Recovery is judged satisfactory if it is not significantly
different from 100% at 95% confidence interval (
P
-value
>0.05) between themedian concentration and the certifiedor
referencevalues (Table2).Recoveryis judgedquestionable if
it is not significantlydifferent from100%at 99%confidence
interval (0.01 <
P
-value < 0.05) between the median
concentration and the certified or reference values (Table 2).
Recovery is judged unsatisfactory if it is significantly
different from 100% at 99% confidence interval (
P
-value
<0.01) between themedian concentration and the certifiedor
reference values (Table 2).
SLV
(
a
)
Studymaterial
.—Tenmaterials includedsix reference
materials (RM) of LGC Promochem LGC RM 7104
(sterilized cream), of existing NIST Standard Reference
Material (SRM) NIST SRM2384 (baking chocolate), NIST
RM8415 (whole egg powder), NISTSRM2383 (baby food
composite), NIST RM 8433 (corn bran), NIST SRM 1549
(nonfat milk powder), and four Nestlé well-characterized
reference samples of infant cereals with milk powder,
chocolatemilkpowder,dieteticmilkpowder (DMP1), andpet
food(PET).Thereferencevaluesofelementconcentrations in
the four in-house reference materials and their associated
reproducibility SD were obtained from Nestlé proficiency
tests performed by a significant number of internal and
external laboratories (between 12 and 45 laboratories,
depending on element and materials analyzed) using robust
statistics (28).
(
b
)
Microwave digestion optimization
.—Main reagents
formicrowave-assisteddigestionareHNO
3
andHCl.H
2
O
2
is
generally used to remove any remaining organic residues of
rich-fat or rich-carbohydratesamples.MDCsystems supplied
by manufacturers typically apply to a single food matrix or
minimize the analytical portionmass to account for themost
reactivefoodmatrix.Thisapproachisnotefficientwhenafull
range of foodmatrixes is analyzed on a routine basis, and an
additional predigestion step is often used to allow various
matrixes to be digested concurrently. Thus, approach of a
single optimized MDC program for duty vessels with only
HNO
3
operating up to 200 C, which has been adapted on a
CEMMarsXpress system,was favored: anoptimal analytical
test portionmass of 0.5 g [based on an empirical maximum
energy release by the food of 3 kcal (8)] was used, and
concurrent digestions of triplicates of three NIST-certified
RMs(bakingchocolate,NISTSRM2384;wholeeggpowder,
NIST RM 8415; and Baby Food Composite, NIST SRM
2383)wereperformed.Similarprograms tothatofCEMMars
Xpress systemwere adaptedonotherMDCsystems used for
MD3 andMD4 ruggedness tests. An optimized program for
MDOsystemCEMMDS2000 (Table3)was alsodeveloped
in parallel, and used for all matrixes during ruggedness test
MD2 andbyone volunteer laboratoryduring theRT.
(
c
)
Linearity
.—A calibration curve was obtained from
seven standards, including a blank (Std0) and six
concentrations of the standard solution (Std1–Std6).
Weighted linear and quadratic regression analyses were
performed and correlation coefficient (R
2
) values were
calculated.
(
d
)
LOD and LOQ
.—The instrumental detection (DL)
and quantification (QL) limits expressed in mg/L were
estimatedusingtheSDapproachbyanalyzing10replicatesof
the lowstandard solution (Std1) according to the formula:
DL=3 SDof themeanof Std1determinations (
n
=10)
QL=10 SDof themeanof Std1determinations (
n
=10)
The DL and QL values of elements in real samples were
hence estimated assuming a sample weight of 0.5 g and a
dilution to 50mL (i.e., a dilution factor of 100). Verification
of QL estimation was performed by checking satisfactory
performance characteristics (accuracy and precision) for
elements present in real matrixes at these estimated
concentrations, e.g., Ca, Cu, Fe,K,Na, P in cornbran (NIST
RM8433),Mg, Zn in sterilized cream(LGCRM7104), and
Mn innonfatmilkpowder (NISTSRM1549).
(
e
)
Selectivity
.—Elementmedian recoverywas calculated
usingall thelinesdisplayedinTable5fromduplicatesofeight
different solutions of spiked foodgrade salts (Table2)witha
spiking concentration corresponding to Std4 prepared from
theCPI commercial solution. Robust
t
-test for comparisonof
median recovery and SD (i.e., median of averages of
duplicatesfoundperelementusingallanalytical lines)ofeach
spiked element in all food-grade salts with and without
additionof ionizationbufferCs 1%(w/v)was performed.
(
f
)
Accuracy
.—Three replicate test portions (
r
= 3) were
collected fromeach of seven (
n
= 7) different food samples
populating four of the nine food triangle sectors, and then
weredigestedandanalyzedon8days(
d
=8)correspondingto
nine series (i.e., 9 elements) of 168 data (
n r d
= 168).
Recoverymedian values (ratio of found and certified values)
and associated RSD were determined. Accuracy was
statistically evaluated.
P
OITEVINETAL.
: J
OURNALOF
AOACI
NTERNATIONAL
V
OL
. 92,N
O
. 5, 2009
1491
Table 8. EstimatedLODandLOQfor thenineelements
in foodmatrixesobtained fromrecommendedand
alternateelement lines
Element
LOD
a
,mg/L LOQ
b
,mg/L LOQ
c
,mg/kg
Calcium
0.5
1.50
150
Copper
0.006
0.02
2
Iron
0.03
0.1
10
Potassium
0.5
2
200
Magnesium
0.2
0.5
50
Manganese
0.0002
0.0005
0.05
Sodium
0.4
1
100
Phosphorus
0.4
1
100
Zinc
0.02
0.05
5
a
Instrumental LODcalculated insolution.
b
Instrumental LOQcalculated insolution.
c
LOQestimated in foodmatrixes (dilution factor 100).
MTE-01
FORWORKINGGROUP/ERPUSEONLY
DONOTDISTRIBUTE
