Assessing aflatoxin B1 distribution and variability in pistachios: Validation of a Monte Carlo modeling method and comparison to the Codex method

• Published in: Food control Vol. 59; pp. 553 - 560
• Main Authors: Wesolek, N., Roudot, A.C.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2016
• Subjects: AFB1; Codex; Probabilistic; RASFF; Sample distribution; Sample distribution; AFB1; Probabilistic; RASFF; Codex
• ISSN: 0956-7135; 1873-7129
• DOI: 10.1016/j.foodcont.2015.06.034

Aflatoxin B1 (AFB1) has genotoxic carcinogenic effects with no threshold. This mycotoxin, produced by Aspergillus fungi species, can be found at a very low incidence in pistachios. More precisely, the contaminant is heterogeneously distributed. Moreover, it is usually found at high concentration levels in contaminated individual nut kernels. For this reason, it is crucial to assess AFB1 distribution and variability accurately in pistachio samples, in order to establish informed sampling schemes for health protection purposes. A modeling method using Monte Carlo simulations for AFB1 distribution in pistachio samples, developed in previous work, required validation. For this purpose, the simulation was adjusted to the distribution observed in 30 kg samples as calculated from the RASFF (Rapid Alert System for Food and Feed) levels reported by Germany for Iranian pistachio imports from September 2002 to September 2004, for the lots tested. As a final result, when the simulated distribution in 10 kg samples was computed from the prior simulation, it was very similar to the distribution observed in representative samples for the same sample size from the RASFF. The simulation method was then compared to another method, for AFB1 distribution and variability assessment, used by the Codex to design aflatoxin sampling plans for tree nuts. Indeed, the distribution observed in 30 kg samples from the RASFF was considered as an input distribution. The output distribution for 10 kg representative samples, as computed using the method, was close to the observed distribution. Results of both simulation methods were very similar. Finally, an extensive comparison of the principles of both methods was performed. •A Monte Carlo simulation distribution method was tested.•30 kg samples contamination distribution was simulated and adjusted to real data.•Computed 10 kg samples distribution was compared to observed data.•The same was achieved for the Codex method.•The predictive performance of both methods was compared.