AOAC Final Action Methods in 2017

542 P ACQUETTE & T HOMPSON : J OURNAL OF AOAC I NTERNATIONAL V OL . 101, N O . 2, 2018

Table 2015.06D. Performance of Method 2015.06 in the SLV and MLT and recommended analytical ranges

Parameter

Na Mg

P K Ca

Mn

Fe

Cu

Zn

Cr

Se

Mo

Low standard, µg/L

500 200 500 1000 1000

5.00

50.0

5.00

20.0

0.800

0.400

0.800

PLOQ, µg/L a

50

20 100 100 100

2.5

5.0

0.50

4.0

0.080

0.20

0.080

PLOQ, mg/100 g b

0.25 0.10 0.50 0.50 0.50

0.012 0.025 0.0025 0.020 0.00040

0.0010

0.00040

PLOQ meets SMPR? c PLOQ meets Codex? d

Yes Yes Yes Yes Yes

No

No

No

Yes

Yes

Yes

Yes

Yes Yes Yes Yes Yes

No

Yes

Yes

Yes

Yes

No

Yes

LOQ, mg/100 g e

0.11 0.0049 0.13 0.56 0.53 0.00050 0.0073 0.00048 0.056 0.00073

0.00031

0.00048

LOQ meets SMPR? c LOQ meets Codex? d Avg. repeatability, % f Avg. reproducibility, % f

Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

Yes Yes

1.5 1.5 1.9 1.4

1.8

2.1

2.9

2.3

1.6

3.0

3.3

2.1

Downloaded from https://academic.oup.com/jaoac/article-abstract/101/2/536/5653941 by guest on 05 May 2020

3.6 3.5 4.3 3.6

4.1

3.8

6.1

6.3

4.3

4.9

5.0

3.6

2 – 850 0.7 – 110 3 – 800 3 – 2000 3 – 1280 0.002 – 1.0 h 0.04 – 20 0.002 – 1.2 0.07 – 18 0.002 – 0.16 i 0.0006 – 0.050 h 0.0002 – 0.10

Recommended

analytical range, mg/100 g g

a The PLOQ is the lowest level at which the measured 3 day mean of a standard concentration, run as an unknown against the calibration curve, is within 5% of its nominal concentration. This is the preferred lower limit of the method, as it limits the amount of calibration bias in the result. How- ever, the sensitivity of ICP-MS typically affords a lower actual LOQ, as measured in the traditional way of measuring blank concentrations over sev- eral days. b PLOQ converted to a product concentration level using the typical dilution factor of 1 g to 50 mL for an RTF. c The lower limit of the analytical range as set forth in the SMPRs (Table 1). d Codex STAN 72 – 1981 minimum levels for infant formulas and Food for Special Medical Purposes products. e LOQs derived from measuring several digested and undigested blanks, each run on 5 separate days. Both sets of blanks yielded similar results. f For 18 samples tested at 10 laboratories in the MLT not counting the Adult High-Fat RTF sample or the samples in which the analyte was at or be- low the quantitation limit (as opposed to

the same 3 – 8% range over a wide range of concentrations. Rather, it is the relative level of the analyte above the LOQ that matters, and any analyte level above about 10 times the LOQ can be determined with roughly the same reproducibility that is due to the sources of variance of the method. For example, in this study, Mo had an average overall mean concentration of 0.0040 mg/100 g in the 18 SPIFAN matrixes and had an average reproducibility of 3.6% (both numbers can be calculated from the data in Table 3), whereas Na had an average level of 42.6 mg/100 g in these products — 10 650 times higher than Mo — and was determined with about the same reproducibility of 3.7% (Table 3). Indeed, if one does not include results lower than the PLOQ and results from a problematic high-fat sample, the average RSD R (see Table 6) for all 12 elements in all 18 SPIFAN matrixes and 6 IDF matrixes ranged from a low of 3.7% for Mg to a high of 7.9% for Fe — all in a narrow range because the sensitivity of the ICP-MS allows for determinations at analyte levels well above the LOQ. In Tables 3 and 4 repeatability and reproducibility that failed the SMPR criteria are indicated with a footnote. If one excludes adult high-fat ready-to-feed (RTF) sample 2, there are very few cases in which the method did not meet the criteria for products. Sample 2 is a difficult one to prepare because it has a high fat level and a lot of insoluble tricalcium phosphate in it, and thus it separates easily and is difficult to homogenize. A different batch

LOQ limits given in the table. This would allow measurement at the CODEX minimum limits for Mn and Cr, for example, but the expected repeatability would be above 5% RSD, and the expected reproducibility would be above 15% RSD, with nonlinearity of the calibration curves expected to produce a bias of >10% in the result.

Collaborative Study Results and Discussion

Method 2015.06 MLT statistical parameters for each of the 12 elements are given in Table 3 for the SPIFAN matrixes and Table 4 for the ISO/IDF matrixes. Parameters include the number of outliers, overall mean, s r and s R , RSD r and RSD R , repeatability and reproducibility limits, and the HorRat value. The HorRat value (7) is included for legacy purposes as a point of comparison with past collaborative studies, but the authors ’ opinion is that it is a meaningless figure of merit in comparison with reproducibility, per se, and does not need to be included in collaborative studies. The premise that reproducibility is expected to get worse as the concentration of the analyte diminishes is clearly not supported by the data of the present study nor most of the SPIFAN MLTs published to date: For example, studies on ultratrace minerals (3), pantothenic acid (8), iodine (9), and inositol (10) reported an RSD R of most samples in