AOACSPIFANMethods-2017Awards

165 / 258

1710 Pacquette & Thompson: J ournal of AOAC International Vol. 98, No. 6, 2015
Collaborative Study Results and Discussion
The key precision performance metrics of the multilaboratory
study per sample matrix are summarized in Table 6. The RSDr
derived from analysis of the blinded duplicates was roughly the
same as for the known duplicates (raw data not shown from the
participating laboratories, but the RSDr obtained is consistent
with the intermediate precision data from the SLV shown in
Table 3). For instance, none of the seven matrixes produced an
RSDr higher than the method’s 10% RSD duplicate criterion for
Cr, 7% for Se, or 5% for Mo. Note, however, that repeatability
of Se for three of the seven matrixes was between 5 and 7%,
further justification for the change of this QC criterion to 7%.
In terms of the repeatability SMPR, two matrixes had RSDr
of >5% for Cr, as did three matrixes for Se. The highest RSDr
observed for the blinded duplicates was 7.0%, and in this case,
as well as the other four cases, the corresponding reproducibility
was only slightly higher.
The RSDR of method 2011.19 for each matrix was, on the
average, about half of the SMPR of 15%. HorRat were similarly
low, averaging 0.46 for Cr, 0.27 for Mo, and 0.32 for Se. The
authors’ opinion is that the RSDR expected from this study is
a function of how far above the instrument quantification limit
we are at the determination stage, not of the absolute level of
the analyte. Methods with good sensitivity, good linearity over
the calibration range, and adequate required system suitability
should be able to produce comparable reproducibility at the low
ppb level, and this appears to be supported by other SPIFANMLT
studies (publications in progress).
The individual sample results submitted by each laboratory
are given in Tables 7–9. Each value given is the mean of known
duplicates, prepared per the method. Then, the blinded duplicate
results are shown for each participating laboratory, for each
matrix. The footnotes indicate which samples were rejected,
either by the method’s QC criteria, or by the AOAC-supplied
statistical package (5). The laboratories could have analyzed new
samples and obtained data to replace the rejected results, but there
was not enough time to do so, or perhaps they did not realize
this was an option. Although there were five cases in which both
blind duplicate samples were rejected (thus Table 6 records the
number of laboratories as seven for that matrix), the footnotes
indicate that retaining the data would keep the RSDR under 15%
in all but one case. The data in Tables 7–9 also indicate why
Laboratory 1 data were totally excluded from the study; except
for a few Mo results, its data were significantly lower than
that of any other laboratory across the board. Also, Laboratory
1 stood out as having the most problems with sensitivity and
linearity (Table 1), and perhaps contamination was an issue due
to the number of (known) duplicate failures for Cr (Table 2). It
may be a coincidence, but that laboratory was using the oldest
ICP/MS instrument, a PerkinElmer ELAN DRC-e, which may
not have had the capability to do the required collisional/reaction
chemistry to eliminate low-mass interferences.
Comments about the performance of the method were
requested. One laboratory pointed out that the 10% powder
reconstitution in the method was different than the 11.1%
reconstitution recommended by SPIFAN and proceeded to use
the latter (it made no discernible difference). Another comment
was that the ICP/MS instrument model DRC-e could not use
ammonia gas for Se determination, which may be the reason for
Laboratory 1’s exclusion from this study.
Conclusions
AOAC Method 2011.19 was successfully studied in
collaboration by eight laboratories using multiple ICP/MS
instrument models and testing a variety of infant, pediatric, and
adult nutritional matrixes. The method demonstrated acceptable
repeatability and reproducibility and met the SPIFAN SMPRs for
reproducibility for all seven matrixes analyzed.
Recommendation
The multilaboratory collaborative study data were
summarized and presented to the AOAC ERP in
September 2014. After reviewing the data, the AOAC ERP voted
to move AOAC 2011.19 to Final Action status, and the method
was approved by the AOAC Official Methods Board as a Final
Action method (6).
Acknowledgments
The authors would like to thank the following collaborators
and their associates:
Yue Fenpeng, Chinese Academy of Inspection and Quarantine
(CAIQ), Beijing, China
Fan Xiang, Entry-Exit Inspection and Quarantine, Shanghai,
China
Yue Zhang and Shuqi Zhang, Zhejiang Test Academy,
Hangzhou City, China
Sudhakar Yadlapalli, First Source Laboratory Solutions,
Hyderabad, India
Isabelle Malaviole, Laboratory Aquanal, Pessac, France
Ashutosh Mittal, Syngene International Ltd, Bangalore, India
Michael Gray, Mead Johnson, Evansville, IN
Marissa Feller, Covance Laboratories, Madison, WI
Diana Mould and Michael Farrow, U.S. Food and Drug
Administration, Atlanta, GA.
References
(1) Cubadda, F., Raggi, A., Testoni, A., & Zanasi, F.
(2) Sharpless, K.E., Thomas, J.B., Christopher, S.J., Greenberg, R.R.,
Sander, L.C., Shantz, M.M., Welch, M.J., &
Wise, S.A.
(3) Pacquette, L., Szabo, A., & Thompson, J
(4) AOAC SMPR 2011.009 (2012) J. AOAC Int. 95, 297. http://
dx.doi.org10.5740/jaoac.int.11-0441
(5) AOAC Interlaboratory Study Workbook for Blind (Unpaired)
Replicates (2013) Version 2.1, AOAC INTERNATIONAL,
Rockville, MD
(6) Official Methods of Analysis (2012) 19th Ed., AOAC
INTERNATIONAL, Rockville, MD.
Method 2011.19
165