50 / 71
1446
Pang et al.:
J
ournal of
AOAC I
nternational
V
ol.
98, N
o.
5, 2015
(
2
)
GC/MS/MS.—
Regarding GC/MS/MS: 14 laboratories
established 556 matrix-matching ISTD calibration curves,
respectively, for 20 pesticides in green tea and oolong tea,
among which those with R
2
≥0.995 reached 549, making up
98.7%; seven had R
2
between 0.990 and 0.995, accounting for
1.3%, and none R
2
less than 0.990.
(
3
)
LC/MS/MS.—
Concerning LC/MS/MS, 24 laboratories
established 958 matrix-matching ISTD calibration curves,
respectively, for 20pesticides ingreen tea andoolong tea samples,
among which those with R
2
≥0.995 reached 893, accounting for
93.2%; 34 had R
2
between 0.990 and 0.995, making up 3.7%,
and 11 R
2
<0.990, making up 1.2%. The above-mentioned
data show that there are 2079 with R
2
≥0.995 out of the 2153
matrix-matching ISTD curves established by 29 laboratories,
accounting for 96.6%. It demonstrates that the matrix-matching
ISTD calibration curves adopted for the method are capable of
realizing accurate quantification for the majority of pesticides by
different laboratories using three different types of instruments.
Error Analysis and Traceability
The 6638 effective data derived in this study were inspected
with Grubbs and Dixon tests for outliers, with 187 outliers
obtained, accounting for 2.8%. The distribution of outliers
derived from these three methods, GC/MS, GC/MS/MS, and
LC/MS/MS, for different teas, different pesticide varieties, and
different samples are tabulated in Supplemental Tables 19–21.
The outliers from 10 samples of three categories determined
by the three different methods for two teas in Supplemental
Table 19–21 are summarized and tabulated in Supplemental
Table 22. Distribution of outliers for different laboratories and
different samples is listed in Supplemental Table 23.
(a)
By GC/MS.—
Supplemental Tables 19, 22, and 23 show
that each of 16 laboratories analyzed 20 pesticide residues in
eight samples (excluding two blank samples) and obtained
1977 effective data; the application of Grubbs and Dixon tests
revealed 65 outliers making up 3.3%. They came from 11
laboratories, with 32 from Laboratory 19, accounting for 49.2%
of the total outliers; outliers from other 10 laboratories are 33,
making up 50.8%, and outliers from each of these laboratories
are less than eight, due to accidental deviations.
Thirty-two outliers from Laboratory 19 all came from No. 4
and No. 5 green tea incurred samples and No. 9 and No. 10
oolong tea aged samples. Test results from these four samples
are 20–50% higher than those from other laboratories. Review
of the experimental raw data record found that this laboratory
established the matrix calibration curves on June 8, 2013 and
then the fortified sample of green tea (No. 1 and No. 2) and
oolong tea (No. 6 and No. 7) were analyzed, while No. 4 and
No. 5 green tea incurred samples and No. 9 and No. 10 aged
samples were tested on June 12, 2013 with an interval of 4 days.
Exact reasons that caused such deviations failed to be found
in the raw data record, and it is assumed that during this 4 day
interval instrument conditions had possibly changed, which may
be the cause of such systematic deviations. The collaborator
considered that his instrument was stable since comparison of a
QC sample had been conducted, but no reasonable explanation
has been yet offered for the cause of such errors.
Laboratory 19 did not carry out continuous inspection of
samples after establishing calibration curves, with a time
interval of 4 days before continuing inspection, which led to
the results from this laboratories 20–50% higher than those
from other laboratories. Such practice is not recommended by
the Study Director, who points out that the matrix-matched
calibration standards should be prepared and used ONLY for the
quantitative analysis of the samples prepared at the same time
under the same conditions. As far as this point is concerned,
Laboratory 19 deviated from the collaborative study operational
procedures unconsciously, which inevitably led to relatively
large deviations in test results compared to those from other
laboratories.
(b)
By GC/MS/MS.—
Supplemental Tables 20, 22, and
23 show that 1704 effective data were obtained from
determination of 20 pesticides in eight samples (excluding
two blank samples) by 14 laboratories; Grubbs and Dixon test
inspection was adopted, and 65 outliers were found (making
up 3.8%). Sixty-five outliers came from 11 laboratories, of
which 25 were from Laboratory 21, accounting for 38.5%
of the total outliers; 19 from Laboratory 18, making up
29.2% of the total outliers; and seven from Laboratory 27,
making up 10.8% of the total outliers. Outliers from these
three laboratories total 51, accounting for 78.5% of the total
outliers; outliers from eight other laboratories total 14, only
making up 21.5%, and outliers from each laboratory are less
than three due to accidental errors.
Regarding Laboratory 21, there are 23 of 25 outliers from
No. 6 and No. 7 oolong tea fortification samples, and the
analytical results from these two samples are 40% lower than
those from other laboratories. In addition, there are very big
Table 10. Distribution range of RSD
r
, RSD
R
, and HorRat values for aged samples
Parameters
of method
efficacy
RSD
r
, %
RSD
R
, %
HorRat
Range
<8
8–15
>15
<16
16–25
>25
<0.50 0.50–1
.00 1.01–2.00 >2.00GC/MS (16 laboratories)
Oolong tea
20(100)
a
0
0
18(90)
2(10)
0
1(5)
19(95)
0
0
GC/MS/MS (14 laboratories)
Oolong tea
19(95)
1(5)
0
0
6(30)
14(70)
0
0
20(100)
0
LC/MS/MS (24 laboratories)
Oolong tea
15(75)
5(25)
0
0
8(40)
12(60)
0
8(40)
12(60)
0
Total
54(90)
6(10)
0
17(28)
17(28)
26(43)
1(2)
27(45)
32(53)
0
a
Data in parentheses are the percentages.