| Abstract: | A computer model that accounts for sampling and analytical variability was developed to simulate the aflatoxin testing program administered by the North Carolina Department of Agriculture (NCDA) to regulate aflatoxin in corn meal. Monte Carlo solution techniques were employed to account for conditional probabilities that rise from multiple samples being used in the testing program. The NCDA testing program was then evaluated by applying the computer model to a hypothetical group of 1000 corn meal lots with the same distribution of aflatoxin concentrations as was observed among aflatoxin assays made by NCDA on commercial lots of corn meal from 1977 to 1980. The average of the 1000 lots assayed was 17.7 parts per billion (ppb). The model predicted that 79.5% of the lots would be accepted and 20.5% of the lots would be rejected by the NCDA testing program. The accepted and rejected lots contained an average of 5.7 and 64.2 ppb aflatoxin, respectively. The testing program accepted 7.3% of the lots with more than 20 ppb aflatoxin (consumers' risk) and rejected 1.0% of the lots with 20 ppb or less (processors' risk). A correct decision was made 94% of the time. |
