Journal of Chromatography A, 1102 (2006) 268–272

Development and validation of a solid phase micro-extraction–gas

chromatography–mass spectrometry method for the determination
of furan in baby-food
Federica Bianchi ∗ , Maria Careri, Alessandro Mangia, Marilena Musci
Dipartimento di Chimica Generale ed Inorganica, Chimica Analitica, Chimica Fisica, Università degli Studi di Parma,
Parco Area delle Scienze 17/A, 43100 Parma, Italy
Received 21 July 2005; received in revised form 17 October 2005; accepted 27 October 2005
Available online 17 November 2005

Abstract
An efficient and simple method for the determination of furan in baby-food (vegetables and fruits) by solid phase micro-extraction–gas
chromatography–mass spectrometry (SPME–GC–MS) was developed and validated. Experimental design was used to investigate the effects
of temperature and time of extraction. The calculated regression model was used to find the experimental conditions providing the optimal SPME
extraction yield. Validation was carried out in terms of limit of detection (LOD), limit of quantitation (LOQ), linearity, precision and trueness.
LOD and LOQ values in the low ng kg−1 were achieved, whereas linearity was established over two order of magnitude. Good precision was
obtained both in terms of intra-day repeatability and between-day precision on two concentration levels (RSD% lower than 3.6%). Recovery
values of 91.5 ± 6.2% and of 96.1 ± 1.3% (n = 3) were calculated at 75 ng kg−1 and 75 ␮g kg−1 level. Finally, the applicability of the method to the
determination of furan in a number of commercial and home-made baby-food samples was demonstrated.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Furan; Solid phase micro-extraction; Baby-food; Food safety; GC–MS; Method validation

1. Introduction
Food safety is regarded as a widespread priority task. In addi-
tion to the actions undertaken by food-related industries, the EU
legislative committees are reviewing current legislation, both
introducing new recommendations aimed to the establishment
of general principles issuing rules to increase quality standards
in food and fixing hazard limits for the EU citizen health [1].
Recently, great attention has been paid to the presence of both
natural and xenobiotic contaminants in raw and processed foods.
Among naturally-occurring substances, great concern is
addressed to the analysis of furan, this compound being classi-
fied by the International Agency for Research on Cancer (IARC)
in the group 2B “as possibly carcinogenic to humans” [2]. The
Food and Drug Administration (FDA) is requesting the submis-
sion of data and information regarding the occurrence of furan
in food, as well as sources of exposure, mechanisms of forma-
∗ Corresponding author. Tel.: +39 0521 905433; fax: +39 0521 905556.
E-mail address: federica.bianchi@unipr.it (F. Bianchi).
tion and toxicology [3]. In fact, although furan is a well known
heat treatment related by-product and has been detected in many
thermally treated foods [4], only few methods principally based
on static headspace techniques have been developed and applied
for furan determination in food [5–10].
SPME is an alternative solvent-free sampling technique
widely used for the analysis of volatile compounds, but nowa-
days only a SPME–GC–MS method with the use of a cryofo-
cussing unit was recently developed for furan determination in
heated foodstuffs [11].
The aim of this work was the development and validation of
a more simple method for the selective and sensitive analysis of
furan in baby-food, based on SPME–GC–MS.
An experimental design was used to study the effects of two
parameters on the SPME extraction, i.e. temperature of extrac-
tion and time of extraction. Data obtained allowed to calculate
the main and interaction effects of the factors under investiga-
tion, thus indicating the optimised extraction conditions.
The SPME–GC–MS method was then validated under opti-
mised conditions. According to the EURACHEM guidelines,
in-house validation was carried out in terms of detection

0021-9673/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2005.10.056
 F. Bianchi et al. / J. Chromatogr. A 1102 (2006) 268–272
limit (LOD), quantitation limit (LOQ), linearity, precision
and trueness. Excellent results in the low ng kg−1 level were
obtained for LOD and LOQ, thus proving the potential of the
method for the determination of furan at ultratrace levels in
baby-food.
The method developed was then applied to determine furan
in commercial and home-made baby-food samples.
2. Experimental
2.1. Chemicals
Furan (+99%, purity), furan-d4 (99.3%, purity) and
methanol (99.8% purity) were purchased from Sigma–Aldrich
(Milan, Italy).
Standard and working solutions, prepared from the
2500 mg l−1 methanolic stock solution by dilution in Ultra
Resi-Analysed purge and trap water (J.T. Baker, Deventer, The
Netherlands), were stored in completely filled vials closed with
hermetic seals at 4 ◦ C until analysis.
2.2. Experimental design
Fresh carrots (4 g) were boiled, minced and homogenized
with 4 ml of Ultra Resi-Analysed purge and trap water. Experi-
ments were carried out on samples spiked with furan at the final
concentration of 1 ␮g kg−1 . The vials were immediately closed
just after the spiking procedure.
A 22 two-level full factorial design (FFD) was performed [12]
to investigate the effects of temperature of extraction (T) and
time of extraction (t): low and high levels were T = 30–80 ◦ C,
t = 10–40 min. This experimental plan allows us to evaluate the
effects of the main factors and their interactions. The order of the
experiments was randomised in order to avoid possible memory
effect due to the analytical procedure.
A F-test comparing the experimental and calculated
responses at the centre of the experimental domain was per-
formed to evaluate the existence of relevant quadratic effects
and a star design was added to the factorial design experiments
if a relevant quadratic effect was observed [13,14].
The final regression model was then calculated using the Cen-
tral Composite Design (CCD) experiments, obtained both from
the FFD and the star design and used to find the optimal extrac-
tion conditions.
All statistical analyses were carried out by using the statistical
package SPSS 10.0 for Windows (SPSS, Bologna, Italy).
2.3. SPME analysis
SPME experiments were carried out using a 75 ␮m carboxen-
polydimethylsiloxane (CAR-PDMS) fiber (Supelco, Bellefonte,
PA, USA) by using manual device. The fiber was exposed to the
headspace of the 10 ml sample vial operating under the opti-
mized extraction conditions, i.e. extraction temperature: 30 ◦ C
and extraction time: 10 min. A constant magnetic stirring was
always applied.
269
The fiber was conditioned in the injection port of the gas
chromatograph at 300 ◦ C under helium flow for 1.5 h prior to
use. Desorption was carried out at the temperature of 230 ◦ C
for 3 min. A fiber blank was run between each sample to reduce
memory effects.
2.4. GC–MS analysis
A HP 6890 Series Plus gas chromatograph (Agilent
Technologies, Milan, Italy) equipped with MSD 5973 mass
spectrometer (Agilent Technologies) was used for GC–MS
analysis. Helium was used as the carrier gas at a flow rate of
1 ml min−1 ; the gas chromatograph was operated in splitless
mode with the PTV injector (Agilent Technologies) main-
tained at the temperature of 230 ◦ C and equipped with a PTV
multi-baffled liner (i.d. 1.5 mm, Agilent Technologies). Chro-
matographic separation was performed on a 60 m × 0.25 mm,
df 0.5 ␮m HP-INNOWAX capillary column (Agilent Technolo-
gies). The following GC oven temperature program was applied:
40 ◦ C for 8 min, 12 ◦ C min−1 to 200 ◦ C, 200 ◦ C for 2 min.
Transfer line and source were maintained at the temperature
of 220 and 230 ◦ C, respectively. The mass spectrometer was
operated in selected-ion monitoring mode (SIM) by recording
the current of the following ions: m/z 68 and 39 for furan
and m/z 72 and 42 for furan-d4 . The corresponding ion ratios
were used to confirm the identification of the analyte. A dwell
time of 100 ms was used for all the ions. Preliminarily, full
scan EI data were acquired to determine appropriate masses
for SIM under the following conditions: ionisation energy:
70 eV, mass range: 35–150 amu, scan time: 3 scan/s. All the
analyses were performed setting the electron multiplier voltage
at 1200 V.
Signal acquisition and elaboration were performed using the
HP Chemstation (Agilent Technologies).
2.5. Validation procedure
Validation of the SPME–GC–MS method was performed on
a blank boiled carrot sample (no signal was detected at the reten-
tion time of the analyte) according to EURACHEM guidelines
[15].
Detection and quantitation limits were calculated following
the approach already used in previous works [14,16] by con-
structing a calibration curve in the 50–500 ng kg−1 range.
Linearity was established over two order of magnitude in the
1–100 ␮g kg−1 range. Furan-d4 was used as internal standard at
the final concentration of 30 ␮g kg−1 . Six concentration levels
were analyzed performing three measurements at each concen-
tration level. Statistical analysis (Bartlett, lack-of-fit and Mandel
test) were performed to check the goodness of fit and linearity
[16,17]. The significance of the intercept (significance level 5%)
was established running a t-test.
Intra-day repeatability and between-day precision, estimated
over 3 days [12] were calculated in terms of RSD% on two
concentration levels, performing three replicates at each level.
Trueness was evaluated in terms of recovery by spiking
blank cooked carrots with 75 ng kg−1 and 75 ␮g kg−1 of furan.
270 F. Bianchi et al. / J. Chromatogr. A 1102 (2006) 268–272
Recovery (R%) [15] was calculated as follows:
cobs
R% = × 100
cspike
where cobs is the mean concentration of the fortified sample and
cspike is the spiked concentration. All the measurements were
replicated three times.
2.6. Samples
The following commercial baby-food samples were pur-
chased in local stores: vegetables (nine samples of car-
rots, French beans, mixed vegetables and zucchini deriving
from different producers) and fruits (nine samples of banana,
apple–banana and pear deriving from different producers). The
vegetables samples were prepared by heating the closed jars
in hot water, whereas the fruit samples were analysed without
previous heating, thus reproducing the typical behaviour of the
consumers.
Home-made baby-food samples were also analysed: fresh
fruit samples (apple, pear, banana and kiwi) and fresh vegetable
samples (potatoes, zucchini and green-peas) were purchased
in local stores. The fruit samples were minced and analysed,
whereas the vegetables were boiled, homogeneised and anal-
ysed.
3. Results and discussion
The aim of this work was to develop a simple and sensitive
method based on SPME–GC–MS analysis for furan detection
and determination in baby-food. All the analyses were per-
formed by using a 60 m × 0.25 mm, df 0.50 ␮m INNOWAX
column in order to obtain the elution of the analyte at a reten-
tion time compatible with quantitative purposes. The column
length as well as the film thickness allowed to elute furan with
an adequate retention time, at 7.5 min, thus obtaining a very
good furan chromatographic signal without the use of a cryofo-
cussing module as reported in a previous work [11]. The proper
selection both of the chromatographic conditions (column, tem-
perature program) and the acquisition mode (SIM) allowed to
avoid matrix interference.
Regarding the preparation of the solutions, obviously, being
furan very volatile and characterised by a low molecular
weight, all the solutions were stored in completely filled
vials. In fact, when partially filled vials were used, a general
decrease in the chromatographic response was evidenced dur-
ing the time as a consequence of the furan partition in the
headspace.
3.1. Fiber selection
In order to obtain the highest furan yield and the absence of
memory effects, fiber selection was performed by testing three
different coatings:
• Polydimethylsiloxane-divinylbenzene (PDMS-DVB) 65 ␮m;
• Divinylbenzene/carboxen/polydimethylsiloxane
(DVB/CAR/PDMS) 2 cm–50/30 ␮m;
• Carboxen-polydimethylsiloxane (CAR-PDMS) 75 ␮m.
With the aim to load the highest amount of analyte, aqueous
solutions containing furan at the concentration of 1 ␮g l−1 were
analyzed for 15 min at the temperature of 40 ◦ C, following a pre-
viously described approach [11,14]. For each fiber, three repli-
cated measurements were performed. Data obtained showed a
different behavior among the fibers. As already obtained by
Goldman et al., the CAR-PDMS resulted the fiber with the better
performance.
3.2. Experimental design
A 22 two-level factorial design was used to evaluate the
significance of the main and interaction effects of the factors
investigated. The experimental domain was defined taking into
account instrumental and operative limits, namely: tempera-
ture values lower than 30 ◦ C could not be maintained over long
time, owing to the thermal variability of the laboratory, whereas
temperature values higher than 80 ◦ C could produce analyte des-
orption from the fiber, whereas extraction times greater than
40 min would determine long analysis times. Eight replicated
measurements at the centre of the experimental domain were
performed in order to evaluate both the experimental error and
the existence of relevant quadratic effects. The presence of cur-
vature was tested by the F-test, as described in Section 2. A star
design was added to the factorial design experiments since rele-
vant quadratic effects were shown obtaining a Fcalc = 125 versus
a Ftab(1,7,α = 0.05) = 5.6.
Using the CCD experiments the following final regression
model was then calculated:
y = 146000(±12000) − 93000T (±11000)
+46000T 2 (±16000)
r2 = 0.91
The model was used to find the experimental conditions pro-
viding the optimal SPME extraction yield, i.e. temperature of
extraction = 30 ◦ C. As for the time of extraction, since it was not
a significant effect, a value of 10 min was used in order to reduce
analyses time.
As a general result, being furan very volatile, it has to be
observed that a low temperature value is sufficient to improve
its partition in the gas phase: on the contrary, higher temperatures
could produce its desorption from the fiber.
The use of a 10 min extraction reduced the analysis time, thus
allowing to perform about 20 analyses in a working day (8 h).
In a further step, the method was validated operating under
these conditions.
3.3. Method validation
The method was validated in terms of detection limit, quanti-
tation limit, linearity, precision and trueness by using the exper-
imental setting providing the optimised conditions.
 F. Bianchi et al. / J. Chromatogr. A 1102 (2006) 268–272
Table 1
Furan content in baby-food samples
Baby-food samples
Processed baby-food Apple–bananab
Bananab
Mixed vegetablesb
Carrotsb
French-beansb
Zucchinib
Home-made baby-food (fruits) Apple
Pear
Banana
Kiwi
Home-made baby-food (vegetables) Potatoes
Zucchini
Green-peas
a Mean of nine independent samples.
b Different producers.
c <LOD.
LOD and LOQ values, expressed as concentration, were cal-
culated by constructing an appropriate calibration curve in the
50–500 ng kg−1 range. Excellent results were obtained for LOD
(25.7 ng kg−1 ) and LOQ (41.7 ng kg−1 ) in boiled carrots, thus
proving the potential of the method for the determination of
furan at ultratrace levels.
Good linearity (y = 5.29x ± 0.05, r2 = 0.998, n = 18) was
demonstrated over two order of magnitude in the 1–100 ␮g kg−1
range by applying Mandel’s fitting test (Fcalc = 8.40,
Ftab(1,16,CI = 99%) = 8.53).
Method precision was evaluated testing two concentration
levels, i.e. 50 ng kg−1 and 100 ␮g kg−1 . Good results were
obtained both in terms of intra-day repeatability and between-
day precision: RSD% values lower than 1.2% at the highest con-
centration and lower than 3.5% at the lowest one were calculated
for intra-day repeatability, whereas between-day precision was
evaluated verifying homoscedasticity and performing ANOVA
on the data acquired over 3 days. ANOVA showed that mean
values were not significantly different among the 3 days obtain-
ing p values of 0.069 and 0.615 for the highest and lowest level,
respectively at 95% confidence level. RSD% lower than 3.6%
at both concentration levels were calculated.
Extraction recoveries of 91.5 ± 6.2% and of 96.1 ± 1.3%
(n = 3) were calculated at 75 ng kg−1 and 75 ␮g kg−1 , respec-
tively by addition of furan to blank boiled carrots samples.
These results show the good efficiency of the developed method
in terms of extraction recovery as well as of precision, with-
out using the standard addition method recommended by
FDA [5].
3.4. Baby-food samples
Finally, applicability of the validated method for the determi-
nation of furan in baby-food samples was demonstrated (Fig. 1).
Different commercial and home-made vegetable and fruit sam-
ples were submitted to analysis: data obtained are reported in
Table 1.
271
Furan (␮g/kg)a
2.78 ± 0.12
4.3 ± 0.2
99.1 ± 9.7
15.38 ± 0.56
50.09 ± 2.94
140.9 ± 2.4
–c
–c
–c
–c
0.336 ± 0.014
0.11 ± 0.025
1.03 ± 0.25
Fig. 1. GC–SIM–MS chromatograms of: (A) a commercial baby-food sample
(zucchini) and (B) an home made baby-food sample (zucchini).
As shown in the table furan was detected and quantitated at
different concentration levels depending both on the matrix and
on the cooking technique. In particular, all the processed baby-
food showed higher furan levels in comparison with the home-
made products. Even though the number of samples considered
does not make possible a definitive conclusion, this behaviour
could be ascribed to the different cooking technologies. In fact,
all the commercial processed foods are submitted to the heating
treatments in closed pots, whereas the home-made baby-food
are cooked in open containers. Under these circumstances, the
analyte, being extremely volatile can be more easily released
from the matrix.
As for the matrices analysed, it was observed that all the
fruit samples revealed a furan content lower than the vegetable
samples. In the case of the processed baby-food samples, this
difference could be explained taking into account that the fruit
samples are generally pasteurised, whereas the vegetables are
sterilised: the different heating treatment could be responsible
for furan production.
Nowadays, no Acceptable Daily Intake values have been
established: however, on the basis of our results, considering
a consumption of 350 g/die of canned baby-food and assuming
a body weight of 7.5 kg of a 6 month baby, this would result in
a maximum exposure of 7 ␮g/kg b.w./day.
Future investigations will be aimed to the analysis of other
matrices, like meat in order to verify if different technological
treatments are able to influence furan production.
4. Conclusions
A simple method for ultratrace determination of furan in
baby-food (vegetables and fruits) was developed and validated.
272 F. Bianchi et al. / J. Chromatogr. A 1102 (2006) 268–272
The analytical approach was based on the use of SPME cou-
pled with GC–MS analysis for food safety evaluation purposes.
The selection of the SPME fiber and the use of an experimen-
tal design allowed to calculate the optimal sampling conditions.
These conditions were found in correspondence with a tempera-
ture of extraction of 30 ◦ C. A time of 10 min was chosen in order
to obtain shorter analyses time. Under these experimental set-
ting values, the method proved accurate and sensitive allowing
the determination of the investigated analyte at low ng kg−1 lev-
els. The method developed was successfully applied to analyse
a number of commercial and home-made baby-food samples,
thus revealing differences related to the production technique.
References
[1] European Commission (2000a, January 12), White paper on food safety
(COM 1999, 719 final).
[2] International Agency for Research on Cancer, IARC Monographs on the
Evaluation of Carcinogenic Risks to Humans, vol. 63, 1995.
[3] http://www.cfsan.fda.gov/∼lrd/fr040510.html, accessed 27 September
2005.
[4] J.A. Maga, Food Sci. Nutrit. 4 (1979) 355.
[5] http://www.cfsan.fda.gov/∼dms/furan.html, accessed 27 September
2005.
[6] http://www.fda.gov/ohrms/dockets/ac/cfsan04.html, accessed 27 Septem-
ber 2005.
[7] http://www.cfsan.fda.gov/∼dms/furandat.html, accessed 27 September
2005.
[8] A. Becalski, S. Seaman, J. AOAC 88 (2005) 102.
[9] T. Kuballa, S. Stier, N. Stricow, Deutsche Lebensmittel-Rundschau 101
(2005) 229.
[10] A. Becalski, D. Forsyth, V. Casey, B.P.Y. Lau, K. Pepper, S. Seaman,
Food Add. Contam. 22 (2005) 535.
[11] T. Goldmann, A. Perisset, F. Scanlan, R.H. Stadler, Analyst 130 (2005)
878.
[12] G.E.P. Box, W.G. Hunter, J.S. Hunter, Statistic for Experimental, Wiley,
New York, 1978.
[13] A.I. Khuri, J.A. Cornell, Response Surfaces, Marcel Dekker, New York,
1987.
[14] F. Bianchi, M. Careri, C. Corradini, A. Mangia, M. Musci, CAC 1
(2005) 129.
[15] The Fitness for Purpose of Analytical Methods: A Laboratory Guide
to Method Validation and Related Topics, EURACHEM Guide,
First English Edition 1.0-1998, LGC (Teddington) Ltd., http://www.
eurachem.ul.pt/.
[16] B. Antolini, F. Bianchi, M. Bottazzi, M. Careri, M. Musci, Chro-
matographia 60 (2004) 323.
[17] N. Draper, H. Smith, Applied Regression Analysis, Wiley, New York,
1981.