SPIFAN Nutrients ERP Book_9-29-15

Fluor-02 SLV FOR ERP USE ONLY DO NOT DISTRIBUTE

checked daily throughout the batch run. The system was recalibrated when the percent error of the check standards was > 5%.

Error, % = calculated concentration – actual concentration actual concentration × 100

Accuracy

Accuracy was determined by fortifying the placebo samples (prior to sample preparation) at 50 and 100 % of the typical level of fluoride found in a powdered infant formula. These recovery experiments were performed over multiple days on two different instruments. In addition, NIST SRM 1849a was analyzed during the course of this validation. Table 2 contains recovery results for fluoride on the five placebo samples respectively. The percent fluoride recovered ranged from a low of 73.8% to a high of 166%. The percent relative standard deviations (RSD) ranged from a low of 2.45% to a high of 15.3%. The overall mean recovery is 96.7% with a RSD of 31.4%. With the exception of three recovery value at 73.8%, 77.1% and 166%, all data meets the criteria specified in the SMPR (80-120% recovery). Results on NIST SRM 1849a can be found in Table 3 &4. The mean value of fluoride was found to be 208 µg/100g (n = 30) with % RSD 5.94. The certified value and range for fluoride are not provided in the SMPR. Repeatability was evaluated on all 12 fortified AOAC SLV test material matrices including the NIST SRM 1849a using multiple instruments. Samples and standards were prepared freshly every day over six days and analyzed in triplicate to calculate the intermediate and overall reproducibility. Precision data for all samples can be found in Tables 5. Fluoride concentration ranged from 52.6 to 231 µg/100 g. Overall precision was < 19% for all samples. Percent relative standard deviations ranged from a low of 2.57% to a high of 18.5 %. Precision

Reproducibility

Reproducibility was determined in the precision experiments conducted over multiple days and involving different lots of reagents made fresh each analysis day. In addition, two instruments were used during this validation study.

Limit of Detection (LOD) and Limit of Quantitation (LOQ)

The LOD was estimated as 3× the signal-to-noise ratio (S/N) and the LOQ as 10× the S/N. The system baseline noise for the LOD and LOQ was determined by measuring peak-to-