It should go without saying that obtaining a precise and accurate estimate of mycotoxin content in a commodity bulk lot is extremely important for products intended for feed use. The mycotoxin estimated value should be thought of as the end of a multi-stage, analytical chain in which sampling, sample preparation and analysis are all critical phases that contribute to the accuracy of the final estimate.
Test procedure critical
The mycotoxin test procedure means the fate of a multiple-ton lot is decided on a few grams of product measured in the units of ppb. The knowledge of the critical points and the variability associated with each phase of the analytical chain is fundamental in order to reduce the total error and increase the certainty and reliability of the final mycotoxin contamination estimate.
It's even possible to calculate the acceptance probability of rejecting a good lot or accepting a wrong lot. Error reduction in the mycotoxin test procedure reduces the number of misclassified lots and is critical for correctly accepting or refusing a lot and minimizing economic or consumer risk.
Mycotoxin sampling plan
Putting a mycotoxin sampling plan into place is key to improving accuracy of mycotoxin estimates. A mycotoxin sampling plan of commodities may be defined as a test procedure combined with a sample acceptance limit. The final mycotoxin analytical result is influenced by the effect of the three distinct phases characterizing the test procedure: sampling, sample preparation and analysis.
The total error associated with the analytical value may be described as the sum of errors resulting from each of the three phases. Since each phase is associated with a certain level of uncertainty, it is not possible to obtain a quantitative value for the mycotoxin contamination with 100 percent certainty.
Furthermore, for a given mycotoxin measurement plan, it is possible to calculate the acceptance probability to reject a good lot (type I error, false positive sample) or to accept a wrong lot (type II error, false negative sample) as a function of the toxin concentration and the risk associated to the specific plan. The determination of type I or II errors, which depict an economic loss or a health consumer risk, respectively, may be described graphically by the plot of mycotoxin concentration versus acceptance probability. This plot takes the name of Operating Characteristic curve (OC curve). The areas delimited by the curve and the limit level describe respectively economic (area above the curve) or consumer risk (area below the curve) magnitude. Each mycotoxin measurement plan is described by a unique OC curve. Since the slope of the OC curve has high economic and health relevance, it is crucial to increase the slope of the OC in order to reduce risks.
All critical points have to be considered in order to reduce sampling error and increase the reliability of the final sample. Sample size, collection of a sufficiently large number/size of incremental samples, choice of the sampling points, homogeneity of sample components in terms of size and specific weight, sample preparation, and analytical methods all have to be planned in order to reduce sampling error.
The first step of the analytical chain is the sampling phase. Sampling error is the largest contributor to total error since fungal development and mycotoxin production are "spot processes" significantly affected by crop variety, agronomic practices, weather conditions during growing and harvest, storage and processing conditions, and toxigenic potential of the different mold species. Bulk moisture can facilitate the development of localized clumps particularly rich in molded kernels. These small percentages of extremely contaminated portions ("hot spots") are randomly distributed in the lot (average value usually registered 0.1 percent). This condition can lead to an underestimation of the real level of mycotoxin if a too small sample size without contaminated particles is analyzed or, instead, to an overestimation of the true level in the case of a too small sample size featuring one or more contaminated particles. In order to perform reliable sampling, each unit within a lot must have the same probability of being chosen. If this is the case, sampling is random and will be theoretically characterized by the absence of systematic errors, reduced variability and increased reliability in applying information from the samples to the entire lot.
Several authors evaluated and quantified the contribution of sampling phase error to final error of the entire test procedure in several products and for several commodities, such as corn and peanuts when contaminated with aflatoxins, fumonisin and ochratoxin A, and reported values from 55 to 77 percent of total variability supported by the sampling phase. Different values and a lower variability have been observed in association with the sampling of wheat contaminated with deoxinilvalenol at 22 percent of the total error.
Aspects of this sampling step, such as aggregate sample size, number/size of incremental samples selected in a commodity bulk lot and techniques applied for physical sample selection, are critical for reducing uncertainty. For instance, increasing aggregate sample size by a factor of four reduces sampling variance to ¼. A reduced number and/or too small of incremental samples (inadequate mass) represent the two most frequent and important errors responsible for magnifying total uncertainty.
The techniques used for physically collecting and selecting samples also are an important source of uncertainty. There is generally agreement that larger bias is observed in cases of static sampling (sampling of resting bulk by probes) than in cases of dynamic sampling (sampling of a moving stream of material), enabling the ability to obtain a high number of incremental samples with regular frequency along the entire flow. Other sources of uncertainty can be found in tools used for samples collection.
In order to reduce sampling error, another important critical point of the sampling step to be considered is the level of contamination, because sampling variability has been reported to be a function of mycotoxin concentration. The variability associated to sampling increases with mycotoxin concentration of a bulk lot and/or a decreasing incremental sample size. Sample preparation
The second step of the analytical chain is the sample preparation phase. The sample must be carefully mixed and ground in order to obtain a homogeneous sample. A small sub-sample is taken for the analysis. In a wide range of mycotoxin and commodity combinations, with reference to total error, sample preparation error is greatly inferior to sampling error, with an average value of 25.9 ± 15.9 percent.
Milling the aggregate sample is very important for correct sample preparation. Increasing the degree of sample grinding leads to more uniform particle distribution and therefore to a variability reduction. Commodity, mycotoxin type, particle dimension and sample size are all factors influencing sampling preparation error. Small particle size is associated with a reduction of sampling preparation variability as well as low level of contamination and/or big sample size. Two different approaches may be used for mixing and reducing particle dimensions: dry milling or slurry mixing.
When dry milling is applied, more differences are evidenced in the variability associated with sample preparation than in the case of slurry mixing in terms of characteristics and performance of the mills and sample size. The variability associated to different mills employed for the same matrix may differ by up to two times.
The analytical phase is the last step of the mycotoxin test procedure and the last step of the sampling procedure. The Association of Official Analytical Chemists (AOAC) and the European Standardization Committee (CEN) maintain a list of standardized methods of analysis that are fully validated by collaborative studies and associated with analysis performance data. Their aim is to evaluate compliance with performance criteria reported in specific international regulations. High Pressure Liquid Chromatography (HPLC), Enzyme-Linked Immuno-Sorbent Assay (ELISA), Thin Layer Chromatography (TLC), Gas Chromatography (GC) and Fluorimetry are considered relevant techniques for mycotoxin quantification. The number and complexity of steps for a specific method for mycotoxin evaluation can amplify variation among results of different aliquots taken from the same laboratory sample. Analytical error, the variability associated to this step, is the lowest contributor to total error (10.2 ± 8.93 % total variance).