Trends in Food Science & Technology 18 (2007) 365e372

Review

A review of the
occurrence, Occurrence of furan in foods
Initial findings by the US-FDA
In 2004, the US Food and Drug Administration (FDA)
formation and reported the first results of tests for furan in selected foods,
mainly heat-processed foods sold in jars and cans (FDA,

analysis of furan in 2004). The FDA reported the results for 334 selected foods,
including many repeat purchases. The majority were baby
foods, and a large proportion of the remainder were infant
heat-processed foods formulae. Other foods included canned vegetables, fruit,
meat and fish, pasta sauces, nutrition drinks, fruit preserves,
beers, and coffees. The analytical method used had a limit
C. Crews* and L. Castle of quantification of about 5 mg/kg (parts per billion) for
most foods and 2 mg/kg for most liquids including coffee.
Defra Central Science Laboratory, The FDA found that many heat-treated foods contained de-
Room 10GA11, Sand Hutton, tectable furan, including almost all of the baby foods sold
York YO41 1LZ, UK (Tel.: D44 (0)1904 462549; in jars and many of those sold in cans. The highest levels
fax: D44 (0)1904 462111; e-mail: c.crews@csl.gov.uk) were for vegetables, particularly beans, squash, and sweet
potatoes, packed in jars or cans.

In 2004, the US-FDA announced the findings of its exploratory

European data
surveys of furan in selected foods. In the 3 years since, there
The European Food Safety Authority (EFSA) followed
have been a number of publications on furan in foods. This re-
the FDA initiative and began to collect information on the
view looks at furan in foods, its metabolism and toxicity. The
methods of analysis, occurrence and formation in food,
important factor of heating is explored along with the precur-
exposure through consumption, and the toxicity of furan.
sors and chemical mechanisms that may be involved. Finally,
EFSA reported provisional findings (EFSA, 2004a,b) which
prospects for limiting the occurrence, formation and retention
included concentration data for some European samples.
of furan in foods are discussed.
The data were summarised with the FDA data and expressed
as ranges over 11 food categories with quantification limits
Introduction
of 1e2 mg/kg.
Furan is a small cyclic ether with aromatic character and
EFSA has subsequently issued a call for more informa-
a low boiling point of 31  C. Early reports of furan in food
tion on the occurrence of furan in foods (EFSA, 2006).
were collated by Maga (1979) who reported that furan was
A revision of the tabulated data compiled from the
present in a number of foods, with the highest levels being
annexe to the EFSA report (EFSA, 2004b) and updated
found in coffee. This has been followed by a series of food
(November 2005) results from the FDA (FDA, 2004a) is
surveys of heat-processed foods and studies into its formation,
provided in Table 1. It covers an enlarged range of food cat-
toxicity and analysis which are summarised in this paper.
egories and provides average furan levels for most of the
Furan is carcinogenic in rats and mice and has been clas-
positive samples calculated from the data available. Furan
sified as ‘possibly carcinogenic to humans’ (International
levels of over 100 mg/kg were found principally in three
Agency for Research on Cancer, 1995). The European
categories of major foods; coffee, baby food, and sauces
Food Safety Authority (EFSA) has expressed the opinion
and soups. Furan was detectable in 262 of the 273 baby
that ‘furan is clearly carcinogenic in rats and mice’ and
food samples reported (96%) with an average level of
that ‘the weight of evidence indicates that furan-induced
28 mg/kg, and in 70 of 71 infant foods at similar levels.
carcinogenicity is probably attributable to a genotoxic
These data alone were too few to draw firm conclusions
mechanism’ (EFSA, 2004a).
on the tendency of different foods to form furan, because
exact details on food composition and the heating condi-
* Corresponding author. tions of time and temperature were not known. Both the
0924-2244/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tifs.2007.03.006
366 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372

Table 1. Levels of furan in foods (mostly canned and jarred) reported by EFSA and FDA

Sample type Number Furan mg/kg

Total Negative Minimum Maximum Average of positivesa
Baby foods 273 11 1 112 28
Infant foods 71 1 1.3 87.3 27
Infant formulaeb 42 14 2.5 27 12
Coffee beans/powder 19 0 239 5050 1500
Coffee brewed 38 4 3 125 50
Beers 14 4 0.8 1 6
Bread 13 12 30 30 30
Bread, toasted 6 5 18 18 18
Crisp breads/crackers 4 0 4.2 18.6 12
Candy and chocolate 20 8 0.5 10.3 3
Desserts/puddings 12 1 1.5 13 4
Fish 9 3 1.5 8 6
Fruit 28 2 0.9 11 4
Fruit juices 51 27 0.5 31 5
Fruit preserves 47 2 0.9 37 7
Gravies 8 0 13.3 174 48
Malt 6 1 16 195 91
Convenience meals 32 0 3 94 35
Meats 15 9 1.7 39 18
Miscellaneous 47 9 1 420c 28
Nutrition drinks/shakes 22 2 1.1 174 29
Sauces 41 4 3.3 46 11
Soups 58 7 3 125 32
Soy sauces 12 0 17.2 91 50
Maple syrup 3 0 8.6 88.3 45
Tortilla and potato chips 6 1 4.4 39.1 16
Vegetable juices 8 0 3.2 40 10
Vegetables (in cans and jars) 50 9 0.8 48 9
Baked beans 26 0 23.3 122 59
Data taken from EFSA (2004a,b) and FDA (2004a).
a
Some values are approximate (estimated from ranges).
b
Includes ready to eat samples and concentrates prepared as for consumption but with cold water.
c
Two samples (caramels) contained 192 and 400 mg/kg with all remaining samples below 100 mg/kg.

US and European surveys were restricted to specific prod-
ucts thought likely to contain furan on account of being
heated in sealed containers, and frequently several samples
of the same brand were analysed. The data therefore prob-
ably do not represent the actual distribution of furan in
foods. EFSA has called for further concentration data for
furan in food, in a database that will be used to estimate
consumers’ exposure (EFSA, 2006).
Foods not cooked in closed containers
Substantial levels of furan (20e200 mg/kg) have been re-
ported in foods not cooked in closed containers, including
potato crisps, crackers and crisp breads (Hoenicke, Fritz,
Gatermann, & Weidemann, 2004) and toasted bread
(Hasnip, Crews, & Castle, 2006). This area is worthy of fur-
ther study, particularly with regard to the levels of furan
present in these products as consumed. It might not be ex-
pected that furan persists in toasted bread and its behaviour
in packeted potato crisps for example is unknown.
Coffee
Furans have long been known as normal components of
coffee flavour volatiles. For example, Merritt, Bazinet,
Sullivan, and Robertson (1963) reported that 6.5% of the
volatiles were furanic compounds. The number of volatile
furans identified in roasted coffee has increased in line
with improvements in analytical techniques with more
than 800 reported by 1994.
High levels of furan are found in roasted coffee beans,
probably on account of the roasting process where the
high temperatures exceed most other food processing pro-
cedures. However, even after accounting for dilution, losses
are substantial during coffee brewing by all of the usual
methods such as expression or filtration, and in the prepa-
ration of instant coffee. Table 2 shows furan in coffee prod-
ucts as reported by Kuballa, Stier, and Strichow (2005).
Automatic coffee machines produced brews with the high-
est levels of furan, because a higher ratio of coffee powder
to water is used giving a lower dilution factor, and because
the closed system favours retention of furan. Much lower
levels were produced by standard home coffee-making
machines and by manual brewing. These limited data
show the substantial reduction on brewing coffee from
coffee powder. Apart from this clear trend, the data are
too few to draw any firm conclusions on, e.g. differences
between countries.
 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372 367

Table 2. Furan in coffee beans and its transfer into brewed coffee Metabolism
Furan is absorbed quickly into the body but is also ex-
Coffee type Brewing method Furan in Furan in
the bean brew
creted with high efficiency. Radiotracer experiments have
(mg/kg) (mg/kg) shown that a 84% of a single oral dose of furan was metab-
Whole beans Home machine 3600e6100 9e33
olized within 24 h in rats, and the remainder exhaled.
Whole beans Manual machine 3600e6100 17e24 Twenty percent of the radioactivity was contained in over
Whole beans Automatic machine 3600e6100 57e115 10 compounds excreted in the urine with a further 22%
Whole beans French press (cafetiere) 3600e6100 33e66 excreted in faeces (Burka, Washburn, & Irwin, 1991).
Regular powder In cup 800e3400 9e66 Repeated doses accumulate in the liver. Absorbed furan is
Decaffeinated 1000e2800 8e31
powder
metabolized rapidly by cytochrome P-450 enzymes (princi-
Espresso 1500e2000 28e60 pally CYP2E1) via ring opening to form carbon dioxide
Instant 200e700 3e15 and cis-2-butene-1,4-dialdehyde (Chen, Hecht, & Peterson,
Instant 2000e2200 14e25 1995; Ravindrananth, Boyd, & Burka, 1984), which binds
Data summarised from Kuballa, Stier, and Strichow (2005). to protein and nucleosides (Burka et al., 1991; Byrns,
Predecki, & Peterson, 2002), and has been found as
a mono-glutathione conjugate in the urine of rats exposed
Other possible dietary sources of furan to radiolabelled furan (Peterson, Cummings, Vu, & Matter,
Various minor sources of furan in the diet are listed in the 2004).
review by Maga (1979), including canned meats. Miscella- Experiments with human hepatocytes in primary culture
neous foods shown by EFSA and the FDA to contain rela- have shown that metabolism by the CYP2E1 is so rapid that
tively high levels of furan include malts (16e195 mg/kg), the rate limiting step in elimination of ingested furan is the
gravies (13e174 mg/kg), caramels (220e400 mg/kg) and speed of its delivery to the liver (Kedderis & Held, 1996).
soy sauce (17e90 mg/kg). Malts, gravies and caramels are
consumed in relatively small quantities but soy sauce may Toxicity
be consumed in larger quantities, particularly in Asian coun- Toxicity studies have shown furan to be a potent carcinogen
tries. The FDA survey included 31 infant formula products affecting many organs. Tests carried out under the US National
(with many duplicated samples). Furan was found in 11 of Toxicology Program (NTP) used furan applied by gavage to
13 samples with added iron and in 6 of 18 samples without rats and mice at doses of between 20 and 160 mg/kg (NTP,
iron, typically at 8e10 mg/kg. 1993). Mortality was increased in both species over 16 days.
Administration of low maximum doses over 13 weeks caused
The effects of domestic cooking on furan formation weight loss, increase in liver and kidney weights, decrease in
Most of the results in the literature to date are for heat- thymus weight and toxic lesions of the liver and kidney in rats
processed foods tested as purchased. There have been a few and mice which increased in severity with dose. There were
studies to investigate if furan persists in food during normal mortalities among the rats (9 of 10 males and four of ten
preparation processes leading to eating. Initial thoughts females) but not for the mice (10 of each sex).
were that such a volatile substance may simply evaporate Furan administered to 50 mice at 8 or 15 mg/kg bw
from the food, e.g. when cans or jars were opened. These 5 days per week for 2 years caused loss of body weight
thoughts proved to be naive. Furan is in fact rather persis- at 15 mg/kg bw, and significantly increases in hepatocellu-
tent in foods. Furan formed in foods following heating in lar adenomas and carcinomas.
sealed containers during industrial processing, is not lost Administration of higher maximum doses (30 mg/kg
to a very significant extent when the foods are warmed furan by gavage 5 days per week for 13 weeks) induced
ready to eat (Hasnip et al., 2006). The exceptions are cholangiocarcinoma of the liver in all of 50 male rats.
vigorous boiling or cooking where furan can be lost, pre- Hepatocellular carcinomas were found in 6 of 40 rats that
sumably by evaporation and by entrainment in the large survived beyond the dosing period, at 9 months. Cholangio-
volumes of steam that are released. On the other hand, carcinomas were found in all 10 surviving male rats at
warming the food even in lightly lidded containers can in- 9 and 15 months.
crease furan levels, so any additional formation appears to
be balanced by evaporative losses (Hasnip et al., 2006). Genotoxicity
Studies have shown furan not to be mutagenic in some
strains of Salmonella typhimurium both with and without
Toxicology S9 metabolic activation (NTP, 1993). Some evidence of
A risk assessment of the toxicity of furan was published mutagenicity has been shown in strain TA100 (Lee, Bian,
by the European Food Safety Authority (EFSA) in 2004. & Chen, 1994). Furan was mutagenic to mouse lymphoma
The crucial conclusion was that furan is carcinogenic to cells with and without S9 activation (McGregor et al.,
rats and mice with a clear dose-dependency and probably 1988) and, in another study, with and without S9 activation.
acting by a genotoxic mechanism. High doses of furan (250 mg/kg bw by i.p. injection)
368 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372
induced structural chromosome aberrations but not sister
chromatid exchange in mice bone marrow cells. Single
oral doses of 200 or 100 mg/kg bw furan did not induce un-
scheduled DNA synthesis in mouse or rat hepatocytes
in vivo (NTP, 1993).
cis-2-Butene-1,4-dial is similar to a,b unsaturated com-
pounds that react with DNA and are known mutagens. It
was directly mutagenic at non-toxic concentrations in a
S. typhimurium strain (TA104) sensitive to aldehydes but
not to several other strains (Peterson, Naruko, & Predecki,
2000). It is likely that furan or cis-2-butene-1,4-dial reacts
with DNA in target cells and can play a role in furan-
induced tumors.
Furan causes loss of ATP after bioactivation to metabo-
lites which cause an irreversible uncoupling of hepatic
mitochondrial oxidative phosphorylation, this activates
cytotoxic enzymes, including endonucleases that produce
DNA double-strand breaks prior to cell death (Kedderis
& Ploch, 1999; Mugford, Carfagna, & Kedderis, 1997).
Analysis for furan
The FDA published the first quantitative method for fu-
ran in food (FDA, 2004b). It used sample preparation under
cold conditions and headspace sampling following incuba-
tion at 80  C. Chromatographic separation used a PLOT
(porous layer open tubular) column with mass spectromet-
ric detection in selected ion monitoring mode. The PLOTQ
column used has a bonded polystyrene-divinylbenzene
based phase which separates small volatile molecules effec-
tively. Quantification was based on standard additions and
used a deuterated furan internal standard. As interest in
furan testing grew several variations of this procedure
were tested and introduced by other laboratories.
Sample preparation
Because of its volatility, headspace sampling is the obvi-
ous method for the analysis of furan. To avoid losses, the
food samples need typically to be chilled (ca. þ4  C) be-
fore handling, and to be homogenised briefly but effectively
using a chilled blender with the sample sitting in an ice-
bath. Puréed and liquid samples can be weighed directly
into the headspace vial and mixed with chilled water prior
to addition of internal standard and spiking solutions. Solid
samples may need homogenisation with cold water using
a top drive homogeniser.
Headspace analysis
Partitioning of furan from the food sample into the head-
space in the vial is affected by time, temperature, and the
mobility of the sample. Effective partitioning has been
ensured by prolonging the equilibration incubation time
(Becalski et al., 2005), or improving the efficiency of auto-
mated shaking by adding glass beads to the headspace vial
(Hasnip et al., 2006). In any event sufficient water must be
present or added to ensure that the sample is completely
mobile before analysis.
Normal headspace procedures involve heating the sam-
ple to promote volatilisation of analyte into the headspace
gas phase. With furan, excessive heating is both inadvisable
and unnecessary. Becalski et al. (2005) showed that in-
creasing the headspace incubation temperature from 30 to
50  C caused only a 50% increase in the furan peak area
of an aqueous standard. Addition of salt more than doubled
the furan signal. In view of the danger of forming extra fu-
ran even at quite low temperatures, most workers have set-
tled on an incubation temperature of 50  C or below and
this gives adequate sensitivity, about 1 mg/kg.
Senyuva and Gokmen (2005) demonstrated furan forma-
tion in green coffee (4 mg/kg) and in tomato and orange jui-
ces on incubation at 40  C for 30 min. For any new sample
type, it is necessary to check that the analytical procedure,
and especially the headspace incubation temperature used,
does not cause the formation of extra furan.
GC conditions
Many different GCeMS conditions have been used but
they are mostly simple variations on a common theme
(Becalski & Seaman, 2005; Bianchi, Careri, Mangia, &
Musci, 2006; Cerny & Davidek, 2003; FDA, 2004; Gold-
mann, Perisset, Scanlan, & Stadler, 2005; Hasnip et al.,
2006; Ho, Yoo, & Tefera, 2005; Reinhard, Sager, Zimmer-
mann, & Zoller, 2004; Senyuva & Gokmen, 2005). Most
workers use PLOT columns or an equivalent. With a starting
temperature of 20e50  C these columns give adequate
retention and separation of furan from other volatiles.
Identification using GCeMS
The identification of furan is assured by checking for the
correct GC retention time (e.g. 2% of that of standards)
along with the correct ratio of the molecular ion for furan
at m/z 68 compared to its fragment ion at m/z 39, again
compared to the ratio seen with standards (e.g. agreement
to 10%). The low abundance of the m/z 39 ion constrains
the detection and quantitation limits.
Quantification using GCeMS
Quantification has been based on standard additions
or external calibration graphs, both incorporating
a deuterium-labelled internal standard. High levels of inter-
nal standard must be avoided to prevent a contribution to
the m/z 68 analyte signal from the [M-d2]þ$fragment ion.
Analytical recovery for the major matrices, brewed
coffee and baby food, is consistently better than 90%
with limits of detection of less than 1 mg/kg. A number of
method performance characteristics, typically measurement
uncertainty and trueness, have been reported in detail and
show that with care the current headspace methods perform
very well.
Headspace sampling by SPME
Several authors have reported SPME (solid phase micro-
extraction) methods for furan. This includes testing for
 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372
furan in coffee (Ho et al., 2005), orange juice (Fan, 2005a),
foods in general (Bianchi et al., 2006; Goldmann et al.,
2005) and model reaction mixtures (Cerny & Davidek,
2003). In direct headspace analysis a portion of the head-
space gas is taken and injected directly into the GCeMS.
In contrast, in SPME a needle coated with a polymeric ma-
terial is first exposed to the headspace vapours to absorb
volatiles (for 10e60 min in the references cited), and is
then desorbed thermally (1e5 min; 90e300  C) in the in-
jection port of the GC to drive off the volatiles onto the
GC column. SPME allows sample concentration and
affords higher sensitivity. Limits of detection in the low
ng/kg range can be achieved for furan (Bianchi et al.,
2006; Goldmann et al., 2005) although in practice the effec-
tive detection limit can be constrained by background levels
of furan in method blanks.
Formation of furan
There appear to be many pathways to furan, with the
major routes beginning with sugars, ascorbic acid, or unsat-
urated fatty acids. Simplified versions of the proposed ma-
jor routes are shown in Fig. 1. The thermal degradation and
rearrangement of carbohydrates is encountered in non-
enzymic (Maillard) browning reactions during food pro-
cessing and cooking. Even in simple component mixtures,
large numbers of furanic compounds are formed. For exam-
ple, in a heated mixture of serine, threonine and sucrose (as
found in green coffee) about 350 furans were identified
(Baltes & Bochmann, 1987). Substituted furans are familiar
products of Maillard reactions but furan itself has been re-
ported less often because its polarity and volatility placed it
outside the working range of chromatographic systems ad-
justed to monitor larger, less volatile compounds. Perez
Locas and Yaylayan (2004) studied the formation of furan
in simple sugar/amino acid systems heated at 250  C.
Ascorbic acid formed the highest level of furan, with dehy-
droascorbic acid also forming considerable quantities.
Heating sugars or amino acids alone did not produce signif-
icant levels of furan, with the exception of erythrose. Furfu-
ral is a common product in browning and other reactions.
Small molecules such as furfural and furoic acid are un-
doubtedly part of the route to furan from the larger precur-
sors chemicals (Becalski & Seaman, 2005).
Studies on the efficiency (molar yield) of furan formed
from some precursor chemicals in models systems (Becalski
& Seaman, 2005; Mark, Pollien, Lindinger, Blank, & Mark,
2006) have tended to be sometime contradictory and diffi-
cult to interpret. It does seem that ascorbic acid/dehydroas-
corbic have the highest potential to form furan, followed by
polyunsaturated fatty acids and then sugars. The molar yield
can be up to about 1% (Stadler, 2006).
Formation of furan from carbohydrates
Yaylayan, Keyhani, and Wnorowski (2000) and Yaylayan,
Machiels, and Istasse (2003) reported furan formation from
369
model Maillard systems containing serine and cysteine and
later studied simple sugar/amino acid combinations heated
at 250  C. Highest levels of furan per mole of starting mate-
rial were produced from mixtures of glycoaldehyde with
L-alanine. Serine produced about 30% of this quantity
when heated with sucrose or ribose, and about 10e25%
when heated with fructose or glucose. Cysteine and alanine
also produced significant furan on heating with glucose and
some other mixtures formed low levels.
The reactions between several amino acids and carbohy-
drates on heating form aldotetrose derivatives in an aldol
condensation. In the example of serine it has been shown
from 13C labelling that pyrolysis of the acid forms acetalde-
hyde and a glycoaldehyde which undergo aldol condensa-
tion leading to the aldotetrose intermediate of furan. The
aldotetrose undergoes cyclisation to form furan.
Labelling studies carried out using serine mixtures with
glucose show both serine and glucose contribute to furan
formation (Perez Locas & Yaylayan, 2004). The major
pathway is through initial formation of 1-deoxyosone in
the presence of amino acids. The 1-deoxyosone undergoes
cleavage to form an aldotetrose which can dehydrate to
form 3-furanone which leads to furan via reduction and
dehydration. An alternative route via 1-deoxyosone begins
with retro-aldol cleavage of glucose.
An important route to furan from glucose begins with
dehydration and retro-aldol cleavage forming 2-deoxy-3-
ketoaldotetrose which leads to 3-furanone. The third route
is from 3-deoxyosone (produced with 1-deoxyosone on
pyrolysis) which forms 2-deoxyaldotetrose by a-dicarbonyl
cleavage and oxidation. The 2-deoxyaldotetrose cyclicises
to furan. The final stages of furan formation from sugars
and Maillard type reaction appear to involve reduction of
3-furanone by formic acid, itself formed in the pyrolysis
of sugars.
Formation of furan from ascorbic acid
In browning reactions ascorbic acid behaves in similar
manner to reducing sugars. It degrades by two parallel mech-
anisms (Yuan & Chen, 1999). In acid aerobic conditions as-
corbic acid is oxidised to dehydroascorbic acid and is then
hydrolysed to 2,3-diketogulonic acid (DKG) by ring
cleavage and addition of water. Decarboxylation of 2,3-
diketogulonic acid forms xylosone which is ultimately
dehydrated to 3-hydroxy-2-pyrone and 2-furoic acid, or
3-deoxypentosulose and ultimately furfural. Ascorbic acid
can form 2-deoxyaldotetrose and then furan as described
by Perez Locas and Yaylayan (2004). Dehydroascorbic
acid cannot be formed in anaerobic conditions, where ascor-
bic acid can undergo hydrolysis, b-elimination and decar-
boxylation to 3-deoxypentosulose which is an intermediate
of 2-deoxyaldotetrose. Furfural can also be formed from
ascorbic acid in both aerobic and anaerobic conditions. Shi-
noda, Komura, Homma, and Murata (2005) have proposed
pathways for the formation of the furan precursors fural-
dehyde and 2-furoic acid.
370 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372

H
HO OH
O
HO
O OH
O
HO
OH
Ascorbic acid
2-deoxy-aldotetrose

H
O HO
- H2O
OH
OH O
O
OH Furan

aldotetroses

H
HO
OH
O O

4-hydroxy-2-butenal

O O

oxidized fatty acid

Fig. 1. Furan formation from sugars, ascorbic acid or fatty acid.

When heated in the absence of water dehydroascorbic
acid can form a cyclic hemiketal, preventing furan formation
(Perez Locas & Yaylayan, 2004). Ascorbic acid produced
a number of substituted furans on heating at 300  C in dry
systems (Vernin, Chakib, Rogachev, Obretenov, & Parkanyi,
1997). The major compounds were furfural (about 70%) and
furoic acid.
When ascorbic acid was mixed in model systems with
single amino acids (glycine or serine), sugar (erythrose)
or unsaturated fatty acid (linoleic), the mixtures produced
far less furan on heating than did ascorbic acid alone
(Mark et al., 2006).
Ascorbic acid contributes to the browning of orange juice
on storage, when 3-hydroxy-2-pyrone, 5-hydroxymethylfur-
fural (HMF), furfural, 5-hydroxymaltol, and 2-furoic acid
are formed. A study of the formation of furanic compounds
(not including furan) in orange juice and juice models
(Shinoda, Murata, Homma, & Komura, 2004; Shinoda
et al., 2005) confirmed that the furan precursors furfural
and 2-furoic acid were derived from ascorbic acid in a reac-
tion stimulated by the sugars or chelating agents.
Heating orange juice to boiling for 5 min produced 1.4 ng/
ml furan, and more was produced on autoclaving for 25 min
at 121  C (Fan, 2005a). Very little furan was produced on
warming apple juice but on autoclaving, far more furan
was produced than when autoclaving orange juice. These
effects were correlated with decomposition rates of ascorbic
acid. Subsequently, it was shown that furan formation by
heating or irradiation was affected by the pH and concen-
tration of sugars and ascorbic acid in solution (Fan, 2005b).
 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372
Formation of furan from fatty acids
Furan is formed from unsaturated fatty acids and in
model systems the yield increases as the degree of unsatu-
ration increases. Becalski and Seaman (2005) showed that
monounsaturated acid (oleic) did not form furan. Linoleic
acid heated for 30 min at 118  C in water formed 125 ng/g
furan and linolenic acid formed 625 ng/g. Similar levels
were produced when the acids were present as triacylgly-
cerols. Mark et al. (2006) showed that furan is formed
from linolenic acid at lower temperatures (110e220  C)
than from ascorbic acid or Maillard systems. Formation
of furan has therefore been linked with the process of
free radical autoxidation (Perez Locas & Yaylayan, 2004).
The effects of lipid oxidation catalysts such as Fe(II) or
of antioxidants, are sometime contradictory. Metal ions can
promote the oxidation of fatty acids and consequently en-
hance formation of furan. Ferric ions increased furan for-
mation in linoleic acid by 79% and in trilinolein by 29%
(Mark et al., 2006). However, the data for linolenic acid
are less clear, with decreased formation reported by Mark
et al. (2006) but an increase reported by Becalski and
Seaman (2005). Given the complicated and competing
reaction pathways available in autoxidation processes,
some of which may lead to furan and some not, limiting
autoxidation may not always lead to a corresponding reduc-
tion in furan formation.
Epoxidised linseed (ELO) and soya bean (ESBO) oils
are used extensively in the food packaging industry as plas-
ticisers in the plastic seals fitted to metal lids for jars. ELO
and ESBO contained more furan than the unepoxidised oils
and produced more furan on heating (Hasnip et al., 2006).
However, the quantities of ELO and ESBO used in food
containers mean that any transfer of furan from this source
into foods is insignificant.
In a similar vein, foods heated in cans and jars make
contact with lacquered metal surfaces. Foods heated in con-
tact with normal can-coating polymers, including polyester
urethane, epoxyphenolic, organosol, and epoxyanhydride
coatings, did not form more furan than when the foods
were simply heated in the absence of any coating material
(Hasnip et al., 2006).
Dietary exposure
EFSA estimated exposure to furan based on the limited
data available. Baby food was of particular interest as
a high proportion of samples sold in jars and cans contained
furan and such foods may form the sole diet of many
babies. The estimated intake based on consumption of
baby food from glass jars was <0.2 to 26 mg furan per
day or <0.03 to 3.5 mg/kg bw per day for a 6 month old
baby weighing 7.5 kg (EFSA, 2004). The daily intake for
adults from canned or jarred vegetables (35 samples) was
estimated to be 1.1 to 23 mg/person, and from beer (12 sam-
ples) was 1.3e50 mg/person. The daily intake from coffee
based on data from 45 samples was 2.4 to 116 mg/person,
making coffee the major dietary source for adults.
371
Prospects for mitigation measures
Since the carcinogenicity of furan is probably attribut-
able to a genotoxic mechanism (EFSA, 2004) levels in
food should be kept ALARA e as low as reasonably
achievable. The obvious way to reduce heat-induced toxins
in food is by changes to the heating regime or by reduction
in the content of precursors.
Reduction of furan in foods is likely to be more chal-
lenging compared to other process contaminants, for two
reasons. First, there may be little room for manoeuvre to
lower heating times and temperatures because the processes
of pasteurisation and sterilisation are for the microbiologi-
cal safety of foods. Second, furan has a wide range of pre-
cursors. Ascorbic acid shows the highest potential to form
furan, followed by polyunsaturated fatty acids and then
sugars (Stadler, 2006). Ascorbic acid and polyunsaturated
fatty acids are regarded as desirable food components
because of their health benefits.
A third mitigation measure that could be explored is to
make use of the volatility of furan. However, again this
may be of limited applicability because for microbiological
reasons canned and jarred foods have to be sealed hermet-
ically and for coffee it would be technically difficult to
purge coffee of furan whilst retaining all the flavour and
aroma substances that the consumer demands.
The best approaches appear so far to involve interven-
tion in the reaction mechanisms. For example, formation
of furfural from ascorbic acid in model orange juice was re-
pressed by the presence of ethanol and mannitol acting as
free radical scavengers (Shinoda et al., 2005). Reduction
of atmospheric oxygen reduces the autoxidation of unsatu-
rated fatty acids and also reduces furan formation from sev-
eral precursors, notably ascorbic acid, as does the addition
of sulphite (Mark et al., 2006). Therefore, modification of
the atmospheres within heating systems might be effective
in reducing furan in foods.
Acknowledgements
This work was co-funded by the UK Food Standards
Agency and the European Union HEATOX project.
References
Baltes, W., & Bochmann, G. (1987). Model reactions on roast aroma
formation. 1. Reaction of serine and threonine with sucrose under
the conditions of coffee roasting and identification of new coffee
aroma compounds. Journal of Agricultural and Food Chemistry,
35(3), 340e346.
Becalski, A., Forsyth, D., Casey, V., Lau, B. P. Y., Pepper, K., &
Seaman, S. (2005). Development and validation of a headspace
method for determination of furan in food. Food Additives and
Contaminants, 22(6), 535e540.
Becalski, A., & Seaman, S. (2005). Furan precursors in food: a model
study and development of a simple headspace method for deter-
mination of furan. Journal of AOAC International, 88(1), 102e106.
Bianchi, F., Careri, M., Mangia, A., & Musci, M. (2006). Development
and validation of a solid phase micro-extractionegas chromato-
graphyemass spectrometry method for the determination of furan
in baby-food. Journal of Chromatography A, 1102(1e2), 268e272.
372 C. Crews, L. Castle / Trends in Food Science & Technology 18 (2007) 365e372
Burka, L. T., Washburn, K. D., & Irwin, R. D. (1991). Disposition of
[14C]-furan in the male F344 rat. Journal of Toxicology and
Environmental Health, 34(2), 245e257.
Byrns, M. C., Predecki, D. P., & Peterson, L. A. (2002). Characteriza-
tion of nucleoside adducts of cis-2-butene-1,4-dial, a reactive
metabolite of furan. Chemical Research in Toxicology, 15(3),
373e379.
Cerny, C., & Davidek, T. (2003). Formation of aroma compounds from
ribose and cysteine during the Maillard reaction. Journal of Agri-
cultural and Food Chemistry, 51(9), 2714e2721.
Chen, L. J., Hecht, S. S., & Peterson, L. A. (1995). Identification of
cis-2-butene-1,4-dial as a microsomal metabolite of furan.
Chemical Research in Toxicology, 8(7), 903e906.
EFSA. (2004a). Report of the Scientific Panel on Contaminants in the
Food Chain on provisional findings of furan in food. EFSA Journal,
137, 1e20, Available at http://www.efsa.eu.int/science/contam/
contam_documents/760/contam_furan_report7-11-051.pdf
EFSA. (2004b). Report of the CONTAM Panel on provisional
findings on furan in food. Annexe corrigendum. Available at
http://www.efsa.europa.eu/etc/medialib/efsa/science/contam/
contam_documents/760.Par.0002.File.dat/furan_annex1.pdf
EFSA. (2006). Invitation to submit data on furan in food and beverages.
Available at http://www.efsa.europa.eu/en/science/data_
collection/furan.html
Fan, X. (2005a). Impact of ionizing radiation and thermal treatments
on furan levels in fruit juice. Journal of Food Science, 70(7),
E409eE414.
Fan, X. (2005b). Formation of furan from carbohydrates and ascorbic
acid following exposure to ionizing radiation and thermal
processing. Journal of Agricultural and Food Chemistry, 53(20),
7826e7831.
FDA. (2004a). Exploratory data on furan in food. Available at
http://www.cfsan.fda.gov/wdms/furandat.html
FDA. (2004b). Determination of furan in foods. Available at
http://www.cfsan.fda.gov/wdms/furan.html
Goldmann, T., Perisset, A., Scanlan, F., & Stadler, R. H. (2005). Rapid
determination of furan in heated foodstuffs by isotope dilution solid
phase micro-extractionegas chromatographyemass spectrometry
(SPMEeGCeMS). Analyst, 130(6), 878e883.
Hasnip, S., Crews, C., & Castle, L. (2006). Some factors affecting the
formation of furan in heated foods. Food Additives and Contami-
nants, 23(3), 219e227.
Ho, I. P., Yoo, S. J., & Tefera, S. (2005). Determination of furan levels in
coffee using automated solid-phase microextraction and gas
chromatography/mass spectrometry. Journal of AOAC
International, 88(2), 574e576.
Hoenicke, K., Fritz, H., Gatermann, R., & Weidemann, S. (2004).
Analysis of furan in different foodstuffs using gas chromato-
graphy mass spectrometry. Czech Journal of Food Science, 22,
357e358.
International Agency for Research on Cancer. (1995). Dry cleaning,
some chlorinated solvents and other industrial chemicals. In:
Monographs on the evaluation of carcinogenic risks to humans
(pp. 394e407), Vol. 63. Lyon: IARC.
Kedderis, G. L., & Held, S. D. (1996). Prediction of furan pharmaco-
kinetics from hepatocyte studies: comparison of bioactivation and
hepatic dosimetry in rats, mice and humans. Toxicology and
Applied Pharmacology, 140(1), 124e130.
Kedderis, G. L., & Ploch, S. A. (1999). The biochemical toxicology of
furan. CIIT (Chemical Industry Institute of Toxicology) Activities,
19(12), 1e10.
Kuballa, T., Stier, S., & Strichow, N. (2005). Furan concentrations in
coffee and coffee beverages. Deutsche Lebensmittel-Rundschau,
101(6), 229e235.
Lee, H., Bian, S. S., & Chen, Y. L. (1994). Genotoxicity of 1,3-dithiane
and 1,4-dithiane in the CHO/SCE assay and the Salmonella/
microsomal test. Mutation Research, 21(4), 213e218.
Maga, J. A. (1979). Furans in foods. Critical Reviews in Food Science
and Nutrition, 4, 355e400.
Mark, J., Pollien, P., Lindinger, C., Blank, I., & Mark, T. (2006).
Quantitation of furan and methylfuran formed in different precursor
systems by proton transfer reaction mass spectrometry. Journal of
Agricultural and Food Chemistry, 54(7), 2786e2793.
McGregor, D. B., Brown, A., Cattanach, P., Edwards, I., McBride, D.,
Riach, C., et al. (1988). Responses of the L5178Y tkþ/tk mouse
lymphoma cell forward mutation assay: mouse lymphoma cell
forward mutation assay. III.72 coded chemicals. Environmental and
Molecular Mutagenesis, 12, 85e154.
Merritt, C., Bazinet, M. L., Sullivan, J. H., & Robertson, D. H. (1963).
Mass spectrometric determination of the volatile components from
ground coffee. Journal of Agricultural and Food Chemistry, 11(2),
152e155.
Mugford, C. A., Carfagna, M. A., & Kedderis, G. L. (1997). Furan-
mediated uncoupling of hepatic oxidative phosphorylation in
Fischer-344 rats: an early event in cell death. Toxicology and
Applied Pharmacology, 144(1), 1e11.
NTP. (1993). Toxicology and carcinogenesis studies of furan (CAS No.
110-00-9) in F344/N rats and B6C3Fl mice (gavage studies). NTP
Technical Report No. 402. U.S. Department of Health and Human
Services, Public Health Service, National Institutes of Health,
Research Triangle Park, NC.
Perez Locas, C. P., & Yaylayan, V. A. (2004). Origin and mecha-
nistic pathways of formation of the parent furan e a food
toxicant. Journal of Agricultural and Food Chemistry, 52(22),
6830e6836.
Peterson, L. A., Cummings, M., Vu, C. C., & Matter, B. (2004). Iden-
tification of the urinary metabolites of furan. Chemical Research in
Toxicology, 17(12), 1764.
Peterson, L. A., Naruko, K. C., & Predecki, D. P. (2000). A reactive
metabolite of furan, cis-2-butene-1,4-dial, is mutagenic in the ames
assay. Chemical Research in Toxicology, 13(7), 531e534.
Ravindrananth, V., Boyd, M. R., & Burka, L. T. (1984). Reactive
metabolites from the bioactivation of toxic methylfurans. Science,
224(4651), 884e886.
Reinhard, H., Sager, F., Zimmermann, H., & Zoller, O. (2004). Furan in
foods on the Swiss market e method and results. Mitteilungen aus
Lebensmitteluntersuchung und Hygiene, 95, 532e535.
Senyuva, H. Z., & Gokmen, V. (2005). Analysis of furan in foods. Is
headspace sampling a fit-for-purpose technique? Food Additives
and Contaminants, 22(12), 1198e1202.
Shinoda, Y., Komura, H., Homma, S., & Murata, M. (2005). Browning
of model orange juice solution: factors affecting the formation of
decomposition products. Bioscience Biotechnology and
Biochemistry, 69(11), 2129e2137.
Shinoda, Y., Murata, M., Homma, S., & Komura, H. (2004). Browning
and decomposed products of model orange juice. Bioscience
Biotechnology and Biochemistry, 68(13), 529e536.
Stadler, R. (2006). Furan: summary of industry activities. Report of
a workshop held on Analytical methods and brainstorming on the
elements to be included in a database. Joint DG SANCO/EFSA/DG
JRC workshop, Brussels, 19th May 2006.
Vernin, G., Chakib, S., Rogachev, S. M., Obretenov, T. D., &
Parkanyi, C. (1997). Thermal decomposition of ascorbic acid.
Carbohydrate Research, 305(1), 1e15.
Yaylayan, V. A., Keyhani, A., & Wnorowski, A. (2000). Formation of
sugar-specific reactive intermediates from C-13-labeled L-serines.
Journal of Agricultural and Food Chemistry, 48(3), 636e641.
Yaylayan, V. A., Machiels, D., & Istasse, L. (2003). Thermal decom-
position of specifically phosphorylated D-glucoses and their role in
the control of the Maillard reaction. Journal of Agricultural and
Food Chemistry, 51(11), 3358e3366.
Yuan, J. P., & Chen, F. (1999). Simultaneous separation and determi-
nation of sugars, ascorbic acid and furanic compounds by
HPLC e dual detection. Food Chemistry, 64(3), 423e427.