4.6.1. Precision

To evaluate the precision profile of the method, data generated each validation level were subjected to analysis of variance (ANOVA). When applying this model, as defined by ISO 5725 [33], the measured test response Yijk (the AFM1 mass fraction in the present case) is defined as the true value (TV) plus the contribution of 3 components:

$$\mathbf{Y}\_{\rm ijk} = \mathbf{T}\mathbf{V} + \mathbf{D}\_{\rm i} + \mathbf{M}\_{\rm ij} + \mathbf{R}\_{\rm ijk} \tag{1}$$

where Di is the between-day variability, Mij is the between-matrix (milk batches from di fferent farms) variability, and Rijk is the within-day variability. The within-day variability gives the precision under repeatability conditions, whereas the sum of all components gives the intermediate precision. The statistical assessment was done with the software package MINITABTM Statistical Software for Windows (Version 15).

#### 4.6.2. Cut-O ff Value

The measured levels (ng/kg) of samples containing AFM1 at STC were taken as basis for the calculation of the cut-o ff value. According to Regulation 519/2014/EU the following equation was used:

$$\text{Cut}\,off = \mathcal{R}\_{\text{STC}} - \text{traulue}\_{(0.05)} \times \mathcal{S}\mathcal{D}\_{\text{STC}} \tag{2}$$

where the *RSTC* is the mean level of AFM1 (ng/kg) calculated from all 24 experiments performed on samples containing AFM1 at STC, *SDSTC* is the corresponding standard deviation of intermediate precision as defined in the previous paragraph, and *tvalue*(0.05) is the one tailed t value for a rate of false negative results of 5%.

#### 4.6.3. False Suspect and False Negative Rate

Using the cut-o ff value and the results from the analysis of negative samples the rate of false suspect results was estimated by first calculating the t-value as follows:

$$\text{travalue} = \frac{\begin{pmatrix} \text{cut } off \, f - \text{mean}\_{\text{ncg}} \end{pmatrix}}{SD\_{\text{ncg}}} \tag{3}$$

where *meanneg* is the mean value of the results obtained from the 24 experiments on the negative samples and *SDneg* is the corresponding standard deviation of intermediate precision.

From the obtained t-value, based on the degrees of freedom calculated from the number of experiments (23 in the present case), the false suspect rate results (probability) for a one tailed distribution was calculated using the spread sheet function "TDIST" from Microsoft Excel.

The false suspect rate for samples containing AFM1 at 50% STC was calculated by applying the same procedure using the mean value of the results obtained from the 25 experiments on samples containing AFM1 at 50% STC and the relevant standard deviation of intermediate precision.

Finally, the false negative rate for samples containing AFM1 at levels above the STC was estimated by calculating the t value as specified here:

$$\text{traulue} = \frac{\left(\text{mean}\_{\text{>STC}} - \text{cutoff}\right)}{SD\_{\text{>STC}}} \tag{4}$$

where *mean*>*STC* is the mean value of the results obtained from the experiments on the samples containing the analyte above the STC, cut-o ff is the value established as above, and *SD*>*STC* is the corresponding standard deviation of the intermediate precision. The probability corresponding to the calculated t value with a one-tailed distribution gives the rate of false negative results for the samples containing the analyte at levels higher than STC.

**Author Contributions:** Conceptualization, V.M.T.L. and C.v.H.; methodology, V.M.T.L. and I.P.; formal analysis and validation, N.G., R.B. and B.C.; statistics, C.v.H.; writing—original draft preparation, V.M.T.L., I.P.; writing—review and editing, supervision, V.M.T.L., M.P. and A.F.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.
