Acrylamide in food: mechanisms of formation and influencing factors during heating of foods

Background: In April 2002, the Swedish National Food Administration and a scienti c group at the University of Stockholm jointly announced that they had shown acrylamide to be formed during the preparation of food and found it to occur in many foodstuffs. These new  ndings were clearly of concern to many types of industrial food processing as well as to home cooking. The Swedish Food Federation (Li) initiated and  nanced the formation of an expert committee to look into the chemical mechanisms. The present review is the  nal report of that expert committee. Design: The study identi ed, examined and put together facts and present knowledge on reaction routes for acrylamide formation in food and causal connections to cooking and food processing conditions. The results are based on literature surveys, examination of the analytical data published by the Swedish National Food Administration and other follow-up studies, contacts with international scienti c networks, and observations from food companies. Results: The exact chemical mechanism(s) for acrylamide formation in heated foods is unknown . Several plausible mechanistic routes may be suggested, involving reactions of carbohydrates, proteins:amino acids, lipids and probably also other food components as precursors. With the data and knowledge available today it is not possible to point out any speci c routes, or to exclude any possibilities. It is likely that a multitude of reaction mechanisms is involved. Acrolein is one strong precursor candidate, the origin of which could be lipids, carbohydrates or proteins:amino acids. Acrylamide is a reactive molecule and it can readily react with various other components in the food. The actual acrylamide level in a speci c food product, therefore, probably re ects the balance between ease of formation and potential for further reactions in that food matrix. There are indications in support of that the Maillard reaction being an important reaction route for acrylamide formation, but lipid degradation pathways to the formation of acrolein should also be considered. Conclusions: Reliable analytical methods to measure acrylamide in foods are available. Model studies are needed to identify precursors and reaction route(s) based on current hypotheses and to elucidate possible further reactions between acrylamide and other food components. Studies are needed to optimize formulation and processing conditions to minimize acrylamide levels, taking other product quality properties into consideration.

Acrylamide appears as a white crystalline solid, is odourless and has high solubility in water (2155 g l ¼ 1 water). Its melting point is 84.5°C, and its boiling point (25 mmH g) 125°C (192.6°C at atmospheric pressure).
Acrylamide is a reactive chemical, which is used as a monomer in the synthesis of polyacrylamides used, for example, in the puri cation of water and in the formulation of grouting agents. Acrylamide is known as a component in tobacco smoke.
Acrylamide is primarily reactive through its ethylenic double bond. Polymerization of acrylamide occurs through radical reactions with the double bond. Acrylamide could also react as an electrophile by 1,4-addition to nucleophiles, e.g. SH -or N H 2 -groups in biomolecules.
Acrylamide is metabolized in the body to glycidamide, a reactive compound formed through epoxidation of the double bond.
The toxicological effects of acrylamide have been studied in animal models. Exposure to acrylamide leads to D N A damage and at high doses neurological and reproductive effects have been observed. Carcinogenic action in rodents has been described, but carcinogenicity to humans has not been demonstrated in epidemiological studies, although it cannot be excluded. The International Agency for R esearch on Cancer (IAR C) has classi ed acrylamide as ''probably carcinogenic to humans'' (group 2A). Neurological effects have been observed in humans exposed to acrylamide. The properties, use and toxic effects of acrylamide are reviewed by IAR C (1) and the European U nion (EU ) (2).

Background: how was acrylamide formation in foods observed?
Reaction product fro m acrylamide observed in humans Compounds that are reactive and therefore short lived in the body can be demonstrated through their stable reaction products (adducts) with biomacromolecules, e.g. haemoglobin (H b) in blood. The adducts to H b are accumulated during the life span of the protein (about 4 months in humans). H b adducts are not indicators of toxic action but could be used for exposure measurements and calculation of intake.
A sensitive method for speci c measurement of adducts to the N -terminal valines in the globin chains in H b and for analysis by gas chromatography -mass spectrometry (G C-M S) has been devel-oped and applied to a wide range of compounds (3,4). Application of this methodology has shown that an adduct from acrylamide is formed by M ichael addition to the ethylenic double bond to N -terminal valine (5). This adduct, N -(2-carbamoylethyl)valine, has been measured in blood from acrylamide-exposed humans (5)(6)(7)(8) and animals (9). In studies of occupational exposure it has been shown that the adduct also occurs in blood from persons without known exposure (6,8), although at higher levels in smokers (since acrylamide occurs in cigarette smoke).
In connection with studies of the leakage of acrylamide at the H allandsa ûs tunnel construction in Sweden, calculations of uptake of acrylamide and evaluation of cancer risk were performed (10). The calculations of uptake (from pharmacokinetic modelling and reaction kinetics) showed that the average background adduct level in unexposed controls would correspond to a daily intake by adults of about 100 mg acrylamide, and there was a preliminary indication that this background level could be associated with a considerable cancer risk. The estimated risk seemed to be higher than the risk from ''background'' exposure of other reactive compounds detected as adducts in persons without known exposure. In this situation it appeared urgent to nd the source of the acrylamide adducts regularly observed in non-exposed persons. The occurrence of acrylamide in tobacco smoke and the ndings of lower background levels in wild animals (Tö rnqvist et al., to be published) led to the hypothesis that acrylamide was formed in cooking.

A crylamide fo rmation during cooking: identi cation and quanti cation
The identi cation of acrylamide in heated foodstuffs originated from a hypothesis for which both direct and indirect proof was obtained. An animal feeding experiment was performed to test the hypothesis on acrylamide formation during heating of foodstuffs. A strong increase in the acrylamide H b adduct level in blood from rats was observed when the animals were fed fried standard feed (11). In this context the identity of the observed H b adduct was veri ed by tandem mass spectrometry (G C-M S:M S) through comparison with isotopesubstituted standards (11).
A G C-M S method for analysis of acrylamide in water (based on bromination) was further developed for the analysis of acrylamide in the animal feed. The content of acrylamide in the fried feed was measured and found to be compatible with the increase in the acrylamide H b adduct level in the rats.
In following experiments the effect of heating (frying, etc.) on the content of acrylamide in different foodstuffs was investigated (12). The G C-M S method that had been applied in the studies of animal feed was further improved and simpli ed. This method, based on a well-known procedure for analysis of acrylamide in water, involves bromination of the ethylenic double bond, the dibromopropionamide formed being the analyte. The bromination is performed at ambient temperature and at pH 1 -3. The G C-M S analysis is performed at raised temperature. It was found desirable to con rm the results by a milder method for analysis of underivatized acrylamide. This was achieved by the development of a liquid chromatographytandem mass spectrometry (LC-M S:M S) method. The analytical results obtained with these two methods (G C-M S and LC-M S:M S) are in full agreement. Analysis at different conditions with the two methods further supported the conclusion that acrylamide is the analyte. It was shown that acrylamide was formed in a temperature-dependent manner in food. Low contents of acrylamide were found in heated protein-rich foods (5-50 mg kg ¼ 1 ) and high contents in carbohydrate-rich foods (100 -4000 mg kg ¼ 1 ), compared with undetectable levels in unheated or boiled foods.
The fact that in the above animal experiment the formed adduct levels tallied with the dietary intake of acrylamide as analysed in the heated feed, further supports the view that we are really dealing with acrylamide. Similar conclusions can be drawn from adduct levels in humans and human consumption on heated foods (12).
This work was carried out in a collaborative project between Stockholm University and Ana-lyCen N ordic AB, with development of methods and analysis of acrylamide in food in the latter laboratory. In view of the results obtained, showing high acrylamide contents in carbohydrate-rich commercial foods, the Swedish N ational F ood Administration, in parallel work, developed an LC-M S:M S method for acrylamide analysis in food (13). They essentially veri ed the results and extended the study to a broader range of foodstuffs (http:::www.slv.se), and also showed that there was good agreement between the analyses carried out at the two laboratories (13). The results from the studies were jointly announced in Stockholm on 24 April 2002.
Subsequent work verifying and extending the analyses of acrylamide in food, mostly analysed with LC-M S:M S, has been presented from several F ood Authorities and other organizations in different countries (e.g. the U K , N orway, the N etherlands, Switzerland, Germany and the U SA).
Chemical mechanisms for acrylamide formation F ood scientists and technologists have had an interest in acrylamide (and:or its derivatives, including polymers), its applications and its possible toxic effects for many years. F or example, there are many reports on can coatings and food packaging, food additives (preservatives, arti cial sweeteners, etc.) and acrylamide polymers of suitable quality with low residual acrylamide monomer levels that are used in, for example, the USA for treatment of poultry, potato, corn and other wastes, with the resulting concentrated solids used as components of blended animal feeds (14 -19).
There are only a few earlier reports on the occurrence of acrylamide in food. F or example, acrylamide has been reported to be present in plant material (potatoes, carrots, radish, lettuce, Chinese cabbage, parsley, onions, spinach and rice paddy) (20). In 1 g plant samples, 1.5 -100 ng acrylamide could be detected. Acrylamide was also reported to occur in sugar (21). The origin of the detected acrylamide in these foods is not known; it may be exogen ous.
To the authors' knowledge, no proposed or proven reaction routes for the formation of acrylamide during food processing have been published. Therefore, described below are the hypotheses that were found to be most relevant and probable in a food processing situation. Therefore, acrolein is a very probable precursor of acrylamide. Simple, fundamental chemical transformations (such as reaction with ammonia liberated from amino acids) can then convert acrolein (or a derivative of it) into acrylamide. The produc-tion of acrylamide through the reaction of acrolein with ammonia can be anticipated. (B) Alternative formation mechanisms of acrylamide do not necessarily involve acrolein. F or example, proteins and:or amino acids can, after a series of transformations, such as hydrolyses, rearrangements and decarboxylations, eventually lead to acrylamide.
Processes A and B are complicated and involve multistage reaction mechanisms which may also include free radical reactions to acrolein or acrylamide (23 -25).

A crolein fo rmation fro m lipids
When oil is heated at temperatures above the smoke point, glycerol is degraded to acrolein, the unpleasant acrid black and irritating smoke (26 -29). The formation of acrolein is known to increase with the increase in unsaturation in the oil and to lead to a lowering of the smoke point. The smoke point is higher for oils with a higher content of saturated fatty acids and lower content of polyunsaturated acids. The smoke points for some of the main oils and fats are: palm 240°C, peanut 220°C, olive 210°C, lard and copra 180°C, sun ower and soyabean 170°C, corn 160°C, margarine 150°C and butter 110°C. U sually, the smoke starts to appear on the surface of heated oils before their temperature reaches 175°C. The oil is rst hydrolysed into glycerol and fatty acids and then acrolein is produced by the elimination of water from glycerol by a heterolytic acidcatalysed carbonium ion mechanism followed by oxidation (30): Besides the above-mentioned mechanism for the formation of acrolein from acylglycerols, acrolein can also be produced as a result of oxidation of polyunsaturated fatty acids and their degradation products (31 -34). Several aldehydic products (including malondialdehyde, C3 -C10 straight chain aldehydes, and a,b-unsaturated aldehydes, such as 4-hydroxynonenal and acrolein) are known to form as secondary oxidation products of lipids (35). Acrolein was also found to form in vivo by the metal-catalysed oxidation of polyunsaturated fatty acids, including arachidonic acid (36).

A crolein fo rmation fro m amino acids, proteins and carbo hydrates
Several sources for the formation of acrolein are known. It may arise from degradation of amino acids and proteins (37,38), degradation of carbohydrates (39), and the M R between amino acids or proteins and carbohydrates (40,41). M any possible routes for the formation of this three-carbon aldehyde, taking the starting point from many different sugars or amino acids, may be proposed. Its formation from methionine by the Strecker degradation in the frame of the M R is one example. Alanine, with its three-carbon skeleton, has also been suggested as a possible source. H owever, ssion reactions of longer carbon chains are common and well known, so at present there is no basis to give priority to any speci c reaction routes.

Formation of acrylamide through amino acid reactions no t involving acrolein
There are also numerous, plausible reaction routes by which amino acids (or proteins) may form acrylamide without going through acrolein. Within the frame of complex, multistage reaction mechanisms, involving hydrolyses, rearrangements, decarboxylations, deaminations, etc., many speci c mechanistic pathways may be suggested. D ecarboxylation and deamination of aspargine, and transformations of dehydroalanine (formed from, e.g. serine or cysteine) are some examples of reaction routes that have been proposed. H owever, also in this case these can only be seen as possible examples, and similarly to above, there is no basis to give priority to any speci c routes.

C onclusion
Since no systematic studies have been performed or reported, there is at present no evidence to indicate any speci c reaction routes for acrylamide formation, or to exclude any possibilities. It is most likely that a multitude of reaction mechanisms is involved, depending on food composition and processing conditions.

Further reactions of formed acrolein and acrylamide
As mentioned above, acrolein can be converted into acrylamide by a series of fundamental reactions. H owever, both acrolein and acrylamide are reactive, because of their double bonds and the amino group of acrylamide. They can readily react further with other reactive groups present in the food matrix or formed during the heating process. F or example, acrylamide can react with small reactive molecules, such as urea [CO(N H 2 ) 2 ] and form-aldehyde (H CH O), or with glyoxal [(CH O) 2 ], aldehydes (R CH O), amines (R 2 N H ), thiols (R SH ), etc. F urthermore, the products shown in the following scheme can even react further in the same mode of reaction: These types of reactive functional group may also be found in macromolecules, such as proteins.
(See adduct formation with valine in the globin chain of haemoglobin, described above. In haemoglobin adducts are formed not only with valine, but also with, e.g. cystein.) The presence or absence of reactive groups (or its concentration) in the food matrix may thus be one explanation for differences in nal acrylamide content in different food systems. The resulting acrylamide level may be due to a balance between formation and further reactions. The low acrylamide levels in heated meat products could, for instance, depend on adduct formation between acrylamide (or acrolein) and proteins.

Factors with a possible in uence on acrylamide formation
A couple of different chemical mechanisms for the formation of acrylamide has been outlined above. As long as the mechanisms are not con rmed, the in uencing factors cannot be established. Thus, what is presented here are attempts to identify what factors would be of importance (regarding processing conditions or product composition) if a speci c reaction route were the prevailing one. Speci c emphasis is placed on the M R , since this reaction system involves many of the basic carbohydrate and amino acid reactions. Another major reaction in foods during processing, which could be of importance, is lipid hydrolysis followed by oxidation of the fatty acids.
A crolein fo rmation fr om lipids Acrolein may be formed from the glycerol part of triglycerides or through oxidation of fatty acids. This means that factors favouring lipid hydrolysis as well as factors favouring lipid oxidation would promote acrolein formation. Temperature is an important factor for both of these reactions. R egarding hydrolysis, pH may also be of importance and high as well as low pH may be supposed to favour acrolein formation. R egarding oxidation, lipid composition is of key importance: the higher the degree of unsaturation, the lower the stability. Protection against oxygen and light will limit the oxidation, and pro-oxidants, such as metals, should be avoided. The protective effect of antioxidants should also be taken into account.

The Maillard reaction as the route fo r acrylamide fo rmation
The M R has been proposed as a route for acrolein formation. The direct formation of acrylamide through amino acid transformations has also been proposed. These amino acid transformations also involve reactions common in the M R system.

Basics of the Maillard reaction
The M R is one of the most important chemical reactions in food processing, with an in uence on several aspects of food quality. F lavour, colour and nutritional value may be affected and certain reaction products have been noticed to be antioxidative, antimicrobial, genotoxic, etc. The practical applications of M aillard chemistry in food processing are, therefore, a matter of balance between favourable and unfavourable effects, and the aim of the food manufacturer is to nd an optimum in this balance. This may be accomplished by in uencing the main variables affecting the M R (42).
The M R takes place in three major stages and is dependent on factors such as concentrations of reactants and reactant type, pH , time, temperature and water activity. F ree radicals and antioxidants are also involved (43).
The early stage (step 1) involves the condensation of a free amino group (from free amino acids and:or proteins) with a reducing sugar to form Amadori or H eyns rearrangement products. The advanced stage (step 2) means degradation of the Amadori or H eyns rearrangement products via different alternative routes involvin g deoxyosones, ssion or Strecker degradation. A complex series of reactions including dehydration, elimination, cyclization, ssion and fragmentation results in a pool of avour intermediates and avour compounds. F ollowing the degradation pathway as illustrated schematically in F ig. 1, key intermediates and avour chemicals can be identi ed.
One of the most important pathways is the Strecker degradation, in which amino acids react with dicarbonyls (formed by the M R ) to generate a wealth of reactive intermediates. Typical Strecker degradation products are aldehydes, e.g. formaldehyde, acetaldehyde and possibly propenaldehyde (acrolein ). Strecker degradation results in degradation of amino acids to aldehydes, ammonia and carbon dioxide (44) and takes place in foods at higher concentrations of free amino acids and under more drastic reactions, e.g. at higher temperatures or under pressure (45).
The nal stage (stage 3) of the M R is characterized by the formation of brown nitrogenous polymers and co-polymers. While the development of colour is an important feature of the reaction, relatively little is known about the chemical nature of the compounds responsible. Colour compounds can be grouped into two general classes: low molecular weight colour compounds, which comprise two to four linked rings, and the melanoidins, which have much higher molecular weights.
Review of factors in uencing the Maillard reaction F actors that are particularly important for the M R are the starting reactants, e.g. type of sugar and amino acid (protein), temperature, time and water activity. The presence of metal salts (pro-oxidants), and inhibitors, such as antioxidants and sul te, may have an impact.
S tarting reactants: reducing sugar and amino acids: proteins. The M R requires reducing sugars, i.e. sugars containing keto-or aldehydes (free carbonyl groups). The reactivity of different sugars can be summarized in the following way (46): The shorter the carbon chain of the sugar, the greater the lysine losses (M R ). Pentoses are more reactive than hexoses and disaccharides in yielding brown colour. Aldoses are more reactive than ketoses both in aqueous solution model systems and at storage (low water content). Among isomeric sugars, stereochemistry is important. Thus, ribose is more reactive than xylose monitored as lysine losses.
All monosaccharides are reducing sugars. (Sugar alcohols do not participate in the M R .) Among the disaccharides all sugars except for sucrose are reducing sugars. In oligosaccharides and starch only the end-terminal monosaccharide is a reducing sugar. Starch and sugars, such as sucrose, lactose and maltose, can easily hydrolyse upon heating above 100°C at slightly acidic pH , resulting in the formation of monosaccharides (reducing sugars). Thus, thermal processing often results in a continuous supply of reducing sugar formed from complex carbohydrates.
M ost studies concerning the reactivity of amino acids have been performed on free amino acids in diluted aqueous solutions. The reactivity among the diamino acids increased with the length of the carbon chain. Among the amino acids studied lysine was most reactive. In proteins and peptides, only free amino groups can react, i.e. N -terminal a-amino groups and V-amino groups.
T emperature and time. The temperature dependence of chemical reactions is often expressed as the activation energy (E a ) in the Arrhenius equation. The higher the value of E a , the more temperature dependent the reaction rate. Activation energy data for the M R have been reported within a wide range, 10 -160 kJ mol ¼ 1 , depending on, among other things, water activity and pH and what effect of the reaction has been measured. The temperature dependence of the M R is also in uenced by the participating reactants. The temperature effect is also affected by the other variables, and different aspects of the M R thus differ in temperature dependence (42).
W ater. Water has both an inhibitory and an accelerating impact on the M R . Water acts partly as a reactant and partly as a solvent and transporting medium of reactants (reactant mobility). In the initial steps of the M R , 3 moles of water are formed per mole of carbohydrate. Thus, the reaction occurs less readily in foods with a high water activity (a w ) value. Water may depress the initial glucosylamine reaction, but enhance the deamination step later in the reaction.
The results from studies in model systems for optimal water concentration or a w (free water) or relative humidity (R H ) vary markedly depending on selected reactants and how the M R is evaluated: as loss in lysine or browning intensity. Several studies have been performed, of which most claim the maximum a w to be between 0.3 and 0.7 (47). However, most data on the in uence of a w are based on studies at relatively low temperatures (30 -60°C). At higher temperature, more relevant to heat processes, considerably lower a w has been shown to be favourable to the M R (42).
The main explanation for an optimum reaction rate at an intermediate a w is that the reactants are diluted at the higher a w , while at a lower a w the mobility of reactants is limited, despite their presence at increased concentrations.
pH . The M R itself has a strong in uence on pH. Therefore, aqueous model systems based on re ux boiling of sugars and amino acids need to be buffered since the pH quickly drops from 7 to 5. Low pH values ( B 7) favour the formation of furfurals (from Amadori rearrangement products), while the routes for reductones and ssion products are preferred at a high pH . H owever, the overall effect of pH is not clear cut, since the reactions take place by all three pathways. In unbuffered water solutions, pH decrease during the M R and buffering with alkali have a catalytic effect.
The reactivity of different amino acids at various pH has been studied. Browning of a glucose solution upon heating was obtained rst when pH exceeded 5 and it increased with increasing pH. The degree of browning varied with the position of the amino group. The function of pH is linked with speci c reaction steps of the M R . Initially, only non-protonized forms of amino acids can form Schiff base. This explains the pronounced change in reactivity (monitored as browning) that happens when pH passes the isoelectric point of the amino group in the reacting amino acid. Thus, optimal pH for the M R varies with the system used and how the reaction is monitored (e.g. lysine losses or browning).
Inhibition of the M aillard reaction. M easures to inhibit the M R in cases where it is undesirable involve lowering the pH value, maintenance of the lowest possible temperatures and avoidance of critical water contents (moistures below 30%, during processing and storage), use of non-reducing sugars and addition of sul te (45). The use of the inhibitor sulfur dioxide constitutes an important way of controlling the M R . It may combine with early intermediates. H owever, sul te only delays colour formation and it is interesting to note that the colour formed in sul te-treated systems is less red and more yellow than in untreated systems.

Maillard reactions and food processing
In exploiting the M R , the key target for the food industry is to understand and harness the reaction pathways enabling the improvement of existing products and the development of new products. While it would be easy to assume that this means the generation of avour and colour, not all M aillard products endow positive characteristics to foods and ingredients. The positive contributions of the M R are avour generation and colour development. The negative aspects are off-avour development, avour loss, discoloration, loss of nutritional value and formation of toxic M aillard reaction products (M R Ps). In applying the M R , there are challenges that are common to the food industry, independent of the type of the product. These challenges can be classi ed: maintenance of raw material quality, maintenance of controlled processes for food production, maintenance of product quality and extension of product shelf-life (42,43).
Flavour:arom a. The most common route for formation of avours via the M R comprises the interaction of a-dicarbonyl compounds (intermediate products in the M R , stage 2) with amino acids through the Strecker degradation reactions. Alkyl pyrazines and Strecker aldehydes belong to commonly found avour compounds from the M R . F or example, low levels of pyrazines are formed during the processing of potato akes when the temperature is less than 130°C, but increase 10-fold when the temperature is increased to 160°C, and decrease at 190°C, probably owing to evaporation or binding to macromolecules. The aroma pro le varies with the temperature and the time of heating. At any given temperature -time combination, a unique aroma, which is not likely to be produced at any other combination of heating conditions, is produced. Temperature also affects the development of aroma during extrusion cooking.
Colour. The coloured products of the M R are of two types: the high molecular weight macromolecule materials commonly referred to as the melanoidines, and the low molecular weight coloured compounds, containing two or three heterocyclic rings (48). Colour development increases with increasing temperature, time of heating, increasing pH and intermediate moisture content (a w ¾ 0.3 -0.7). In general, browning occurs slowly in dry systems at low temperatures and is relatively slow in high-moisture foods. Colour generation is enhanced at pH \ 7. Of the two starting reactants, the concentration of reducing sugar has the greatest impact on colour development. Of all the amino acids, lysine makes the largest contribution to colour formation and cysteine has the least effect on colour formation.
A ntiox idative capacit y. There are several reports on the formation of antioxidative M R Ps in food processing. The addition of amino acids or glucose to cookie dough has been shown to improve oxidative stability during the storage of the cookies. H eat treatment of milk products before spray drying has been reported to improve storage stability, as has heat treatment of cereals (42).
The antioxidant effect of the M R P has been extensively investigated (49). It has been reported that the intermediate reductone compounds of M R P could break the radical chain by donation of a hydrogen atom: M R P was also observed to have metal-chelating properties and retard lipid peroxidation. M elanoidines have also been reported to be powerful scavengers of reactive oxygen species (50). R ecently, it was suggested that the antioxidant activity of xylose-lysine M R Ps may be attributed to the combined effect of reducing power, hydrogen atom donation and scavenging of reactive oxygen species (51).

N utritive value.
Loss of protein quality is often associated with the M R , especially in cereal products and milk powder produced by heat treatment. U sually the essential amino acid having an extra free amino group, e.g. lysine, is most vulnerable. If the essential amino acid is also the nutritionally limiting amino acid, the in uence of the M R on the protein quality is substantial. This is not a problem in cooking meat and sh, since these food items are very rich in protein. Loss of protein quality in terms of nutritional value is a more serious problem for heat treatment and dehydration, especially of cereals, milk and their mixtures (breakfast cereals, gruels, bread, biscuits), since carbohydrates dominate over proteins in these food items and the proteins levels are also generally low.
T ox ic effects. The possibilities that M R Ps could be mutagenic and:or carcinogenic were explored with the Ames test around 20 -25 years ago. In general, weak genotoxicity:mutagenic activities were found for known M R Ps. M ost attention over the past few decades has been paid to the food mutagens found in the crust from cooked meat and sh. Chemically, these compounds belong to a class of heterocyclic amines, currently amounting to around 20 different species. M ost of them have been classi ed as possible food carcinogens (group 2B) by the IAR C, based on long-term studies on rodents. The precursors of the heterocyclic amines are free amino acids and, for more than half of the 20 species, also creatine (a natural energy metabolite present in muscle cells only). R educing sugars up to equimolar amounts compared with amino acids and:or creatine enhance the yields of heterocyclic amines markedly.
Thus, the M R and:or pyrolysis have been claimed to be important mechanisms for the formation of these heterocyclic amines, where Strecker aldehydes, pyrazines or pyridines and creatine have been suggested to play an important role. The yields of these food-borne carcinogens increase with time and temperature, especially from 150°C and above. The highest concentrations of heterocyclic amines are found in the crust of panfried, grilled or barbecued meat and sh. In addition, gravies prepared from dried meat juice collected from pan residues or oven roasting could be rich in heterocyclic amines. Pro-oxidants, water activity in the optimal range for the M R and high temperatures (200 -400°C) enhance their yield. The average daily exposure for heterocyclic amines is around 0.5 mg per person, with a range between 0 and 20 mg. Antioxidants, an excess of carbohydrates, cooking temperatures below 200°C and moisture contents above 30% reduce the occurrence of heterocyclic amines. M oreover, heterocyclic amines rarely occur in plant foods, even during well-done cooking (52).
To the authors' knowledge, no reports in the literature so far have studied acrylamide formation linked with the M R . In all studies single, randomly selected samples were analysed. The main original results reported by the Swedish N ational F ood Administration were con rmed by the later studies, but it must be concluded that substantial variations are found within a given food group and in cases when repeated analyses of the same product have been performed considerable variations have been found between samples. This makes the data premature as a basis for conclusions on the mechanisms for acrylamide formation. An attempt to summarize the results published so far is presented in Table 1. Table 1, presenting hitherto reported acrylamide data of heat-treated foods, can only give indications as to which factors are important in acrylamide formation. Table 1 indicates that high temperatures are needed for the acrylamide to form. N o acrylamide formation has so far been demonstrated at temperatures below 100°C and it is probable that the products reported have reached temperatures well above this level. There are several examples of exaggerated acrylamide formation on overheating. F ried products of plant origin seem to give the highest concentrations, but frying fat is not a prerequisite for acrylamide formation.

Conclusions and ideas from data and observations presented so far
The data indicate strongly that acrylamide formation is mainly a surface phenomenon. This has also been veri ed by other data presented to the authors by companies. This implies that water activity may be an important factor, although there are strong links between temperature and water activity in a frying or baking process. F rom the analyses reported it is very dif cult to draw conclusions on variations between different plant raw materials. Corn products are possibly lower in acrylamide content than comparable products made from potato or other cereals. H owever, it is dif cult to explain these possible differences or draw conclusions on reaction mechanisms, since detailed data on chemical composition (reducing sugars, speci c amino acids, etc.) are lacking. This is a general dif culty. There are insuf cient details on chemical composition to be able to suggest which are the important precursors and formation mechanisms.
Some data and observations, however, indicate that acrylamide formation is increased by increased concentration of (reducing) sugar in the raw materials or ingredients. This strongly supports the M R mechanisms. The M R hypothesis is also supported by several other observations, parallel to browning, the in uence of temperature and water activity, etc. In fried products the proposed route via acrolein formed from lipids should also be considered. The relatively high temperatures combined with low water activity which favour acrylamide formation, also favour free radical reactions. Thus, antioxidants and other free radical scavengers or quenchers could act as inhibitors.
The M R occurs wherever non-enzymic browning is induced by heat treatment, e.g. extrusion cooking, roasting, popping, baking, pan-frying, deep-fat frying, barbecuing and autoclavation. M ost unprocessed foods contain the necessary starter reactants, i.e. amino acids:proteins and reducing sugar. Conventionally cooked foods are subjected to a relatively high temperature for a relatively long time, and the surface of the food dries out to give a crust with a low a w , favouring the M R .
It is also clear from the composition of cereals and potato (see below), and most other raw food material, that they contain all necessary precursors, e.g. protein:free amino acids, carbohydrates:sugars and lipids, to initiate both M R s and lipid oxidation or other degradation routes during processing and storage. The extent to which these foods contain the optimal proportion of precursors and modifying components in terms of enhancers or inhibitors for acrylamide formation remains to be established.
In this context one must also bear in mind the high potential for acrylamides to be consumed through further reactions with other components in the food product. Similarly, acrolein or other precursors could react with other food components and take reaction routes not leading to acrylamide. Consequently, the nal acrylamide level in a food product may be due to the balance between formation and further reactions, controlled by the chemical composition of that speci c food. Low acrylamide levels may thus be a result of further reactions (or altered reaction routes) in that speci c food matrix. As already speculated, the low acrylamide levels demonstrated in meat products could be a result of adduct formation of acrylamide with proteins or other components. All of the reaction mechanisms mentioned so far (M R , lipid hydrolysis and oxidation, etc.) are also known to proceed in meat systems.

Raw material composition
The raw material studied so far has been mainly cereals and potato, and as shown in Table 1, heattreated products from these materials contain the highest concentrations of acrylamides, in several samples exceeding 500 mg kg ¼ 1 . These plant materials are storage organs containing large quantities of starch, protein and cell-wall materials as well as lipids, ash, polypheno ls and a large number of low molecular weight compounds such as sugars and free amino acids. It is well known that there exists a large variation in chemical composition in most plant materials. This variation is dependent on both genetic and environmental factors. The main mechanisms for acrylamide formation discussed so far are related to fat degradation and reactions involving sugars, amino acids and proteins, not the least the M R . Other components such as starch and dietary bre may be involved by modulating the processing conditions, such as the water activity. On a dry matter basis, wheat contains 60 -73% starch, 9 -16% crude protein, 9 -18% dietary bre, 2 -3% fat, 2 -5% sugars (glucose, fructose and fructo-oligosaccharides as well as maltose in germinated products) and 1 -2% ash. The crude protein contains all the common amino acids and also signi cant amounts of free amino acids. R ye and dehulled oats contain less starch but more dietary bre, especially water-soluble and viscous dietary bre. Oats also have a higher content of fat, which in this cereal is also present in the starchy endosperm. R ye has a higher content of sugars, especially fructo-oligosaccharides. Corn has higher starch and fat contents, but a lower content of dietary bre than wheat.

Analytical methods: can we trust the data?
The methods used to analyse acrylamide in foods were described brie y in the background section. G C-M S methods and LC-M S:M S methods have been used. Analyses have now been performed by a number of laboratories (using somewhat different methods). When the same type of product has been analysed in several countries, generally good agreement has been obtained between the results. G ood correlation has also been demonstrated when identical samples have been analysed by different methods at different laboratories (13). This strongly supports that the analytical methods are reliable, that it really is acrylamide that is measured, and that the reported data can be trusted. F urther support comes from the nding that H b adduct levels in animals tallied with aclylamide intake as analysed in the feed (11). Similar conclusions can be drawn from adduct levels in humans and human consumption of heated foods (12).
The same conclusion was reached by the F ood and Agriculture Organization:World H ealth Orga-nization (F AO:WH O) Consultation in G eneva, on 25 -27 June 2002. Their summary report states: ''Sensitive and reliable methods are available to identify and measure acrylamide in foodstuffs. The measurement uncertainty is small in relation to the between-sample and within-lot variability expected for acrylamide levels '' (http:::who.int:fsf).

Conclusions
The exact chemical mechanisms of acrylamide formation in heated foods are not known. Several plausible mechanistic routes may be suggested, involving reactions of carbohydrates, proteins:amino acids, lipids and probably also other food components as precursors. With the data and knowledge available today it is not possible to point out any speci c routes, or to exclude any possibilities. A multitude of reaction mechanisms is probably involved, depending on food composition and processing conditions. Acrolein is one strong precursor candidate. Current data indicate that the M R may be an important reaction route for acrylamide formation, but lipid degradation pathways to the formation of acrolein should also be considered.
Acrylamide is a reactive molecule and it can readily react with various other components in the food. The actual acrylamide level in a speci c food product therefore probably re ects the balance between ease of formation and the potential for further reactions in that food matrix.
M ore research is needed before any rm conclusions can be drawn concerning precursors, reaction route(s) and conditions for acrylamide formation in terms of reactants, time, temperature, pH , water activity, etc. M ore data regarding acrylamide levels in a broader range of food products are also needed. R eliable analytical methods to measure acrylamide in foods are available.

Research needs: suggestions for further studies
M odel studies, based on current hypotheses, to identify chemical mechanisms for acrylamide formation (precursors, reaction conditions, possible inhibitors, etc.); model studies to elucidate possible further reactions between acrylamide and other food components; kinetic studies of acrylamide formation in model systems; studies in foods:food models: in uence of processing parameters; in uence of ingredients:possible precursors; optimization of formulation and processing conditions to minimize acrylamide levels, taking into consideration other product quality properties; continued mapping of acrylamide and acrolein in different foods; development of simple methods to measure acrylamide in foods.
Could anything be done while awaiting the nal answers? The knowledge is still too limited to draw conclusions regarding cooking practices during industrial processing or food preparation at home. M ore answers from further research are needed. In the meantime the only obvious practical advice would be to avoid overheating. As long as the chemical mechanisms remain unknown, further practical recommendations are dif cult to make. H owever, there are some indications that the M R could be involved, suggesting that factors such as levels of free (reducing) sugars and amino compounds in raw materials and ingredients should be taken into consideration. Although there is little available information on the possible in uence of lipids, an interim piece of advice would be to be aware of this possibility and to control lipid degradation and oxidation in frying oils as far as possible, until more knowledge is obtained about this.

Postscript
D uring the publishing of this article the rst reports on the chemical mechanisms of acrylamide formation were presented. They all showed that acrylamide can be formed by a reaction between amino acid and sugar, and identi ed asparagine as a main amino acid.