159 / 258
1704
Pacquette & Thompson:
J
ournal of
AOAC I
nternational
V
ol.
98, N
o.
6, 2015
factor with which to work. The Study Director allowed this small
variation to the method as it will certainly not affect the validity
of results if the correct dilution factor is used.
Participants were reminded more than once that the written
(now official) method mandates the use of several QC/system
suitability solutions including a blank check (must be less
than the PLOQ), a calibration verification standard (must be
within 4% of its nominal concentration before and after samples
are run), and a control sample (the concurrent analysis of SRM
1849a). The criterion for results was not explicitly stated because
these laboratories do not have working control charts for such;
however, the SRMexpected results were known from the prework
phase, and the laboratories could see the SRM results concurrent
with their samples. No laboratories indicated they discarded any
data because of these QC solutions failing.
A final key suitability requirement was the analysis of
duplicates that had to agree within 10% for Cr, 7% for Se,
and 5% for Mo. These duplicates are not to be confused with
the blinded duplicates supplied in the collaborative study. The
duplicate requirement is present not so much to improve the
confidence interval by using the mean of two results (although that
can be useful for concentrations near the PLOQ), but rather to
indicate the presence of substantial indeterminate errors before
the data are allowed to enter the pool of multilaboratory data. A
10% agreement between duplicates is a common, if somewhat
arbitrary, criterion used in many GB (China) official methods, and
others. In this case, the SLV and MLT data indicate that a 10%
criterion is well above the excellent repeatability or intermediate
precision expected of this method (
see
SLV data later in this
report), but this level is maintained for Cr because small levels
of Cr contamination were sometimes unavoidable and usually
irrelevant to results since most adult products had relatively high
levels of Cr (infant formulas are not fortified with Cr or Mo).
With the inclusion of the other QC tests in this method, especially
the use of the control sample, the possibilities of other sources of
indeterminate errors are small [e.g., pipetting the wrong amount
of internal standard (IS) or a poorly made set of standards], and
so Se and Mo have duplicate RSD requirements closer to 3×
the typical short term precision of about 1.5–2.0%. Indeed, the
expected duplicate precision for Se was originally set at 5% RSD
for this method, which was optimal for the authors’ laboratories,
but analysis of these MLT data indicated too many rejections at
that level, and so a 7% RSD requirement is now set for Se. The
requirement for Mo is still 5% duplicate precision. Table 2 shows
the number of failures in analyzing the 14 MLT samples using the
original criteria (i.e., 5% RSD duplicate precision for Se). There
are an inordinate number of failures (4/14 or almost 30%) for
Laboratory 1 Cr results; this is the laboratory whose data were
later entirely rejected from the study. However, it appears that
only Laboratory 1 had this problem with Cr determination. In
contrast, five laboratories had more than one sample rejected
for Se when the duplicate RSD criterion were set to 5%. If the
duplicate precision criterion was set to 7% RSD, only eight total
failures occurred instead of the 19 shown in Table 2, out of a
total of nine laboratories × 15 samples = 135 determinations,
or about a 6% rejection rate. This may be higher than the
<1% rejection rate for Mo because Se concentrations are
routinely low, about 2–3x above the PLOQ in all the samples
tested. Also, the laboratory that had the most Se data rejected,
Laboratory 9, was also the one that had a compromised PLOQ
as shown in Table 1. This underscores the importance of having
optimal sensitivity for Se analyses in infant/pediatric formulas.
Generally, we have observed that ICP/MS units that are not fitted
with hydrogen gas for collisional reaction/reduction of argon
interference cannot readily obtain the 0.2 ppb PLOQ in solution.
As can be seen, setting a single criterion for duplicate precision
to cover all concentration levels encountered and for all matrixes
is difficult, but this does not mean it should not be done. This is
perhaps the best way to avoid out-of-specification results due to
systematic errors and rejecting the data before any unnecessary
retesting or regulatory action begins, and this kind of suitability
criterion should be strongly considered for any dispute resolution
method, even chromatography-based methods in which it may
take much more time to get the duplicate result.
Upon completion of the sample analyses, participating
laboratories were asked to send all of their data to the Study
Director. An Excel spreadsheet was supplied by the Director, with
a template for adding the sample weights, duplicate results, and
spaces for all the calibration and QC results. Participants were
also asked to report any deviations to the method and any relevant
comments based on their experiences with the method.
All data were statistically analyzed in a spreadsheet (5) using
AOAC INTERNATIONAL guidelines to determine overall
mean, repeatability SD (s
r
), RSD
r
, reproducibility SD (s
R
),
RSD
R
, and Horwitz ratio (HorRat). Cochran’s (
P
= 0.025,
one-tail) and Grubbs’ (single and double,
P
= 0.025, two-tail)
tests were used to determine statistical outliers.
SPIFAN SMPRs for repeatability were ≤5% RSD and
requirements for reproducibility were ≤15% RSD in products
above a concentration of 10 ng/g Se and 20 ng/g Cr/Mo on an
RTF basis.
Method
The Final Action method, as now published (6) and given
below, is the updated version the participants used for this study.
In particular, Ge was substituted as the IS for Ni, Cr, and Mo,
and there is an option to analyze more elements concurrently. The
QC/system suitability was more explicitly stated, and the revised
7% duplicate criterion for Se added.
Table 2. MLT duplicate samples failing to meet the original
duplicate precision criterion of 10% RSD for Cr and 5% RSD
for Se and Mo
Lab
No. Cr failures No. Mo failures No. Se failures
a
1
4
0
1 (0)
2
0
0
0
3
0
0
3 (1)
4
1
0
2 (0)
5
1
1
4 (2)
6
7
8
0
0
0
9
0
0
5 (4)
10
1
0
4 (1)
11
0
0
0
a
Failures under 7% RSD criterion shown for Se in parenthesis.
159