AOAC OMA# 2011.14 (Final Action Review)-OMB

2011.14 (Oct. 2018) – MTE-03 Final Action Review FOR ERP USE ONLY DO NOT DISTRIBUTE

considered invalid data and both replicates for that analyte and matrix were excluded from statistical treatment. This was most often seen for determination of low levels of Cu, Fe, and Mn in placebo products. All remaining data were statistically analyzed for overall mean, intra-laboratory repeatability (s r ), repeatability relative standard deviation (RSD r ), inter-laboratory reproducibility (s R ), reproducibility relative standard deviation (RSD R ) and HorRat. Cochran (p=0.025, 1 tail) and Grubbs (single and double, p=0.025, 2-tail) tests were utilized to determine outliers (7). In some instances, statistical outliers were identified but because they had low impact on precision characteristics, these were retained in the data set for calculation of the precision of the method. Raw data are presented in Tables S1-25 and method performance data are summarized by analyte and matrix in Tables 2011.14A-I. NIST SRM 1849a.- Laboratory 6 had 3 outliers (Ca, K, and Na) by the Cochran test, but only the Na data for Laboratory 6 were excluded from statistical analysis due to the wide spread between duplicates compared to all other participating laboratories. A summary of interlaboratory results for the NIST SRM 1849a compared to certified values obtained by ICP-AES are presented in Table 3. All of the mean values except that for Cu were within the certified range. The mean value for Cu was 19.47 mg/kg, 0.05 mg/kg below the certified range. Overall, the interlaboratory data indicated no bias of the candidate method relative to the ICP-AES certified values for NIST SRM 1849a. Repeatability ranged from 1.3% for Fe to 3.96% for K and reproducibility ranged from 2.54% for Na to 5.26% for Zn. HorRat values ranged from 0.50 for Cu to 1.12 for K, within the generally acceptable range of 0.5-2.0 (7).