxin sampling plan design (2x10 kg < 8 ng/g AFBj / 10 ng/g AFT) to test a USA crop that averages 0.42 ng/g AFT. The USA must sample 11,149 lots in order to accept 11,000 lots for export to the EU. The USA sampling plan removed 149 lots (1.36%) from the export market and reduced the average aflatoxin concentration from 0.42 ng/g AFT among all lots tested to 0.15 ng/g AFT among the 11,000 lots accepted for export to the EU. The average aflatoxin concentration among the 149 lots rejected for export in the USA was 23.4 ng/g AFT. When the 11,000 USA lots are imported into the EU, either 220,550 or 1,100 USA lots are tested for aflatoxin by EU authorities based upon a 2, 5, or 10% inspection rate, respectively. For a 5% EU inspection rate (550 lots), the simulation predicts that 2.8 lots are rejected and 547.2 lots are accepted by the EU sampling plan at destination. The total number of lots accepted at destination is 10,997.2, which is the 547.2 lots accepted by the EU sampling plan plus the 10,450 lots not sampled. The EU rejected 0.51% (2.8/550) of the lots tested at destination. The average aflatoxin concentration among lots tested, accepted, and rejected at destination by the EU sampling plan is 0.15,0.11 and 8.59 ng/g AFT, respectively. The percent lots rejected at destination and the average aflatoxin concentration in the lots tested, accepted, and rejected is the same regardless of the inspection rate (2,5 or 10%) used by EU authorities at destination. Even though the USA rejected 1.36% of the lots tested at origin and reduced the aflatoxin contamination by 64.2% (0.42 vs. 0.15 ng/g AFT) when using an identical sampling plan design at origin (2x 10 kg < 8 ng/g AfEl 110 ng/g AFT) as that used by the EU at destination, the EU still rejects 0.51% of the USA lots tested at destination. These results indicate that the USA cannot assume that all export lots will be accepted at destination as a result of using a sampling plan at origin that is equivalent (has the same level of false positives and false negatives) to that used by the EU at destination. The percentage of lots accepted and rejected in the USA and in the EU for various USA sampling plan designs and crop contamination levels (Tables 2 and 3) is discussed below. World Mycotoxin Journal 3 (2) Effect of increasing the crop contamination level on lots accepted and rejected at origin and at destination Reviewing Table 2 it can be seen for a given USA sampling plan design, as the crop contamination level increases from 0.42 to 6.0 ng/g AFT, the following happens: 1. More USA lots have to be sampled at origin to identify 11,000 lots for export (more USA lots are rejected at origin from the export market). 2. The average aflatoxin concentration among the 11,000 lots identified for export increases. 3. The number or percentage of lots rejected by the EU at destination increases. 4. The average aflatoxin concentration among all lots accepted by the EU at destination increases. Effect of decreasing the accept/reject limit of USA sampling plans on lots accepted and rejected at origin and at destination Reviewing Table 3 it can be seen that for a given crop contamination level, as the accept/reject limit decreases from a dual limit of 8 ng/g AFBj /10 ng/g AFT to a single limit of 8,6,4,2, or 0.5 ng/g AFT, the following happens: 1. More USA lots have to be sampled at origin to identify 11,000 USA lots for export (more USA lots are rejected at origin from the export market). 2. The average aflatoxin concentration among the 11,000 USA lots identified for export decreases. 3. The number or percentage of lots rejected by the EU at destination decreases. 4. The average aflatoxin concentration among all lots accepted by the EU at destination decreases. Reducing USA lots rejected in the EU The USA almond industry would like to keep the percentage of lots rejected by the EU as low as feasibly possible. Tables 2 or 3 show that simply having a USA sampling plan that is equivalent to or exceeding the performance of the EU sampling plan (2x10 kg S 8 ng/g AFBj / 10 ng/g AFT) does not guarantee that no USA lots will be rejected at destination or the percentage of USA lots rejected in the EU will not exceed some predefined unacceptable level if crop contamination levels exceed certain thresholds. Observing results in Table 2 or 3 suggests that that the percentage of lots rejected in the EU is a function of the average aflatoxin concentration among all 11,000 lots accepted in the USA for export. This can be better demonstrated by plotting (Figure 4) the percentage of lots rejected in the EU versus the average aflatoxin concentration among all 11,000 lots exported to the EU for all 24 simulation results in Table 2 or 3. Each point on the curve represents a combination of a potential USA sampling plan design and crop contamination level in Table 2 or 3. For example, a USA sampling plan would have to keep the average aflatoxin level 165 IIMMeretal. ! All almond crops and all USA sampling plans 100% USA testing 7 ±-• EU testing: 2x10 kg <; 10T/8B1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 11 1.2 1.3 1,4 1.5 1.6 Avg. afiatoxin concentration of tote accepted in USA {ng/g AFT) Figure 4. EU rejects versus average aflatoxin concentration in USA lots exported to the EU. among the 11,000 lots accepted for export to the EU below approximately 1.0 to 1.1 ng/g AFT to keep the percentage of USA lots rejected by the EU at destination below 5% (Figure 4). As crop contamination level increases, USA sampling plan would have to be designed (use more samples and/or lower accept/reject limits) so that lots accepted for export would average less than approximately 1.0 ng/g AFT. For example, if the USA crop contamination level before testing is 6.0 ng/g, a Codex style sampling plan (2x10 kg) with a 4 ng/g AFT accept/reject limit (2x 10 kg < 4 ng/g AFT) would be required to keep the average aflatoxin concentration among USA lots accepted for export below 1.0 ng/g and to keep USA lots rejected in the EU below 5% (Table 2 or 3). The USA sampling plan acts like a filter with varying degrees of efficiency at removing aflatoxin contaminated lots from the lot distribution being tested for possible export to the EU. The percentage of lots rejected at destination in the EU depends on the average aflatoxin contamination among all 11,000 lots accepted for export to the EU, which depends on the USA sampling plan design (as manifested by the accept/reject limit or the percentage of USA lots rejected at origin), which in turn depends on the USA crop contamination level. The EU sampling plan design also acts as a filter to remove contaminated lots at import. However, the EU sampling plan design was not changed or was held constant though all 24 simulations. 166 Predicting USA export lots rejected in the EU Because of the time required to transport USA lots from California to the EU, time for the EU to inspect USA imported lots, and time for the EU to report inspection results to the US, the USA almond industry cannot determine the consequences (number or percentage of lots rejected in the EU) of their sampling program for some period of time after the lots have been tested for aflatoxin in the USA. Using the observations in the previous paragraph about the effects of USA crop contamination levels and sampling plan designs on the percentage of lots rejected in the EU, there may be a way for the almond industry as a whole or for an individual exporter to predict in 'realtime' the percentage of lots that will be rejected several months later at destination in the EU. These two methods include keeping a record of either (a) the average aflatoxin concentration among sample test results associated with all USA lots tested for export (average aflatoxin concentration among all sample test results is a measure of the crop contamination level) or (b) the percentage of lots rejected at origin in the USA. Crop contamination levels (as measured by averaging all sample test results) and percentage of USA lots rejected at origin are two tangible parameters that can be measured in 'real-time' during the marketing of export lots by an World Mycotoxin Journal 3 (2) Almond lots accepted and rejected in the export market individual exporter or collectively by the industry as a whole to predict the percentage of USA lots rejected in the EU when those lots are eventually tested at a later date in the EU. From simulation results in Table 2 or 3, the percentage of lots rejected in the USA and in the EU by each of the six proposed USA sampling plan designs are plotted in Figure 5 versus crop contamination levels from 0.42 to 6.0 ng/g AFT. There are two sets of curves shown in Figure 5, one set of curves represents the percentage of lots rejected in the USA and one set of curves represents the percentage of lots rejected in the EU. Each curve in Figure 5 (for both the USA and the EU set of curves) is associated with a specific USA sampling plan design and provides an indication how each proposed USA sampling plan would perform (percent lots rejected) in the USA and in the EU for USA crop contamination levels from 0.42 to 6.0 ng/g AFT. Figure 5 can be used by the USA almond industry to keep the percentage of lots rejected by the EU as low as feasibly possible. If the USA almond industry plans to use a sampling plan that is equivalent to the EU sampling plan (2x 10 kg < 8 ng/g AFBj /10 ng/g AFT), then that USA sampling plan would be expected to keep USA rejects in the EU below 5% for crop contamination levels up to 4.25 ng/g AFT (plan 1 in the EU set of curves). The USA almond industry will reject about 12.8% of USA lots tested at origin (plan 1 in the USA set of curves). For example, if the USA almond industry is rejecting about 15% of lots tested for export using a USA sampling plan that is identical to the EU plan (plan 1 or 2x10 kg < 8 ng/g AFBt /10 ng/g AFT), then the industry is sampling a USA crop that averages 4.95 ng/g AFT (using plan 1 in the USA set of curves) and should expect 5.9% of lots tested by the EU (using plan 1 in the EU set of curves) to be rejected at destination (Figure 5). If this predicted percentage of USA lots rejected in the EU (5.9%) is unacceptable to the USA almond industry, they would have to change their sampling plan by lowering the accept/reject limit to 6 ng/g AFT (plan 3 or 2x10 kg s 6 ng/g AFT), which will bring EU rejects down to 4.8% when the USA is sampling a lot distribution at 4.95 ng/g AFT. If the USA changes the sampling plan to 2x 10 kg < 6 ng/g AFT, then lots rejected in the USA would increase from 15 to 17.5%. Reducing the accept/reject limit for the USA sampling plan has a desirable effect of reducing lots rejected at destination, but results in higher rejection rates in the USA and greater economic burden on exporters. Lots rejected in EU Lots rejected in USA USA almond sampling plans 1: 2x10 kg s 8 ng/g AFB/10 ng/g AFT 2:2x10 kg <;8 ng/g AFT 3:2x10 kg £6 ng/g AFT 4:2x10 kg s 4 ng/g AFT 5:2x10 kg £2 ng/g AFT 6:2x1 Okg z 0.5 ng/g AFT EU: 2x10 kg £ 8 ng/g AFB/10 ng/g AFT 0.0 15 20 25 Almond tots rejected (%) Figure 5. USA lots rejected in the USA and in the EU for six different USA sampling plan designs and a range in USA crop contamination levels. World Mycotoxin Journal 3 (2) 167 IMiMeretal. 4. Conclusions Simulation results show that even when the USA and the EU are using the same sampling plan with accepted protocols for sampling, sample preparation, and analytical methods to detect aflatoxin in shelled RTE almond lots, USA lots will be rejected at destination due to random variability associated with the aflatoxin test procedure (sampling, sample preparation, and analysis). The magnitude of the USA lots rejected at destination depends on the USA crop contamination level and the ability of the USA sampling plan to detect and remove aflatoxin contaminated lots from the export market. As the USA crop contamination level increases, the accept/reject limit of the USA sampling plan design will have to decrease to keep USA lots rejected in the EU to acceptable levels. After reviewing results of this study, the USA almond industry decided to modify the pre-March 2010 sampling plan (3x5 kg < 2.0 ng/g AFT) with a sampling plan identical to the new post-March 2010 EU sampling plan (2x10 kg < 8 ng/g AFB1 /10 ng/g AFT). The percentage of lots rejected in the EU is predicted to be minimal (less than 5%) unless crop contamination levels are encountered that exceed 4.2 ng/g AFT on the average. Methods were developed using the percentage of USA lots rejected at origin to help USA exporters recognise when to adjust their aflatoxin sampling plan design (lower the accept/reject limit) to keep USA lots rejected in the EU at a minimum, which will help balance the economic impact of increased rejections at origin. While model results were specific to simulating the sampling of RTE shelled almond destined for the EU, the model can be used for other mycotoxins, other commodities, and other destinations if the appropriate inputs can be provided. References Almond Board of California, 2009. Doc #5418. Almond Board of California, Modesto, CA, USA. Codex Committee on Contaminants in Foods (CCCF), 2008. Aflatoxin sampling plans for aflatoxin contamination in ready-to-eat treenuts and treenuts destined for further processing: almonds, hazelnuts, and pistachios, Alinorm 08/31/41, Appendix IX, 31st CAC Session, Geneva, Switzerland, 30 June to 4 July, 2008. Available at: www. codexalimentarius.net/download/report/700/al31_41e.pdf. 168 European Commission (EC), 2007. Commission Decision 2007/5631 EC of 1 August 2007, amending Decision 2006/504/EC on special conditions governing certain foodstuffs imported from certain third countries due to contamination risks of these products by aflatoxins as regards almonds and derived products originating in or consigned from the United States of America (notified under documents number C (2007) 3613). Official Journal of the European Union L 215:18-20. European Commission (EC), 2010a. Commission Regulation (EU) No 165/2010 of 26 February 2010, amending Regulation (EC) No 1881/2006 setting maximum levels for certain contaminants in foodstuffs as regards aflatoxin. Official Journal of the European Union L 50: 8-12. European Commission (EC), 2010b. Commission Regulation (EU) No 178/2010 of 2 March 2010, amending Regulation (EC) No 401/2006 as regards groundnuts (peanuts), other oilseeds, tree nuts, apricot kernels, liquorice and vegetable oil. Official Journal of the European Union L 52:32-43. Food and Agriculture Organization of the United Nations (FAO), 2003. Worldwide regulations for mycotoxins in food and feed in 2003. Food and Agriculture Organization of the United Nations, Rome, Italy, 165 pp. Giesbrecht, F.G. and Whitaker, T.B., 1998. Investigations of the problems of assessing aflatoxin levels in peanuts. Biometrics 54: 739-753. Whitaker, T.B., Giesbrecht, F.G. and Wu, J., 1996. Suitability of several statistical models to simulate observed distributions of sample test results in inspection of aflatoxin-contaminated peanut lots. Journal of AOAC International 79:981-988. Whitaker, T.B., Slate, A.B., Adams, J.G., Birmingham, T. and Giesbrecht, EG., 2010. 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 B, and total aflatoxins. World Mycotoxin Journal 3:35-44. Whitaker, T.B., Slate, A.B., Hurley, J.M. and Giesbrecht, F.G., 2006. Sampling almonds for aflatoxin, part II: estimating risk associated with various sampling plan designs. Journal of AOAC International 90: 778-785. Whitaker, T.B., Slate, A.B., Jacobs, M., Hurley, J.M., Adams, J.G., and Giesbrecht, F.G., 2006. Sampling almonds for aflatoxin, part I: estimation of uncertainty associated with sampling, sample preparation, and analysis. Journal of AOAC International 89: 1027-1034. World Mycotoxin Journal 3 (2)
© Copyright 2026 Paperzz