Journal of Chromatography A, 1132 (2006) 211–218

Collaborative trial validation study of two methods, one based on high

performance liquid chromatography–tandem mass spectrometry and on
gas chromatography–mass spectrometry for the determination of
acrylamide in bakery and potato products
Thomas Wenzl a,∗ , Lubomir Karasek a , Johan Rosen b , Karl-Erik Hellenaes b ,
Colin Crews c , Laurence Castle c , Elke Anklam a
a European Commission – Joint Research Centre – Institute for Reference Materials and Measurements, Retieseweg 111, 2400 Geel, Belgium
b Swedish National Food Administration, Chemistry Division 1, Box 633, 75126 Uppsala, Sweden
c Central Science Laboratory, Sand Hutton, Y041 ILZ York, United Kingdom

Received 22 May 2006; received in revised form 3 July 2006; accepted 6 July 2006
Available online 8 August 2006

Abstract
A European inter-laboratory study was conducted to validate two analytical procedures for the determination of acrylamide in bakery ware
(crispbreads, biscuits) and potato products (chips), within a concentration range from about 20 ␮g/kg to about 9000 ␮g/kg. The methods are based
on gas chromatography–mass spectrometry (GC–MS) of the derivatised analyte and on high performance liquid chromatography–tandem mass
spectrometry (HPLC–MS/MS) of native acrylamide. Isotope dilution with isotopically labelled acrylamide was an integral part of both methods.
The study was evaluated according to internationally accepted guidelines. The performance of the HPLC–MS/MS method was found to be superior
to that of the GC–MS method and to be fit-for-the-purpose.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Acrylamide; Collaborative trial study; Food products; Gas chromatography–mass spectrometry (GC–MS); High performance liquid
chromatography–tandem mass spectrometry (HPLC–MS/MS); Method validation

1. Introduction by, e.g. setting a maximum allowed concentration in the various

Since the discovery of the occurrence of the Maillard reaction
product acrylamide in food in 2002, many studies have been car-
ried out in order to explore the various pathways of formation,
to find appropriate ways for reduction and to improve the ana-
lytical techniques for determination. A huge number of papers
have been published to-date of which more than 60 are on ana-
lytical methods [1,2]. Still today, European and international
workshops are carried out in order to discuss the improvements
and state-of-the art. Proficiency tests (PTs) have been and are
still offered in order to assess the capability of analytical lab-
oratories. As acrylamide is formed by heat processing of food,
it is difficult to treat it as a contaminant that can be regulated
∗ Corresponding author. Tel.: +32 14 571 320; fax: +32 14 571 783.
E-mail address: thomas.wenzl@ec.europa.eu (T. Wenzl).
food products, especially as much acrylamide is formed during
household cooking. For this reason, method criteria were not set
from a regulatory or standardization point of view. Nevertheless,
it was felt, that validated methods would be an asset. The method
performance criteria obtained from collaborative trial method
validation can serve for further improvement of methods and to
be taken as guidance. It is of utmost importance, that sound data
are produced in order to make comparisons of values found in
food products, e.g. to assess the effectiveness of reduction mea-
sures [3] or to ensure the quality of entries in data bases [4,5].
Validated methods, proficiency tests and reference materials are
excellent quality tools to ensure comparability of data produced
in different laboratories. Proficiency tests have shown that two
main approaches are used for acrylamide: high performance liq-
uid chromatography (HPLC) hyphenated to triple quadrupole
mass spectrometry (MS/MS) and gas chromatography (GC) of
the mono- or dibromo derivative of acrylamide with electron

0021-9673/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2006.07.007
212 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218

ionization and mass spectrometric detection (EI–MS) of the tocol had to be reported to the organisers of the collaborative
fragment ions [6,7]. A third procedure, involving GC–MS analy- trial.
sis of underivatised acrylamide, is practiced but less commonly.
Labour-intensive analysis protocols are required in order to sep-
2.2. Test samples
arate the small (molecular mass = 71 u) and polar acrylamide
from the very complex matrices that are frequently provided
Samples were either purchased in local markets or produced
by food. Single stage HPLC/MS analysis and analysis of the
on a pilot plant of the German Federal Research Centre for Nutri-
native acrylamide by GC–MS (EI) can suffer from interferences
tion and Food (Detmold, Germany). They prepared potato crisps
from some food matrices. Even when applying the more specific
that would not be found in retail stores, by applying special
HPLC–MS/MS in selected reaction monitoring (SRM) mode, it
process conditions. Food matrices covered in this study were
is necessary to focus on sample preparation. Ion suppression
butter biscuits, toasted bread, spiced biscuits, potato crisps and
effects that most likely happen in the ion source (electrospray
mashed potato powder. For a proper assessment of the trueness of
ionization is mainly applied for acrylamide analysis) can com-
the results, a candidate reference material (ERM® -BD272 from
promise quantitation. For that reason a two stage solid phase
the Federal Institute for Materials Research and Testing, Berlin,
extraction clean-up is used frequently to both clean-up and
Germany) with a preliminary certified value was included in
enrich acrylamide.
the study. All samples were ground with suitable mills to have
Different problems are encountered in the GC–MS anal-
particle sizes below 500 ␮m. Special attention was given to the
ysis of acrylamide of which one is the limited solubility of
temperature of the samples during grinding. In order to avoid
acrylamide in organic solvents that impedes acrylamide extrac-
the introduction of heterogeneity in the samples by the grinding
tion from food. The derivatisation of acrylamide to dibromo-
process, e.g. due to fatty particles that stick together, all sam-
propanamide followed by extraction from the aqueous phase
ples were frozen in liquid nitrogen prior to processing and kept
into ethyl acetate, allows analyte enrichment by concentration
at temperatures of at least −20 ◦ C throughout. The potato chips
of the extract, reduces the polarity of the molecule for better GC
samples were ground at temperatures below −110 ◦ C. The white
performance, and improves detectability due to the shift of the
bread crumb sample that was used as a blank sample was pre-
molecular mass to a higher level and thereby providing a mass
pared by drying white bread crumb at room temperature prior
spectrum with more specific mass/charge ratios (m/z) compared
to grinding. Out of 11 test samples in total, nine contained acry-
to native acrylamide.
lamide below 1000 ␮g/kg. The other three samples were chosen
The study described in this paper was performed in close
to cover the range of up to about 9000 ␮g/kg. Since acrylamide
collaboration of the European Commission’s Directorate Gen-
concentrations up to a few thousand ␮g/kg are sometimes found
eral Joint Research Centre (EC DG-JRC), the Swedish National
for potato crisps, it was decided to cover this range of very
Food Administration (SNFA), the Central Science Laboratory
high acrylamide concentrations as well [4,5]. An overview of
(CSL) and the Nordic Committee on Food Analysis (NMKL).
the composition of the set of test samples together with their
The CSL and SNFA developed and optimised the two meth-
assigned acrylamide content is given in Table 1. A quantity of
ods under investigation, whereas EC DG-JRC was responsible
about 3 g of each test sample was filled in amber glass bottles
for the preparation of test materials, homogeneity, and stability
with PTFE lined screw caps and coded by a number between 1
testing, dispatch of samples and evaluation of results. NMKL
and 2000. Two glass bottles per sample were packed into insu-
checked the design of the study. The test materials covered a
lated boxes and sent by express mail to the participants under
range of concentration of acrylamide in food of about 20 ␮g/kg
dry ice cooled conditions. Most participants reported that the
to about 9000 ␮g/kg.
samples arrived frozen.
The method validation using the collaborative trial approach
was performed and evaluated according to the international
accepted guidelines [8,9]. Table 1
Test samples included in study and homogeneity results
2. Validation study Origin Acrylamide
content (␮g/kg)
2.1. Design of the study Butter biscuits (A) From retail store 16
Toasted bread From retail store 38
Forty-six Laboratories from 13 EU Member States and Butter biscuits (B) From retail store 96
five other countries from all over the world with experience Speculoos (spiced biscuits) From retail store 249
in acrylamide analysis were contacted to participate in the Potato crisps (A) Specially prepared 324
Spiked mashed potato powder From retail store 500
study. Of these, 32 expressed their interest in participation Commercial potato crisps (A) From retail store 628
in the study and were therefore supplied with test materi- Crisp bread Candidate reference 980
als. Finally, 25 laboratories submitted results. The study was material:
designed as a blind duplicate study. The participants were pro- ERM® -BD272
vided with the method protocols, instruction guidelines, and Potato crisps (B) Specially prepared 2512
Commercial potato crisps (B) From retail store 4051
the test samples. They were requested to follow the method Potato crisps (C) Specially prepared 9082
protocols as much as possible. Any deviation from the pro-
 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218
2.3. Homogeneity
Homogenisation of test samples was performed by using dif-
ferent means. Granular samples were homogenised in a concrete
mixer, whereas samples with higher or very high fat content
were homogenised by repeated grinding steps. Homogeneity of
the samples was assessed by internationally agreed procedures
[10]. From each test sample, 10 randomly selected sub sam-
ples were taken and analysed in duplicate using the described
HPLC–MS/MS method. Sufficient homogeneity was stated if
the analysis of variances (ANOVA) did not show significant
differences between the results of the replicate analysis of the
samples in a particular sample container and the results obtained
for different sample containers. In addition the variability of the
results resulting as a consequence of sample in-homogeneity
had to be below 30% of the standard deviation that would be
provided by the Horwitz standard equation [11].
3. Experimental
3.1. Protocol for HPLC–MS/MS method [12]
The HPLC–MS/MS analysis procedure consists of the aque-
ous extraction of 2.0 g test sample with 40 mL of water. A
quantity of 400 ng isotopically labelled acrylamide (deuterium
content >98%) is added as internal standard to the extractant. A
10 mL aliquot of the aqueous phase is passed through a precon-
ditioned Isolute Multimode® solid phase extraction cartridge
(6 mL, 500 mg, International Sorbent Technology, Hengoed,
Mid Glamorgan, UK). The eluate is collected and loaded on a
Isolute ENV+® solid phase extraction cartridge (6 mL, 500 mg,
International Sorbent Technology). The cartridge is rinsed with
4 mL of water followed by the elution of acrylamide with 2 mL
of 60% methanol in water. The elution solvent volume is reduced
by, e.g. a stream of nitrogen to a final volume of about 500 ␮L.
A quantity of 10 ␮L is injected on a Hypercarb column (e.g.
50 mm × 2.1 mm, Thermo Electron Corporation, Waltham, MA,
USA), run isocratically with 0.1% acetic acid in water at a flow
rate of 400 ␮L min−1 . For acrylamide the transitions m/z 72 > 55,
72 > 54 and 72 > 44 are recorded, whereas the transition 75 > 58
is used for the isotopically labelled internal standard. Details of
the analytical procedure and the single laboratory validation of
the method are published elsewhere [12].
3.2. Protocol for GC–MS method [13,14]
An amount of 400 ng of isotopic labelled acrylamide dis-
solved in water is added to 2.0 g of test sample followed by a
mixture of ethyl acetate and cyclohexane (3 mL, 1:1, v/v). The
mixture is shaken to extract the fat from the sample and then
cold water (20 mL) is added to the mixture which is placed on
a mechanical shaker for a further 15 minutes. After centrifuga-
tion (3000 × g, 5 min), a 10 mL aliquot of the aqueous phase
is transferred to a glass vial and derivatised by adding 15 mL
of a mixture of potassium bromide, hydrobromic acid and sat-
urated bromine water. Bromination takes place in the dark at
4 ◦ C, within 1 h. Excessive derivatisation reagent is destroyed
213
with a few drops of an aqueous sodium thiosulphate solution.
The resulting 2,3-dibromopropanamide is extracted with 8 mL
of ethyl acetate. Half of that volume is pipetted into a clean
glass vial, dried over anhydrous, granular sodium sulphate and
after transfer of the dried extract into another clean vial reduced
with, e.g. a stream of nitrogen to a volume below 0.5 mL. To
avoid uncontrolled dehydro-bromination in the hot injector of
the GC, 50 ␮L of triethylamine are added in order to deliberately
dehydro-brominate the analyte. Finally 2-bromopropenamide
is analysed by GC–MS (EI) on a mid-polar GC column, e.g.
containing 50% phenyl–50% polymethylsilicone as stationary
phase. The mass spectrometer is run in selected ion monitoring
mode. The monitored mass charge ratios are 106, 133, 135, 149
and 151 for 2-bromopropeneamide; and 152 and 154 for the 13 C
labelled internal standard.
Although d3 -acrylamide can be applied as internal standard,
preference should be given to the 13 C labelled analogue, since
the deuterated compound would loose one deuterium atom dur-
ing dehydro-bromination, which would result in an increased
overlap of the mass spectra of the analyte and the internal stan-
dard. Thus, the before mentioned mass to charge ratios are not
suitable, if deuterated acrylamide is applied.
4. Results and discussion
4.1. Results obtained by the HPLC–MS/MS
Sixteen laboratories reported results. At first the samples
were decoded and the data were compiled. Subsequently the
data were investigated for outliers. For that purpose different
statistical means were applied. “Mandel’s h” and “Mandel’s k”
statistics provide a graphical representation of the laboratories
consistency [9]. The Mandel’s h statistic, shown in Fig. 1, is a
measure of how much the mean values reported by each labo-
ratory for each sample differs from the mean value amongst all
participants. The data for the test samples are sorted for each
participant from the left to the right in the order of ascending
Fig. 1. Mandel’s h parameters for the whole set of samples that was considered
in the evaluation of the study; critical limits are indicated by dashed lines.
214 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218

Table 2
Results reported by the participants applying HPLC–MS/MS

Grey cells mark outliers; <LOD: below limit of detection, <LOQ: below limit of quantitation, (*) participant reported instrument problems when analysis was
performed.

acrylamide content. Fig. 2 shows the k statistics (sorting identical
with Fig. 1). It compares the variability of the results reported by
each laboratory for each sample with the sum of variances for all
laboratories. The critical values for the 5% significance level are
marked by horizontal lines. If a value lies beyond these limits,
it should be further investigated as it could represent an outlier.
As can be seen the participants 4 and 16 reported for a number
of samples results that were suspected as being outliers. Partici-
pant 16 stated problems with the HPLC–MS/MS instrument on
one particular day, when some of the samples were analysed.
This could be the reason for the large differences between the
replicate analyses, as can be seen in Table 2. In general, from the
distribution of Mandel’s h parameter it can be concluded that the
deviations from the assigned values are rather systematic than
random. This is reflected in the either one sided positive or one
sided negative deviation from the assigned value of the whole
set of results reported by a particular participant. In addition to
the determination of Mandel’s h and k parameters, Cochran’s
and Grubb’s tests were conducted to identify numerical out-
liers [8,9]. Outliers identified by one or the other determination
method (Mandel’s parameters, Cochran’s and Grubb’s tests) are
marked in Table 2 by grey cells.
 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218 215

Table 2 shows a compilation of the reported HPLC–MS/MS

Fig. 2. Mandel’s k parameters for the whole set of samples considered in the
evaluation of the study; the critical limit is indicated by a dashed line.
For the candidate reference material that was included into the
study, outliers were identified by estimating the combined mea-
surement uncertainty provided by the material and the method
and comparing it with the difference of the reported value and
the certified value. The average repeatability standard deviation
was applied as a measure of precision, since measurement uncer-
tainty data of the individual participants were not available. An
outlier was identified at the 95% confidence level, if Eq. (1)
was fulfilled. Results that were identified by this method being
outliers are marked in Table 2 by grey cells.
 arithmetic means obtained for the spiked mashed potato pow-
|x̄i − XCRM | > 2 SD2r + u2CRM (1)
x̄i is the average value of laboratory i; XCRM the certified value;
SDr the repeatability standard deviation; and uCRM is the stan-
dard uncertainty of the candidate reference material.
The calculated arithmetic mean value of each sample was
compared with estimates obtained by applying robust statistics,
respectively with the major mode of Kernel density plots [15].
The different estimates were in good agreement for all samples.
data and indicates data that were identified as outliers. After
removal of the latter, repeatability and reproducibility standard
deviations were calculated. In order to estimate the quality of the
precision parameters, they were compared with the respective
precision estimates derived from the Horwitz equation. The cal-
culated precision parameters are listed for each sample in Table 3
together with the Horrat values [16]. The precision parameters
are expressed as relative standard deviations, in order to make
comparability of the results easier. As can be seen these param-
eters were not calculated for the butter biscuits sample with low
acrylamide content. The reason was that for 5 of 32 analyses the
acrylamide content was reported as below limit of quantitation
(LOQ). However, this is a good indication of what the LOQ of
the method is.
The Horrat values were in general below one, indicating
that the results were more precise than it would be allowed
by the Horwitz equation. In regarding the two different food
categories separately, it becomes evident that the Horrat ratios
for repeatability (HoR ) were lower for the bakery ware sam-
ples (0.3–0.5) than for the potato products (0.5–0.8). This could
be due to the fact that the bakery samples are a simpler matrix
causing fewer problems in sample preparation than the potato
products samples. However, it should be emphasised that the
obtained reproducibility relative standard deviations (RSDR )
(5.4–13.2%) could be used to set benchmarks for the target stan-
dard deviations applied in future proficiency tests.
The trueness of the results was estimated by comparing the
der sample for the potato products and the candidate reference
material with the spiking level and the preliminary certified
acrylamide content for the bakery products, respectively. The
spiking level of the mashed potato powder was 500 ␮g/kg. The
assigned value for that sample varied slightly depending on the
method of its determination. It was found between 482 ␮g/kg
(Huber H15 estimate) and 485 ␮g/kg (arithmetic mean), giving
a relative bias of between 3.6 and 3.0%. Such slight devia-
tion was also found for the candidate reference material. For

Table 3
Method performance characteristics obtained in validation study (HPLC–MS/MS and GC–MS method)
HPLC–MS/MS GC–MS

Assigned acrylamide RSDr (%) RSDR (%) Hor HoR Overall mean value RSDr (%) RSDR (%) Hor HoR
content (␮g/kg) GC–MS (␮g/kg)

Butter biscuits (A) 16 – – – – – – – – –

Toasted bread 38 5.5 8.5 0.3 0.3 – – – – –
Butter biscuits (B) 96 7.8 11.8 0.5 0.5 82 18.2 30.8 1.2 1.3
Spiced biscuits 249 3.7 10.4 0.3 0.5 218 5.7 27.4 0.4 1.4
Potato crisps (A) 324 6.0 12.7 0.5 0.7 563 81.9* 81.9* 7.1* 4.7*
Spiked mashed potato powder 500 5.4 8.8 0.5 0.5 562 6.6 7.7 0.6 0.4
Commercial potato crisps (A) 628 8.9 13.2 0.8 0.8 768 27.4* 27.4* 2.5* 1.6*
Crisp bread (ERM® -BD272) 980 3.1 5.4 0.2 0.3 – – – – –
Potato crisps (B) 2512 5.9 11.7 0.6 0.8 – – – – –
Commercial potato crisps (B) 4051 4.3 8.9 0.5 0.7 – – – – –
Potato crisps (C) 9082 5.0 9.1 0.7 0.8 – – – – –

RSDr : repeatability relative standard deviation, RSDR : reproducibility relative standard deviation, Hor : Horrat ratio for repeatability, HoR : Horrat ratio for repro-
ducibility, (*) RSDr was found to be larger than RSDR —consequently RSDR was put at the level of RSDr .
216 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218
Table 4
Results reported by the participants applying GC–MS after derivatisation
Grey cells mark outliers; missing results were indicated by “x”; <LOD: below limit of detection, <LOQ: below limit of quantitation, nq: not quantifiable.
the latter, the mean value of the reported results was com-
puted to 958 ␮g/kg, which is negligible higher than the median
(952 ␮g/kg) or the Huber H15 estimate (950 ␮g/kg). However,
the expanded uncertainty (coverage factor of 2) of the candidate
reference material was given to be 90 ␮g/kg. Therefore, no sig-
nificant difference between the preliminarily certified value and
the assigned value was found at the 95% confidence level. Con-
sequently, the method is considered as unbiased. Summarising
the validation results for the HPLC–MS/MS method it can be
stated that the method performance characteristics are very good
and that the method is fit for the purpose to control acrylamide
in two matrix types investigated.
4.2. Results obtained by the GC–MS method
Only nine participants (12 invited) reported results back to
the organisers of the study. This number would just represent
one more than the minimum number of data required by the
international guidelines provided there are no outliers. Unfor-
tunately, the reported data sets were not complete. The results
are compiled in Table 4 and missing data are marked by an “x”.
As can be seen the variability of the data is quite high. This
is caused by either problems with the GC–MS instrumentation
used in the laboratories (participant 24) or by problems with
the method itself. In particular, participant 22 reported severe
difficulties in the analysis of 2-bromopropenamide. This labo-
ratory could not quantify samples if triethylamine was added
prior to injection (missing values). The values given by this
participant were generated after omitting the dehydrobromina-
tion step. Another participant did not report results, but stated
that the performance of their GC–MS instrument decreased
fast by applying the analysis procedure as required in this
study.
From in total 11 samples that were evaluated for the HPLC–
MS/MS method, complete GC–MS data sets were received just
for the spiked mashed potato powder sample and for the spiced
biscuit sample. For three other samples, eight data sets were
received and seven data sets or fewer were reported for the
remaining six samples. This indicates problems with the method
too. In order to compare the results obtained with this analysis
procedure to those obtained by the HPLC–MS/MS protocol,
samples with at least eight data sets were evaluated according
 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218
to the international guidelines. The results received from four
of nine participants were identified as outliers for at least one
sample. After exclusion of outliers, method performance char-
acteristics were calculated as listed in Table 3.
However, compared to the data gained by HPLC–MS/MS, the
spiked potato powder sample showed a higher arithmetic mean
of 562 ␮g/kg, instead of 485 ␮g/kg (nominal value 500 ␮g/kg).
This is in contrast to the spiced biscuits sample, for which
the mean value obtained by HPLC–MS/MS was 249 ␮g/kg
compared to 218 ␮g/kg that were calculated from the GC–MS
results. Much larger was the mismatch of the assigned values of
the HPLC–MS/MS study and the overall mean values obtained
by GC–MS for the two evaluated potato crisps samples. The
difference was 22 and 74%. With the exception of the spiked
potato powder sample, the Horrat ratios for repeatability (Hor )
and reproducibility were for all samples inferior in compari-
son to those gained with the HPLC–MS/MS method. For two
samples (both potato crisps samples) the calculated repeatability
standard deviation was larger than the reproducibility standard
deviation. Consequently the reproducibility standard deviation
was set to the level of the repeatability standard deviation. The
respective values in Table 3 are marked by an asterisk.
In conclusion, on the basis of the available GC–MS data and
considering the reported background information a reliable esti-
mation of the precision parameters is not possible. However, it
should be stressed, that GC–MS protocols with derivatisation
are very appropriately used in some specialised laboratories for
the analysis of acrylamide in food. This has been demonstrated
by previous proficiency test studies [6,17].
5. Conclusions
From the validation results obtained in the described collab-
orative trial it can be concluded that the HPLC–MS/MS method
studied is superior to the GC–MS method investigated. In addi-
tion, it was observed that the precision of the results obtained
by the HPLC–MS/MS method were for all samples even better
than preset by the Horwitz equation. Therefore, it can be con-
cluded that this method is undoubtedly fit for the purpose. The
experimentally determined precision parameters could now be
used as a benchmark for the target standard deviation applied in
future proficiency tests, since it has been demonstrated that such
a level of precision is achievable.
Acknowledgements
We thank the following participants in the collaborative trial
study:
Food and Drug Administration, Centre for Food Safety and
Applied Nutrition, College Park, MD, USA.
National Measurement Institute, Sydney, Australia.
Danish Institute for Food and Veterinary Research, Soborg,
Denmark.
Voedsel en Waren Autoriteit, Regio Zuid, Eindhoven, The
Netherlands.
217
Nestlé Research Center, Lausanne, Switzerland.
Laboratorio Centrale Sagit S.r.l., Cisterna di Latina, Italy.
Landesuntersuchungsanstalt für das Gesundheits- und Vet-
erinärwesen Sachsen, Dresden, Germany.
Scientific Institute of Public Health, Louis Pasteur, Department
of Pharmaco-Bromatology, Brussels, Belgium.
National Veterinary and Food Research Institute (EELA),
Department of Chemistry, Helsinki, Finland.
Danone Vitapole, Analytical Support, Palaiseau, France.
Laboratoire Départemental de la Sarthe, Le Mans, France.
National Food Administration Sweden, Uppsala, Sweden.
Staatliches Untersuchungsamt Hessen, Standort Wiesbaden,
Wiesbaden, Germany.
Lebensmittelchemisches Institut (LCI) des Bundesverbandes
der Deutschen Süßwarenindustrie e.V., Köln, Germany.
Landesamt für Landwirtschaft, Lebensmittelsicherheit und
Fischerei Mecklenburg-Vorpommern, Abteilung für Schad-
stoff und Rückstandsanalytik, Rostock, Germany.
Food Research Laboratory, 4/F Public Health Laboratory Cen-
tre, Hong Kong, China.
Central Science Laboratory, Food Safety and Quality Group,
York, United Kingdom.
TUBITAK MRC Food Institute, Kocaeli, Turkey.
RHM Technology Ltd., High Wycombe, United Kingdom.
Centro Técnico Nacional de Conservas Vegetales-Laboratorio
del Ebro, San Adrián, Navarra, Spain.
National Food Research Institute, Tsukuba, Japan.
Centro Nationale de Alimentación, Majadahonda-Madrid,
Spain.
Brewing Research International, Nutfield, United Kingdom.
Chemisches Untersuchungsamt Hagen, Hagen, Germany.
Federal Institute for Research and Testing (BAM), Berlin, Ger-
many.
General Chemical State Laboratory, Food Division & Division
of Environment, Athens, Greece.
We thank Dr. Haase from the Federal Research Centre for
Nutrition and Food, Detmold, Germany for the preparation of
potato chips test materials and the excellent support, as well as
NMKL for reviewing the study design.
References
[1] http://www.foodriskclearinghouse.umd.edu/acrylamide.cfm, accessed on
23.03.2006.
[2] L. Castle, S. Eriksson, J. AOAC Int. 88 (2005) 274.
[3] D. Taeymans, J. Wood, P. Ashby, I. Blank, A. Studer, R.H. Stadler, P. Gonde,
P. Van Eijck, S. Lalljie, H. Lingnert, M. Lindblom, R. Matissek, D. Muller,
D. Tallmadge, J. O’Brien, S. Thompson, D. Silvani, T. Whitmore, Crit.
Rev. Food Sci. Nutr. 44 (2004) 323.
[4] D. Lineback, T. Wenzl, O.-P. Ostermann, B. de la Calle, E. Anklam, D.
Taeymans, J. AOAC Int. 88 (2005) 246.
[5] http://www.irmm.jrc.be, accessed on 23.03.2006.
[6] H. Klaffke, T. Wenzl, E. Anklam, C. Fauhl, W. Mathar, R. Palavinskas, R.
Wittkowski, J. AOAC Int. 88 (2005) 292.
[7] L.M. Owen, L. Castle, J. Kelly, L. Wilson, A.S. Lloyd, J. AOAC Int. 88
(2005).
[8] W. Horwitz, Pure Appl. Chem. 67 (1995) 331.
[9] International Organization for Standardization, ISO 5725-2, 1994.
218 T. Wenzl et al. / J. Chromatogr. A 1132 (2006) 211–218

[10] M. Thompson, R. Wood, J. AOAC Int. 76 (1993) 926.
[11] W. Horwitz, L.R. Kamps, K.W. Boyer, J. Assoc. Off. Anal. Chem. 63 (1980)
1344.
[12] J. Rosen et al., submitted to Analyst, in press.
[13] J.S. Ahn, L. Castle, D.B. Clarke, A.S. Lloyd, M.R. Philo, D.R. Speck, Food
Addit. Contam. 19 (2002) 1116.
[14] S. Nemoto, S. Takatsuki, K. Sasaki, T. Maitani, J. Food Hyg. Soc. Jpn. 43
(2002) 371.
[15] P.J. Lowthian, M. Thompson, Analyst 127 (2002) 1359.
[16] W. Horwitz, R. Albert, J. AOAC Int. 79 (1996) 589.
[17] T. Wenzl, E. Anklam, J. AOAC Int. 88 (2005) 1413.