Journal of Food Composition and Analysis 24 (2011) 111120

Contents lists available at ScienceDirect

Journal of Food Composition and Analysis

journal homepage: www.elsevier.com/locate/jfca

Original Article

Simultaneous analysis of 21 elements in foodstuffs by ICP-MS after closed-vessel

microwave digestion: Method validation
Sandrine Millour, Laurent Noel *, Ali Kadar, Rachida Chekri, Christelle Vastel, Thierry Guerin
Agence Francaise de Securite Sanitaire des Aliments, Unite contaminants inorganiques et mineraux de lenvironnement, equipe metaux lourds et elements mineraux, 23,
Avenue du General de Gaulle, F-94706 Maisons-Alfort Cedex, France

A R T I C L E I N F O

A B S T R A C T

Article history:
Received 27 January 2010
Received in revised form 8 April 2010
Accepted 14 April 2010
Available online 30 July 2010

This paper describes a validation process for the simultaneous analysis of 21 elements: lithium (Li),
aluminium (Al), vanadium (V), manganese (Mn), cobalt (Co), nickel (Ni), copper (Cu), zinc (Zn), gallium
(Ga), germanium (Ge), arsenic (As), strontium (Sr), molybdenum (Mo), silver (Ag), cadmium (Cd), tin
(Sn), antimony (Sb), tellurium (Te), barium (Ba), mercury (Hg) and lead (Pb) in food samples by
inductively coupled plasma-mass spectrometry (ICP-MS) after closed-vessel microwave digestion. This
validation was realized in parallel with the analysis of the 1322 food samples of the second French Total
Diet Study (TDS) by the National Reference Laboratory (NRL) of the French Food Safety Agency (AFSSA).
Several criteria such as linearity, limits of quantication (LOQ), specicity, precision under repeatability
conditions and intermediate precision reproducibility were evaluated. Furthermore, the method was
supervised by several internal and external quality controls (IQC and EQC). Results indicate that this
method could be used in the laboratory for the routine determination of these 21 essential and nonessential elements in foodstuffs with acceptable analytical performance.
2010 Published by Elsevier Inc.

Keywords:
Validation process
Essential and non-essential trace elements
ICP-MS
Microwave digestion
Quality controls
Foodstuffs
French Total Diet Study
Dietary exposure
Toxic food contamination levels in a
population
Food analysis
Food and nutrition data
Food composition
Toxicological reference value

1. Introduction
In 2006, the French Food Safety Agency (AFSSA) started the
second French Total Diet Study (TDS) to estimate dietary exposure
to essential and non-essential elements in 1322 different food
samples. The goal of the TDS is an indirect assessment of foods
found in a typical market basket and prepared for normal
consumption. Often used, it is the most reliable way to estimate
the dietary intake and the exposure to toxic chemicals by large
population groups (WHO GEMS/Food, 2006; Zukowska and Biziuk,
2008). First, the TDS assesses the impact of common household
cooking methods on the decomposition of the least stable chemical
compounds and the formation of new chemical compounds (WHO
GEMS/Food, 2006; Zukowska and Biziuk, 2008). Then data on the
concentration of contaminants, minerals and nutrients, combined
with dietary consumption data, make it possible to evaluate the
levels of dietary exposure in population groups. Thus, the

* Corresponding author. Tel.: +33 1 49 77 26 90; fax: +33 1 49 77 26 50.

E-mail address: l.noel@afssa.fr (L. Noel).
0889-1575/$ see front matter 2010 Published by Elsevier Inc.
doi:10.1016/j.jfca.2010.04.002

estimated intakes are compared with nutritional and toxicological

reference values, such as the Recommended Daily Allowances
(RDA) or the Recommended Daily Intake (RDI) for nutrients and the
Provisional Tolerable Weekly Intake (PTWI) for toxic elements
(WHO, 1995, 2010).
In the context of this second French TDS, the National Reference
Laboratory (NRL) for heavy metals put in place the validation of a
multi-elemental method by inductively coupled plasma-mass
spectrometry (ICP-MS) to conrm that the analytical procedure
employed was suitable for its intended use (multi-elemental
routine analysis for studies of exposure through dietary intake and
monitoring plans) (Fig. 1).
First, it was necessary to dene the analytical requirements of
the method (which element in which type of food, concentration
range to be covered, acceptable uncertainty in the result, etc.) in
accordance with the second French TDS monitoring exigencies
(Millour et al., 2010). The method used for the analysis relied on
the accredited method by the French Committee of Accreditation
(COFRAC) for Cd, Pb, As and Hg in foodstuff of animal origin (Noel
et al., 2005). This validation was based on the determination of
generally well-dened and relevant performance characteristics

[()TD$FIG]

112

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

Fig. 1. Validation process of the ICP-MS methodology.

(linearity, limit of quantication (LOQ), specicity, precision under

repeatability conditions and intermediate precision repeatability)
NF EN 13804 (AFNOR, 2002a); NF V03-110 (AFNOR, 1998). Before
ICP-MS methodology was routinely used for the TDS on unknown
samples, a set of experiments was required to evaluate whether
the method would be able to meet the criteria described above.
This preliminary validation study generated enough information to
guarantee that the analytical method will provide, in routine use,
accurate measurements without being affected by other elements
present in the sample, assuming that the methodology and related
handling procedures remained unchanged. From this preliminary
validation study, the decision to accept the analytical procedure
set-up as valid for routine use was also based on the quality of
future results, which were produced using the analytical procedure
during the TDS, through an in-study validation and the implementation of internal quality controls (IQC) (Fig. 1). These IQC were
essential to oversee the study and to ensure the reliable results in
each analytical sequence, such as the calibration curve to monitor
the linearity, blanks and the checking of LOQ, internal standards to
obtain information about sensitivity drift in different mass region
and to follow matrix effects, a standard every eight samples and at
the end of the sequence to control the instrumental drift, spiked
standard solutions on unknown samples to control the specicity,
certied reference material (CRM) to monitor the trueness and
duplicates to control the repeatability (Millour et al., 2010). Several
computer-based aids were also used to make data processing
easier and to follow the evaluation of individual and overall IQC
with their specic acceptance criteria and nally, calculation of all
percentage to conclude whether the results are acceptable or not.
The aim of the present work was to demonstrate that the
method developed by the NRL about the analysis of 21 essential
and non-essential trace elements (lithium (Li), aluminium (Al),
vanadium (V), manganese (Mn), cobalt (Co), nickel (Ni), copper
(Cu), zinc (Zn), gallium (Ga), germanium (Ge), arsenic (As),
strontium (Sr), molybdenum (Mo), silver (Ag), cadmium (Cd), tin
(Sn), antimony (Sb), tellurium (Te), barium (Ba), mercury (Hg) and
lead (Pb)) by ICP-MS after closed-vessel microwave digestion could
apply to numerous types of matrices representative of French diet,
while keeping adequate analytical performances. The results
obtained with IQC used in the study and external quality controls
(EQC) have also been presented.

2. Material and methods

2.1. Reagents and gas
All solutions were prepared with analytical reagent grade
chemicals and ultrapure water (18 MV cm) obtained by purifying
distilled water with the Milli-QTM PLUS system associated with an
Elix 5 pre-system (Millipore S.A., St Quentin-en-Yvelines, France).
(a) Nitric acid: Suprapur HNO3 (67% v/v) was purchased from VWR
(Fontenay-sous-Bois, France).
(b) Standard solutions: (1) Standard stock solutions containing
1000 mg L1 of each element and gold (Au) were purchased
from Analytika (Prague, Czech Republic) and were used to
prepare calibration standards. To avoid the mercury memory
effect, gold chloride was added to maintain Hg as Hg2+ in
solution (Lo and Wai, 1975; Varian, 1996). Working standards
were prepared daily in 6% of HNO3 and were used without
further purication.
(c) Tuning solution: a 10 mg L1 multi-element solution (Merck,
Damstad, Germany) was used to prepare a tuning solution
containing several elements such as indium (In), uranium (U),
Ba, Li, that allowed to sweep a wide range of mass.
(d) Internal standard solutions: 1000 mg L1 standard stock solutions of scandium (Sc), yttrium (Y), In, rhenium (Re), bismuth (Bi)
were purchased from Analytika (Prague, Czech Republic). The
ve internal standards were added to all samples, calibration
standards and blanks at the same concentration, to obtain
information on changes in sensitivity in different mass regions.
The software handles multiple internal standard in interpolated way: the target analyte is converted by a factor proportional
to its mass difference between two internal standards in a prole
of internal standard across the mass range.
(e) Certied reference materials: BCR 278R (mussel tissue) from
the Community Bureau of Reference, IAEA 407 (sh tissue)
from the International Atomic Energy Agency, INCT MPH 2
(mixed herbs of Polish origin) from the Institute of Nuclear
Chemistry and Technology, were all purchased from LGC
Standards (Molsheim, France).
(f) Ultrapure grade carrier (argon (Ar), 99.9995% pure) was
supplied by Linde (Montereau, France).

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

2.2. Samples
Food matrices used for the validation method have been chosen
among the 1322 food composite samples of the second French TDS.
Each of the samples was composed of up to 15 sub-samples of
equal weight of the same food item and was prepared as normally
consumed by the average consumer. These composite samples
were grouped into 41 food groups representative of the French
food habits as, for example, vegetables, fruits, molluscs and
crustaceans, shes, and cooked dishes, etc. (Sirot et al., 2009). The
method was validated by analysing various types of matrices from
different foodstuffs of animal and vegetal origin (salmon, shrimp,
oyster, pork, carrot, spinach, apple, yoghurt, milk, etc.) as
diversied as possible.
2.3. Sample digestion procedure
Samples were digested using a Multiwave 3000 microwave
digestion system (Anton-Paar, Courtaboeuf, France), equipped
with a rotor for eight 80 mL type quartz vessels (operating
pressure, 80 bar).
The sample digestion procedure was performed according to
the NF EN 13805 standard (AFNOR, 2002b). Before use, quartz
vessels were decontaminated in a bath of 10% of nitric acid 67% (v/
v), then rinsed with ultrapure water and dried in a 40 8C oven. From
0.2 to 0.6 g of dietary samples were weighed precisely in quartz
digestion vessels and wet-oxidized with 3 mL ultrapure water and
3 mL ultrapure HNO3 (67%, v/v) in the microwave digestion
system. One randomly selected vessel was lled with reagents only
and taken through the entire procedure as a blank. The digestion
procedure has previously been optimised (Noel et al., 2003). After
cooling at room temperature, sample solutions were quantitatively
transferred into 50 mL polyethylene asks. Then, 100 mL of the
internal standard solution (1 mg L1) was added to a nal
concentration of 2 mg L1; the digested samples were then lled
with ultrapure water to the nal volume before analysis by ICP-MS.
2.4. ICP-MS determination procedure
2.4.1. Instrumentation
ICP-MS measurements were performed using a VG Plasma
Quad ExCell (Thermo Electro, Courtaboeuf, France). The sample
solutions were pumped by a peristaltic pump from tubes arranged
on a CETAC ASX 500 Model 510 auto-sampler (CETAC, Omaha,
Nebraska, USA). Further details of instrumental settings are given
in Table 1.
2.4.2. Optimization
The isotopes 7Li, 27Al, 51V, 55Mn, 59Co, 60Ni, 63Cu, 65Cu, 64Zn,
66
Zn, 71Ga, 74Ge, 75As, 88Sr, 92Mo, 98Mo, 107Ag, 111Cd, 114Cd, 118Sn,
121
Sb, 125Te, 137Ba, 138Ba, 202Hg, 206Pb, 207Pb, 208Pb were selected as
analytical mass in ICP-MS standard mode. For six elements (Cu, Zn,
Mo, Cd, Ba and Pb), several specic isotopes were considered and
monitored according to the sensibility of the element and/or
possible isobaric and polyatomic interferences (Nardi et al., 2009).
For example, 111Cd was analysed rather than 114Cd because of the
isobaric interference of tin (114Sn). Furthermore, interference
correction formulas were included in the software of the ICP-MS
instrument for 75As (Noel et al., 2005).
Torch position, ion lenses and gases output were optimised daily
with the tuning solution (1 mg L1) to carry out a short-term
stability test of the instrument, to maximise ion signals and to
minimise interference effects from polyatomic ions and doubly
charged ions. In all experiments, a relative standard deviation (RSD)
of 3% was achieved. To obtain precise and accurate results, signal of
elements was monitored by real time display (RTD) which showed

113

Table 1
VG PlasmaQuad operating conditions and acquisition parameters.
Operating conditions
Nebulizer
Spray chamber
Sampling cone
Skimmer cone
RF power
Reected power
Standard mode
Plasma gas ow
Nebulizer gas ow
Auxiliary gas ow
Expansion stage
Intermediate stage
Analyser stage
Acquisition parameters
Mass range
Number of channels
Dwell time
Number of sweeps
Total acquisition time
Peak jumping mode

Concentric type pumped at 0.9 mL min1

Scott-type double-pass water cooled
Nickel, 1.0 mm orice
Nickel, 0.75 mm orice
1350 W
<5 W
15 L min1
0.73 L min1
0.90 L min1
2.7 mbar
2.0  104 mbar
4.6  106 mbar
7208 a.m.u
500
160 ms
500
60 s

the constant sensitivity over time for the selected masses and the
ratios of masses with three readings, calculated for each sample.
2.4.3. Calibration
Multi-element standard solutions were used for external
calibration. Five standards with weighed linear regression and
internal standardisation were prepared at concentrations ranging
from 0 to 20 mg L1 for Li, V, Co, Ni, Ga, Ge, Mo, Ag, Cd, Sn, Sb, Sr, Te,
Ba, Hg, Pb, from 0 to 50 mg L1 for Mn, Cu, As and from 0 to
100 mg L1 for Al and Zn. The calibration curve was drawn from six
points, including the calibration blank (0 mg L1).
2.5. Calculations and statistical methods
Concentrations were expressed in milligrams of elements per
kg (mg kg1) of fresh material. The various criteria of the method
for these 21 elements were estimated according to the NF EN V03110 (AFNOR, 1998), NF EN V03-115 (AFNOR, 1996), NF EN 13804
(AFNOR, 2002a) and NF ISO 5725 (ISO, 1994) standards. Data were
subjected to linear regression and analysis of variance using
Microsoft Excel software.
3. Results and discussion
Several criteria were evaluated to monitor the linearity on a
range of dened concentrations, to dene LOQ, to follow the
specicity and the absence of interferences, to control the
repeatability (precision under repeatability conditions) and the
trueness and then, to monitor the intermediate precision
reproducibility (Table 2).
3.1. Range of linearity of the calibration graph
The concentration ranges of the analytes were dened
according to our experience as NRL for heavy metals and our
analytical work for the rst French TDS (Leblanc et al., 2005). ICPMS is well known for beneting from a linear broad dynamic range
(about eight orders of magnitude). Although sample dilutions
could reduce matrix effects, it was not feasible to dilute each
sample individually due to the wide range of concentrations
encountered (going to mg kg1 to mg kg1 according to elements)
and the wide variety of studied foodstuffs. Experiments were
performed with standard solutions of ve concentration levels and
internal standardization (Table 2). The statistical tests were
derived from the analysis of variance applied to the least-squares

114

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

Table 2
Criteria selected for the ICP-MS validation method.
Characteristics to be evaluated

Procedures to be followed

Number of determinations

Linearity

Analysis of standards

Limits of quantication

Analysis of blanks

Specicity (interferences, matrix effects)

Analysis of spiked samples at

appropriate concentrations
Replicate analysis of samples

20 standard solutions of ve concentration levels (020 mg l1, 50 mg l1 or

100 mg l1) depending on the analyte.
At least 21 blank samples after correction for typical sample weight
(0.3 g and 0.6 g) and dilutions
At least 10 spiked samples over working range

Precision under repeatability conditions

Trueness
Intermediate precision reproducibility

Reference materials (CRMs, if available)

Analysis of duplicate sample solutions
from the same homogeneous sample

regression (according to the Fischer tests) (Feinberg and De la

Rochette, 1997); NF EN V03-110, (AFNOR, 1998). Results obtained
for the statistical tests (Data not shown) indicated that the linear
regression model was acceptable for the 21 analytes in the dened
range. Apart from these statistical tests, a determination coefcient r2 > 0.995 was considered acceptable and checked, as IQC,
during the study (Millour et al., 2010). The calibration graph was
determined at the beginning of the analytical sequence and was
assessed by measuring a mid-range standard (5 or 10 mg L1
depending on the elements) after the calibration curve, every ve
samples and at the end of the run (to estimate the variation of the
instrumental response and to check the quality of the analytical
sequence). At the end of the TDS, 96% of calibration curves had a
satisfactory determination coefcient (r2 > 0.995 based on a
minimum of four points) and 95% of mid-range standards had a
concentration between 70 and 130% of the theoretical value for all
the 21 elements (quality criteria dened for the study) (Millour
et al., 2010).
3.2. Limits of quantication (LOQ)
This parameter was essential because the calculation of
exposure results obtained by multiplying individual food consumption data by the average concentrations of elements was
dened from this value. As dened in a previous work (Noel et al.,
2005), the best approach for determining LOQ in ICP-MS was with
the standard deviation based on the sample blanks. So, LOQ were
rst evaluated according to the NF EN 13804 standard (AFNOR,
2002a) and were dened six times the standard deviation of the
average from the 21 different sample blanks quantied on several
months after correction for sample weight and dilutions. Table 3
reported LOQ for each element based on a 0.6 g test portion and a
dilution 1/50. These limits were equivalent to those of the rst TDS
(Leblanc et al., 2005) for the elements of interest. In the second
time, LOQ were also experimentally veried at the concentration
level of the evaluated LOQ, according to the XP T 90-210 standard
(AFNOR, 2003). It consisted in measuring a spiked standard
solution on ultrapure water corresponding to the target value of
LOQ of each element and in monitoring that the found value was
acceptable. Every measurement of concentration (corresponding
to ve batches of concentrations between 0.02 and 5 mg L1) was
fractionated in 10 identical solutions and analysed in repeatability
conditions. Results of veried LOQ with both conditions to be
satised (trueness and precision) are presented in Table 3.
It should be noted that the LOQ were in compliance with the
Commission Regulation (EC) N8 333/2007 (EC, 2007), laying down
the methods of sampling and analysis for the ofcial control of the
levels of Pb, Cd and Hg in foodstuffs. In routine analysis, blanks
values could vary considerably due to possible contamination
hazards during the preparation of samples (air, water, reagents,
glass, etc.) (Pizzolon and Hoenig, 2005). During each analytical

2 measurement values of at least 10 various samples at different levels in

the dened linear range
At least 5 analysis of 3 CRMs (mussel, sh, herb)
At least 10 duplicate sample solutions (CRMs if available), by different
analysts over a period of several months

sequence, as IQC, three blanks were analysed to monitor possible

cross contamination or memory effects. At the end of the TDS, 72%
of blank values were lower than LOQ. Blanks with concentrations
>LOQ were due to random pollution for few elements (Li, Ga, Cu
and Zn) (Millour et al., 2010). It should be noted that the mean
concentration of blanks was subtracted from the value of samples.
These results were not complied with the QC but it was considered
not critical, because values for Li and Ga remained low despite
levels >2  LOQ and those of Cu and Zn stayed very low compared
to the high levels found in food.
3.3. Specicity
To check the specicity of the method (lack of interferences),
recoveries of spiked standards solutions in the dened calibration
range were measured in several matrices from animal and vegetal
origin. Spikings (n  10) were realized before the digestion of
samples, subjected to the complete analytical procedure. The
analytical procedure to realize this test was the following one.
Three samples of the same matrix were analysed: one sample with
a known spiked multi-elemental standard solution and two
samples without spiking. Thus, it was possible to calculate the
mean content of each analysed element and to verify if theoretical

Table 3
LOQ, values of trueness and precision criteria for validation of a targeted LOQ.
Element

LOQ dened
(NF EN 13804)
(mg kg1
fresh weight)

Trueness
criteria
(critical
value = 10.0)

Precision
criteria
(critical value = 20%)

Targeted LOQ
veried
(XPT 90-210)
(mg kg1
fresh weight)

Li
Al
V
Mn
Co
Ni
Cu
Zn
Ga
Ge
As
Sr
Mo
Ag
Cd
Sn
Sb
Te
Ba
Hg
Pb

0.001
0.472
0.020
0.015
0.002
0.053
0.023
0.100
0.002
0.003
0.010
0.013
0.011
0.084
0.001
0.003
0.001
0.002
0.090
0.010
0.005

7.4
0.4
8.5
1.8
2.5
7.2
0.1
5.5
4.0
0.3
1.5
3.0
5.7
9.5
6.5
0.6
7.3
0.4
0.02
0.4
6.4

8%
4%
4%
12%
4%
7%
19%
12%
8%
16%
4%
6%
2%
1%
6%
18%
4%
15%
9%
7%
14%

0.002a
0.417e
0.017c
0.017c
0.002a
0.083d
0.017c
0.083d
0.002a
0.002a
0.008b
0.017b
0.008b
0.083d
0.002a
0.002a
0.002a
0.002a
0.083d
0.008b
0.005a

Spiked deionized water at 0.02 mg L1 (a), 0.1 mg L1 (b), 0.2 mg L1 (c), 1 mg L1 (d)
and 5 mg L1 (e).

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

115

Table 4
Results of specicity test.
Li

Al

Mn

Co

Ni

Cu

Zn

Ga

Ge

As

0.981
0.032
0.170
0.327

0.995
0.038
0.739
1.842

1.018
0.028
0.200
0.279

1.013
0.032
0.860
0.768

0.993
0.017
0.188
0.168

0.962
0.027
0.016
0.282

0.964
0.025
0.105
0.571

0.980
0.033
1.849
1.546

0.962
0.016
0.112
0.166

0.972
0.011
0.023
0.123

1.004
0.016
0.392
0.380

3.169
0.608
0.521

3.250
0.139
0.401

3.106
0.635
0.717

3.250
0.414
1.120

3.106
0.391
1.118

3.169
1.370
0.055

3.106
1.431
0.184

3.250
0.620
1.195

3.106
2.293
0.671

3.250
2.457
0.188

3.169
0.237
1.032

Sr

Mo

Ag

Cd

Sn

Sb

Te

Ba

Hg

Pb

Slope
Standard deviation (s) of the slope
Intercept
s of the intercept

0.984
0.033
0.007
0.351

0.954
0.023
0.295
0.229

0.987
0.072
0.195
0.772

0.956
0.024
0.180
0.206

0.960
0.013
0.197
0.135

0.920
0.029
0.465
0.289

0.925
0.038
0.296
0.405

0.915
0.038
0.542
0.387

0.970
0.015
0.153
0.150

0.962
0.025
0.157
0.258

TCritical value
Tobserved hypotheses test of slope 6 1
Tobserved hypotheses test of intercept 6 0

3.250
0.495
0.021

3.106
2.048
1.291

3.250
0.184
0.253

3.250
1.843
0.876

3.106
3.008
1.456

3.106
2.810
1.609

3.250
1.992
0.730

3.106
2.217
1.399

3.106
2.000
1.021

3.169
1.510
0.609

Slope
Standard deviation (s) of the slope
Intercept
s of the intercept
TCritical value
Tobserved hypotheses test of slope 6 1
Tobserved hypotheses test of intercept 6 0

The various matrices analysed for the test of specicity are the following ones: mutton, semi skimmed milk, veal, chocolate drink, corn, cre`me frache, apple compote, melon,
apple, ratatouille, leek, turkey cutlet, coffee, steamed salmon, orange juice, beer, water, dried, fruits, gruyere type cheese, onion, Pollack, canned tuna, turkey, cucumber,
courgette, buns.

spiking was found (by calculation taking into account the various
weights of the three samples).
Added concentrations varied between 2 and 100 mg L1
depending on the element. It was very difcult to obtain
representatively spiked standard solutions on unknown samples
since initial concentration differed greatly from element to
element within a given matrix and from one matrix to another.
The spiked concentration of the element should normally be equal
to 50200% of the initial concentration of the element in the
sample, except when the sample concentration was close to the
LOQ. In the latter case, any added concentration was considered
appropriate. The line of regression was tested against the line of
unity (slope = 1, intercept = 0) by simultaneously testing the
hypotheses of slope different from 1 and intercept different from
0, using Students t-test; for the 21 elements, Tobserved < Tcritical value

(Table 4). The slope and intercept of this regression line were not
different from 1 and 0, respectively. As a consequence, the
specicity of the method was considered acceptable for all
elements. This conclusion was conrmed by IQC because the
analytical sequence included two different spiked standard
solutions on random samples to check the specicity of the
method during the study. At the end of the TDS, more than 90% of
all spiked standards were between 80 and 120% of the target value,
except for Al and Zn (Millour et al., 2010). However, 8897% of
spikings were between 70 and 130% for these two elements.
Furthermore, the ve added internal standards (Sc, Y, In, Re, and Bi)
permitted to correct the drift and to compensate the matrix effects
in ICP-MS. At the end of the TDS, 86% of internal standards (81% of
Sc, 85% of Y, 88% of In and Re and 89% of Bi) were well included in
the range dened by the QC (30% of the theoretical value).

Table 5
Results of repeatability (precision under repeatability conditions) test (mg kg1).
sra
Li
Al
V
Mn
Co
Ni
Cu
Zn
Ge
As
Sr
Mo
Ag
Cd
Sn
Sb
Te
Ba
Hg
Pb
a

0.0004
0.4590
0.0646
0.0037
0.0010
0.0106
0.0214
0.2625
0.0006
0.0031
0.0178
0.0047
0.0317
0.0013
0.0018
0.0003
0.0006
0.0203
0.0016
0.0013

s2r

CVr mean (%)

7

1.98E
2.11E1
4.18E3
1.39E5
1.02E6
1.12E4
4.59E4
6.89E2
3.04E7
9.69E6
2.35E4
2.19E5
1.01E3
1.76E6
3.31E6
9.11E8
3.51E7
4.11E4
2.57E6
1.59E6

4.7
13.5
3.9
6.7
5.9
5.6
3.7
5.5
9.2
9.0
2.1
6.6
9.8
4.5
10.5
10.5
8.6
4.8
4.3
5.3

Max (s)2
6

1.40E
8.04E1
2.50E2
1.10E4
9.70E6
1.03E3
2.61E3
5.13E1
1.52E6
8.86E5
2.17E3
1.63E4
8.47E3
1.05E5
2.40E5
4.09E7
1.70E6
2.46E3
8.33E6
8.37E6

Cochran test (critical value = 0.4800)b

0.3516
0.1908
0.3015
0.3890
0.4725
0.4618
0.2838
0.3724
0.2509
0.4571
0.3436
0.3722
0.4215
0.2993
0.3631
0.2244
0.2421
0.2994
0.1620
0.2635

21 various samples at different levels in the dened linear range for the 21 elements and n = 2 measurement values.
Critical value at the 1% signicance level (p = 21; n = 2).
The various matrices analysed for the test of repeatability are the following ones: liver, cooked ham, boiled mussel, shrimp, cooked Pollack, canned tuna, salmon, foie gras,
mashed potatoes, veal, turnip, potatoes chips, peas, tomato, bean, spinach, whole milk, non fat yoghurt, tebbouleh, chicken, scallop, boiled salmon, smoked salmon, mutton,
juice of pineapple, sh cakes, French fries, corn, fromage blanc, grape juice, cartonned soup, chocolate cake, chocolate biscuit, sweet pancake, mayonnaise, semolina, tomato
sauce, babybel1 type cheese, chocolate mousse, cheese, gruyere cheese, goat cheese, hard-boiled egg, pizza, dried fruits, oyster, dark chocolate, pear, Roquefort type cheese
and pate.
b

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

116

3.4. Repeatability (precision under repeatability conditions)

3.5. Trueness

The precision of analytical procedure was usually expressed as

the variance or the coefcient of variation (CV) of a series of
measurements. The homogeneity of within-test variances was
veried by a Cochrans test. The Cochrans test statistic C was
compared to the critical values for Cochrans test at the 1%
signicance level. Since all C-statistic values were less than these
critical values, there were therefore no deviating measurement
values according to Cochrans test and the precision under
repeatability conditions was estimated constant. In Table 5,
Cochrans test shows that the variances of the 20 various samples
were homogeneous for the all elements with max (s)2 < critical
value and the precision results show acceptable repeatability
(within-test variance) for 20 elements. Values of mean CVr did not
exceed 10% except for Sn and Sb (10.5%) and Al (13.5%). For Ga, 99%
of concentrations in analysed food matrices were below the LOQ.
During the TDS, all samples were analysed in duplicate to
eliminate any batch-specic error, to verify the homogeneity of
samples and to monitor the repeatability. Thus, variability was
calculated for all duplicates with a QC dened acceptable if CV
20% (mean CVr = 6.6% and k = 3 established for Pb, Cd, As and Hg
(Noel et al., 2005)) when the mean value was 5  LOQ or had a CV
40% when the mean value was close to the LOQ. At the end of the
study, 96% of duplicates were considered acceptable, satised to
quality criterion (Millour et al., 2010) and conrmed this criterion
of performance.

The trueness could have been assessed using an available CRM,

which covered the elements of interest with suitable concentrations (Jorhem, 2004; Thompson et al., 2002). Despite the increasing
number of available CRMs, it was not always possible to nd CRMs
typical of the matrices subjected to analysis or that contained all
elements, notably in the case of TDS. For this reason, some
compromise was necessary. By analysing the same selected CRMs
(with different compositions and/or different analyte concentrations) throughout the study, bias and random error could be
detected, notably through the use of control charts (Jorhem, 2004).
Three CRMs (mussels, herbs, and sh) were selected to perform the
analysis of the trueness (n = 5) according to the V03-115 standard
(AFNOR, 1996). These CRMs were chosen thanks to their
availability, their matrix composition and the considerable
number of certied values: BCR 278R (mussels) had 7 certied
values; and IAEA 407 (sh) and INCT MPH 2 (mixed herbs) had 14
certied values, and one indicative value for Mo in INCT MPH 2 and
Sn in IAEA 407. All results obtained were corrected for moisture
content. Unfortunately, four elements (Ga, Ge, Te, and Ag) had no
certied or indicative values (In IAEA 407, an indicative value of
0.037 mg kg1 for Ag was <LOQ set in Table 3).
Tables 68 summarized the results obtained for 17 elements on
each CRM. The measured mean was generally in the condence
interval (CI) of CRMs, except for seven elements in BCR 278R (Mn,
Cu, Zn, As, Cd, Hg and Pb); eight elements in IAEA 407 (Al, Ni, Zn, As,

Table 6
Results of trueness and intermediate precision reproducibility in IAEA 407.
CRM

Mean (mg kg1)

s (mg kg1)

CVr (%)

CVR (%)

CI (IQC) (mg kg1)

0.040
0.060

6.1
3.2

10.8

0.3770.993a
0.6200.740b

0.8
1.4

7.6
3.4

13.4

7.620.0a
12.415.2b

1.48
1.43

0.06
0.09

4.0
2.1

8.5

1.001.86a
1.341.52b

Theoretical value
Theoretical value

3.46
3.52

0.12
0.08

3.5
0.8

7.2

2.464.58a
3.443.60b

Co

Estimated mean
Theoretical value

0.11
0.10

0.007
0.01

6.4
3.3

10.2

0.070.13a
0.090.11b

Ni

Estimated mean
Theoretical value

0.90
0.60

0.07
0.05

7.6
2.8

14.2

0.330.87a
0.550.65b

Cu

Estimated mean
Theoretical value

3.35
3.28

0.22
0.08

6.7
0.8

9.9

2.304.26a
3.203.36b

Zn

Estimated mean
Theoretical value

64.9
67.1

3.3
0.9

5.2
0.4

5.6

47.087.2a
66.367.9b

As

Estimated mean
Theoretical value

13.7
12.6

0.7
0.3

5.1
0.8

6.1

6.918.3a
12.312.9b

Sr

Estimated mean
Theoretical value

3
5

2.0
1.3

Cd

Estimated mean
Theoretical value

0.184
0.189

0.014
0.004

1.1
0.7

5.8

Sn

Estimated mean
Theoretical value

0.10
0.10

0.01
0.01

8.5
13.3

12.5

0.060.14a
0.060.13b

Sb

Estimated mean
Theoretical value

0.008
0.011

0.001
0.001

6.0
3.0

11.6

0.0060.016a
0.0100.012b

Hg

Estimated mean
Theoretical value

0.224
0.222

0.006
0.006

2.7
0.9

6.2

0.1550.289a
0.2160.228b

7.4
5.6

11.2

Li

Estimated mean
Theoretical value

Al

Estimated mean
Theoretical value

Estimated mean
Theoretical value

Mn

Pb
a
b

0.661
0.685
11.2
13.8

144
130

Estimated mean
0.08
0.006
Theoretical value
0.12
0.020
h
i
CVR M
p with k = 3, n = 1 samples and M = reference value or indicative value for Sn.
: CI M  k  100
n
Condence interval of the CRM.

10

91169a
125135b
0.1320.246a
0.1850.193b

0.080.16a
0.100.14b

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

117

Table 7
Results of trueness and intermediate precision reproducibility in INCT MPH 2.
CRM

Mean (mg kg1)

s (mg kg1)

CVr (%)

CVR (%)

Al

Estimated mean
Theoretical value

491
670

13.3
55.3

2.7
0.6

12.7

Estimated mean
Theoretical value

0.023
0.082

2.8
5.7

8.5

0.6661.238a
0.7891.115b

Mn

Estimated mean
Theoretical value

8
6

4.3
2.0

7.9

134248a
179203b

Co

Estimated mean
Theoretical value

0.217
0.210

0.013
0.008

5.8
4.0

9.5

0.1470.273a
0.1850.235b

Ni

Estimated mean
Theoretical value

1.44
1.57

0.09
0.08

6.2
3.4

8.3

0.862.28a
1.411.73b

Cu

Estimated mean
Theoretical value

6.98
7.77

0.12
0.27

1.7
2.3

9.2

5.4410.10a
7.248.30b

Zn

Estimated mean
Theoretical value

2.8
1.1

8.5
2.1

9.0

23.543.6a
31.435.6b

As

Estimated mean
Theoretical value

0.01
0.01

6.0
4.0

12.8

0.1050.277a
0.1680.214b

Sr

Estimated mean
Theoretical value

0.97
1.35

3.7
2.4

10.0

26.348.9a
34.940.3b

Mo

Estimated mean
Theoretical value

0.42
0.52

0.016

3.9

6.8

0.360.68a

Cd

Estimated mean
Theoretical value

0.207
0.199

0.01
0.01

4.0
2.7

5.6

0.1400.260a
0.1840.214b

Sb

Estimated mean
Theoretical value

0.058
0.066

0.007
0.005

5.7
5.1

10.5

0.0360.096a
0.0560.075b

Ba

Estimated mean
Theoretical value

0.8
1.3

2.2
2.6

8.9

22.842.3a
30.035.0b

Hg

Estimated mean
Theoretical value

0.002
0.001

6.9
3.0

10.4

0.0130.023a
0.0160.019b

4.2
3.6

7.5

1.512.81a
1.932.39b

0.851
0.952
189
191

33.3
33.5
0.204
0.191
26.1
37.6

34.2
32.5
0.019
0.018

Pb
a
b

Estimated mean
2.02
0.08
Theoretical value
2.16
0.12
h
i
CVR M
p with k = 3, n = 1 samples and M = reference value or indicative value for Mo.
: CI M  k  100
n

CI (IQC) (mg kg1)

369972a
659681b

Condence interval of the CRM.

Table 8
Results of trueness and intermediate precision reproducibility in BCR 278R.
CRM

Mean (mg kg1)

s (mg kg1)

CVr (%)

CVR (%)

CI (IQC) (mg kg1)

Mn

Estimated mean
Theoretical value

7.04
7.69

0.33
0.12

4.7
1.0

7.0

5.3810.0a
7.467.92b

Cu

Estimated mean
Theoretical value

8.33
9.45

0.34
0.06

4.1
0.5

7.7

6.6212.3a
9.329.58b

Zn

Estimated mean
Theoretical value

1.7
0.9

4.9
0.9

9.0

58.2108a
81.484.8b

As

Estimated mean
Theoretical value

5.49
6.07

0.12
0.07

2.1
0.7

11.1

3.348.80a
5.946.20b

Cd

Estimated mean
Theoretical value

0.324
0.348

0.009
0.003

2.9
0.7

6.3

0.2440.452a
0.3410.355b

Hg

Estimated mean
Theoretical value

0.183
0.196

0.005
0.005

2.6
1.5

5.3

0.1370.255a
0.1870.205b

0.03
0.02

1.7
0.7

8.2

1.402.60a
1.962.04b

Pb
a
b

71.5
83.1

Estimated mean
1.85
Theoretical value
2.00
h
i
CVR M
p

: CI M  k  100 n with k = 3, n = 1 samples and M = reference value.

Condence interval of the CRM.

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

118

Table 9
Results of trueness for Ga, Ge, Ag and Te spiked with 5 and 20 mg L1 on veal, carrot
or corn.
Mean (mg L1)

CI (mg L1)a

s (mg L1)

CVr

Theoretical value

5.0 and 20.0

0.50 and 2.0

Ga

4.8
20.7

4.35.7
17.322.7

0.05
0.5

1.0
2.4

Ge

4.9
20.5

4.35.7
17.322.7

0.02
0.3

0.4
1.5

Ag

4.9
16.3

4.15.9
16.024.0

0.02
0.5

0.5
3.0

Te

4.9
4.35.7
0.09
1.9
22.3
17.322.7
0.5
2.2
h
i
CVR M
a
p
: CI M  k  100
with
k
=
3,
n
=
5
samples,
M
=
theoretical
value
and
n
CVR = 10% for Ga, Ge, Te and 15% for Ag.

Sr, Cd, Sb and Pb) and three in INCT MPH 2 (Al, Cu and Sr). However,
each individual results as well as the measured means were well
included in the CI (n = 1; k = 3; p = 99%) around the reference value,
except for Ni in IAEA 407 and Sr in INCT MPH 2 (slightly below IC).
It should be noted that Ni was overestimated (0.9 mg kg1 instead
of 0.6 mg kg1) due to polyatomic interferences (44Ca16O or
23
Na37Cl) caused by the high concentration of Na and Ca in the
CRM (Nardi et al., 2009). According to Tables 68, the CVr for the 17
elements was 9% and was less than the CVR dened for the
reproducibility (dened in the following paragraph). So, the
trueness was generally satisfactory and a good accordance was
obtained between found and expected values.
For Ga, Ge, Ag and Te without certied or indicative values, the
analysis was carried out with spiked standard solutions (5 and
20 mg L1) on two unknown samples (carrot, veal or corn).
Obtained values were close to the theoretical concentration of
spiked standard solutions on unknown samples and CVr was below
5% for these four elements (Table 9). So, the trueness was
considered acceptable. It should be noted that for Ag, mean
concentrations (5 and 20 mg L1) were underevaluated in comparison with the theoretical value. The major cause was postulated to
be due to chloride in the sample that caused precipitation of silver
chloride (Yang et al., 2002). Similar problems were also reported in
aqueous samples by US EPA which recommended that samples
were digested using mixed acid HNO3 + HCl (2 + 1) in aqueous
samples containing concentrations up to 0.1 mg L1 of Ag (US EPA,
1994). Indeed, Ag is only slightly soluble in the presence of chloride
unless there is a sufcient chloride concentration to form the
soluble chloride complex. However, as compromise, the multielementary analysis requires the use of the unique HNO3 media
and the obtained mean concentrations remained with the CI set for
the study.
3.6. Intermediate precision reproducibility
Based on the NF V03-110 standard (AFNOR, 1998), the
intermediate precision reproducibility was investigated by repeating the analysis (n > 10) of a same homogeneous sample on
different days and several months, with four different analysts and
with two different microwave systems. Intermediate reproducibility variance was calculated as: s2R s2L s2r , where s2L was the
variance that measured sample variations and s2r was the
repeatability variance. This reproducibility was realized with the
same CRMs than it was done previously. For each element, as
expected, CVR > CVr, obtained previously (Tables 68). Calculated
CVR were in the range of 5.314.9%, according to elements and
CRMs. For practical reasons in the context of routine multielemental analysis, two groups of elements have been established
according to the order of magnitude of the intermediate precision:

The rst group is constituted by V, Mn, Co, Cu, Zn, Sr, Mo, Cd, Ba,
Pb and Hg that usually had a CVR lower than or equal to 10%. It
should be noted that this estimation was similar to those
previously reported for Hg in BCR 278R (5.3%) and IAEA 407
(6.2%) and Pb in BCR 278R (8.2%) and INCT MPH 2 (7.5%) (Noel et al.,
2005). However, as expected, the CVR of Hg in INCT MPH 2 was
higher due to its low level near the LOQ (0.01 mg kg1) and a
slightly higher CVR is also observed for Pb in IAEA 407.
A CVR of 15% was set for the second group, which considered the
other elements Li, Al, As, Ni, Sn and Sb, usually exceeding a CVR of
10% but never exceeded 15% in the CRMs tested. IQC put in place
during the TDS, notably through control charts, corroborated these
results and gave a higher percentage of unacceptably found values
in comparison with the reference value for these elements (Millour
et al., 2010).
For Ga, Ge, Ag and Te without certied or indicative values, a
CVR was estimated considering the results obtained for the
trueness and the IQC of the second French TDS (Millour et al.,
2010). Thus, a CVR of 10% will be set for Ga, Ge and Te (less than 8%
of non-conforming spiking) and it will be necessary to apply a less
restrictive CVR of 15% for Ag.
The Z-scores were used as IQC to check these three CRMs, as
previously described (Millour et al., 2010). At the end of the study,
94% of results (2 < Z < 2) were considered acceptable, 2% were
questionable (2 < jZj < 3) and 4% unacceptable (jZj > 3).
To summarize the intermediate precision reproducibility, two
groups of elements have been dened according to the:
A CVR of 10% for V, Mn, Co, Cu, Zn, Ga, Ge, Sr, Mo, Cd, Te, Ba, Pb
and Hg
A CVR of 15% for Li, Al, As, Ni, Ag, Sn and Sb
Based on the set CVR, the uncertainty of each result for all the
elements was dened as:

k  CVR  M
p
100  n

with k = 2 (p = 95%), M the mean value, and n the number of

independent replicates (n = 2).
Thus, the uncertainties for duplicates were dened according to
the two groups of CVR: U = 0.141  M and U = 0.212  M for a CVR
of 10 and 15%, respectively.
3.7. External quality controls (EQC)
To ensure and conrm the analytical accuracy of the multielemental method, the NRL regularly participated in prociency
test schemes (PT-Schemes) as EQC, such as the Central Science
Laboratory-Food Analysis Performance Assessment Scheme (CSLFAPAS) for multi-elemental and multi-matrices analyses (fruit
juice, canned sh, wine, etc.), and the Community Reference
Laboratory for Heavy Metals (Istituto Superiore di Sanita` ISS). Zscore, which expresses the difference between the laboratory
mean value and the target value, was used to evaluate results.
Performance was considered satisfactory if Z-score was between
2 and +2; when the absolute value of Z-score was between 2 and
3, results were questionable. When the absolute value of Z-score
was superior or equal to 3, results were unacceptable. It was not
possible to participate in interlaboratory comparison with all
elements of the method but nine elements (As, Al, Co, Cu, Mo, Cd,
Sn, Hg, and Pb) have been monitored in several matrices. In 1998
2008, for Pb, Cd and Hg, 95.5% of results for Z-score were acceptable
for FAPAS (n = 46) and 97.8% for ISS (n = 44) (Noel et al., 2009).
During the period 20072009, for six elements (As, Al, Co, Cu, Mo,

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

119

Table 10
Results of EQC with Z-score during the period 20072009 for six elements.
Analyte

Sample (year)

As

Milk powder (2007)

Milk powder (2007)
Canned crab meat (2007)
Milk powder (2008)
Milk powder (2008)
Milk powder (2008)
Canned sh (2008)
Meat (2008)
Canned crab meat (2009)
Milk powder (2009)
Canned sh (2009)

Al

Assigned value (mg kg1)

Observed value (mg kg1)

Z-scorea

0.051
0.104
11.4
0.125
0.189
7.6
0.343
0.770
11.2
0.732
0.654

0.053
0.126
11.6
0.143
0.209
9.1
0.383
0.849
9.94
0.844
0.755

0.2
1.0
0.2
0.7
0.5
1.6
0.6
0.6
1.0
0.9
0.9

Infant formula cereal (2008)

Infant formula cereal (2009)
Milk powder (2009)

0.992
0.972
1.84

0.977
1.72
1.69

0.1
4.8
0.5

Co

Canned sh (2008)

0.007

0.007

0.2

Cu

Canned crab meat (2007)

Canned crab meat (2009)

Mo

Infant formula cereal (2008)

Infant formula cereal (2009)

Sn

Vegetable puree (2008)

Fruit juice (2009)
Tomato paste (2009)

: Z-score

X f X c
p , where Xf is the actual analytical result, Xc is the
n

sR =

13.7
12.1

15.1
12.4

0.382
0.425

0.389
0.456

214
91.5
255

162
80.8
233

0.9
0.2
0.1
0.4
3.4
1.4
1.2

certied value, sR the reproducibility standard deviation, n the number of independent replicates (n = 1).

and Sn) results were considered acceptable (Table 10) except for Sn
in vegetable puree (Z-score = 3.4) and for Al in infant formula
cereal (Z-score = 4.8). The unacceptable results for Al were due to
problem of random pollution. So, performance in these prociency
tests were regularly considered satisfactory and showed the ability
of the NRL to provide reliable and accurate results.
4. Conclusion and prospects
After this validation, the selected criteria of performance
(linearity, LOQ, specicity, trueness, repeatability, intermediate
precision reproducibility) demonstrated that the use of both
microwave digestion in closed-vessel for sample preparation and
the ICP-MS for detection permitted an accurate determination of
21 elements in several types of food samples (matrices of animal
and vegetal origin). This validation was supported by appropriate
IQC to achieve a satisfactory quality level in the case of routine
analysis. Furthermore, the dened CVR will be used to delimit the
CI for CRMs in future studies. Obviously, it should be noted that this
multi-elemental method is an acceptable compromise and some
elements such as Ag and Sn could be more accurately quantied by
specic method, notably using a mixture of HNO3 and HCl to
stabilize these elements.
This validation will be followed to the accreditation by the
French committee of accreditation (COFRAC) in order to obtain an
extension of the in-house method concerning the analysis of four
elements in matrices of animal origin by ICP-MS. Finally, as NRL,
the laboratory will transfer its method to veterinary departmental
laboratories participating in the various national plans of food
monitoring for health risk assessment.
Acknowledgements
The participants are warmly thanked for their willing cooperation. Thanks are also due to J.-C. Leblanc, AFSSA DERNS,
coordinator of the 2nd French TDS for his helpfulness. The authors
would like to thank the general food directorate of the Ministry of

Food, Agriculture and Fisheries, and the general health directorate

of the Ministry of Health and Sports for their nancial contribution.
References
AFNOR, 1996. Analyse des produits agricoles et alimentaires- Guide pour lutilisation des materiaux de reference. AFNOR NF V03-115. Saint Denis, France:
Association Francaise de Normalisation.
AFNOR, 1998. Analyse des produits agricoles et alimentaires Procedure de
validation intra-laboratoire dune methode alternative par rapport a` une methode de reference Cas de methodes danalyse quantitatives. AFNOR NF V03110; Association Francaise de Normalisation, Saint Denis, France.
AFNOR, 2002a. Produits Alimentaires Dosage des Elements Traces Crite`res de
performance, generalites et preparation dechantillons. AFNOR NF EN 13804.
Association Francaise de Normalisation, Saint Denis, France.
AFNOR, 2002b. Produits alimentaires Determination des elements traces
Digestion Sous Pression. AFNOR NF EN 13805. Association Francaise de Normalisation, Saint Denis, France.
AFNOR, 2003. Protocole devaluation dune methode alternative danalyse physicochimique quantitative par rapport a` une methode de reference. AFNOR XP T 90210. Association Francaise de Normalisation, Saint Denis, France.
EC, 2007. Commission Regulation (EC) No 333/2007 of 28 March 2007 laying down
the methods of sampling and analysis for the ofcial control of the levels of lead,
cadmium, mercury, inorganic tin, 3 MCPD and benzo(a)pyrene in foodstuffs, Oj
No. L 88, 29.03.07, 0029-0038.
Feinberg, M., De la Rochette, A., 1997. Evaluation of non linearity testing procedures
on simulated data. Journal of AOAC International 80, 7987.
ISO, 1994. Applications of statistics Accuracy (trueness and precision) of measurements methods and results. ISO 5725. International Organization for
Standardisation, Geneva, Switzerland.
Jorhem, L., 2004. Certied reference materials as a quality tool in food control: much
used-often misused-sometimes abused. Accreditation and Quality Assurance 9,
305310.
Leblanc, J.C., Guerin, T., Noel, L., Calamassi-Tran, G., Volatier, J.L., Verger, P., 2005.
Dietary exposure estimates of 18 elements from the 1st French Total Diet Study.
Food Additives and Contaminants 22, 624641.
Lo, J.M., Wai, C.M., 1975. Mercury loss from water during storage: mechanisms and
prevention. Analytical Chemistry 47, 18691870.
Millour, S., Noel, L., Chekri, R., Vastel, C., Kadar, A., Guerin, T., 2010. Internal quality
controls applied in inductively coupled plasma mass spectrometry multielemental analysis for the second French Total Diet Study. Accreditation and
Quality Assurance 15 (9), 503513.
Nardi, E.P., Evangelista, F.S., Tormen, L., SaintPierre, T.D., Curtius, A.J., De Souza, S.S.,
Barbosa Jr., F., 2009. The use of inductively coupled plasma mass spectrometry
(ICP-MS) for the determination of toxic and essential elements in different types
of food samples. Food Chemistry 112, 727732.
Noel, L., Dufailly, V., Lemahieu, N., Vastel, C., Guerin, T., 2005. Simultaneous analysis
of cadmium, lead, mercury and arsenic content in foodstuffs of animal origin by

120

S. Millour et al. / Journal of Food Composition and Analysis 24 (2011) 111120

inductively coupled plasma/mass spectrometry after closed vessel microwave

digestion: method validation. Journal of AOAC International 88, 18111821.
Noel, L., Leblanc, J.C., Guerin, T., 2003. Determination of several elements in
duplicate meals from catering establishment using closed vessels microwave
digestion with inductively coupled plasma mass spectrometry detection: estimation of daily dietary intake. Food Additives and Contaminants 20, 4456.
Noel, L., Vastel, C., Chekri, R., Chafey, C., Testu, C., Guerin, T., 2009. Evaluation of 10years NRL prociency tests for lead, cadmium and mercury analysis in foodstuff
of animal origin. Microchemical Journal 92, 7379.
Pizzolon, J.C., Hoenig, M., 2005. Analyse des Echantillons Alimentaires par ICP MS.
Developpement et Routine: ou` sont les differences? Canadian Journal of
Analytical Sciences and Spectroscopy 50, 221227.
Sirot, V., Volatier, J.L., Calamassi-Tran, G., Dubuisson, C., Menard, C., Dufour, A.,
Leblanc, J.C., 2009. Core food of the French food supply: 2nd Total Diet Study.
Food Additives and Contaminants 26, 623639.
Thompson, M., Ellison, S.L.R., Wood, R., 2002. Harmonized guidelines for singlelaboratory validation of methods of analysis. Pure and Applied Chemistry 74,
835855.

US EPA, 1994. Determination of Metals and Trace Elements in Water and Wastes by
Inductively Coupled Plasma-Atomic Emission Spectrometry, US EPA Method
200.7, Supplement I, Revision 4.4.
Varian UltraMass Application Report, 1996. Determination of Mercury by ICP-MS.
World Health Organisation GEMS/Food Total Diet Studies, 2006. Report of a Joint
USFDA/WHO. In 4th International Workshop on Total Diet Studies, Beijing,
China. Retrieved December 12, 2008 from: http://www.who.int/foodsafety/
publications/chem/en/index.html.
World Health Organisation, 1995. Evaluation of Certain Food Additives and Contaminants. 44th Report of the Joint FAO/WHO Expert Committee on Food
Additives, WHO Technical report series 859, Geneva, Switzerland. Retrieved
December 12, 2008 from: http://www.who.int/ipcs/publications/jecfa/en/.
World Health Organisation. Retrieved March 29, 2010 from: http://www.who.int/
ipcs/food/jecfa/glossary/en/.
Yang, X.J., Foley, R., Low, G.K.C., 2002. A modied digestion procedure for analysing
silver in environmental water samples. The Analyst 127, 315318.
Zukowska, J., Biziuk, M., 2008. Methodological evaluation of method for dietary
heavy metal intake. Journal of Food Science 73, 2127.