Chapter 7

Review of methods of analysis

This review of analytical methods presents assessments of their applications, limitations and the resources required. The objective of the review is to provide guidance on the selection of compatible methods for the nutrients and some other constituents. The continuous developments in analytical chemistry make it almost impossible to ensure that the review is comprehensive and takes into account all recent developments. The review does not provide detailed analytical protocols; for these the reader needs to consult the relevant specialist texts.

In this review, for each nutrient (or group of nutrients), tables summarize the available methods. Estimates of capital costs have been given in three categories: low, where the method requires basic equipment that would usually be found in a laboratory; medium, where specialized instrumentation is required but normally costing less that US$5 000; high, indicating the need for specialized equipment usually costing more than US$10 000.

The proximate system of analysis

The proximate system for routine analysis of animal feedstuffs was devised in the mid-nineteenth century at the Weende Experiment Station in Germany (Henneberg and Stohmann, 1860, 1864). It was developed to provide a top level, very broad, classification of food components. The system consists of the analytical determinations of water (moisture), ash, crude fat (ether extract), crude protein and crude fibre. Nitrogen-free extract (NFE), more or less representing sugars and starches, is calculated by difference rather than measured by analysis.

Although some of the methods used historically in the proximate system of analysis are not recommended for the preparation of food composition databases (e.g. crude fibre), it is useful to consider the concepts involved as they have dominated views on the composition of foods and food analysis. This system was developed at a time when the chemistry of most food constituents was only partially understood, and the growth of nutritional sciences has shown that for nutritional studies a more detailed and biochemically oriented approach to food analysis is needed. Nevertheless, proximate analysis, including the original methods, still forms the basis for feed analysis, and the analysis of foods for legislative purposes in many countries.

Many people find the concept and term “proximates” useful to represent the gross components that make up foods; the actual analytical methods then become independent. Others believe that the definition of proximates is based on the original methods prescribed by Henneberg and Stohmann, and that method substitution, e.g. dietary fibre instead of crude fibre, negates the use of the term.

Water

Values for water remain an essential constituent in food composition databases because water content is one of the most variable components, especially in plant foods. This variability affects the composition of the food as a whole. The range of methods for water analysis is summarized in Table 7.1.

The methods are based on the direct or indirect measurement of water removed from the food, changes in physical properties that change systematically with water content, or the measurement of the chemical reactivity of water (Egan, Kirk and Sawyer, 1987; AOAC International, 2002; Sullivan and Carpenter, 1993; Southgate, 1999; Bradley, 1998).

For the majority of foods in food composition databases, drying methods are adequate; although slight methodological differences can be observed, these differences are rarely significant. The AOAC Official Methods recommend a lower drying temperature (70 °C) for plant foods to minimize the destruction of carbohydrates. Where this occurs it is usually better to use vacuum drying or freeze-drying.

Vacuum drying is most efficient if a slow leak of dry air is passed through the oven. This approach has the advantage that the analytical portions can be left unattended for long periods. Vacuum drying at 60–70 °C is preferable to drying in an air oven, particularly for foods that are rich in sugars. However, for most foods drying in an air oven is satisfactory for food composition database purposes.

Freeze-drying requires more capital investment but has the advantage that it dries the foods under mild conditions. Freeze-dried material is light, easily transported and can also be ground very easily. The process does, however, usually leave some residual moisture in the freeze-dried material, which must be removed to give values that are comparable with other drying methods.

Drying in a microwave oven is very quick but requires continuous surveillance to avoid charring. Drying with infrared lamps has been very successfully automated (Bradley, 1998). Both of these methods, however, are more suitable for routine quality control.

All the methods mentioned so far are unsuitable for foods with a high content of volatile components because these are driven off with the water. The Dean and Stark method can be used for such foods where a value for the moisture content is required. In this method the water is distilled off as an azeotropic mixture with an immiscible solvent such as toluene, xylene or tetrachloroethylene. The method is an AOAC-approved method for spices and cheese, and has achieved good levels of precision (AOAC International, 2002).

The Karl Fischer method is especially useful for foods with very low moisture content and for hygroscopic foods that are difficult to dry using conventional methods. The levels of accuracy achieved are rarely required for food composition databases.

The physical methods for measuring water content require expensive, highly specialized instrumentation and are most suitable where there is a very high throughput of similar samples.

Near infrared reflectance (NIR) methods, for example, have been widely applied for the analysis of cereal grains. The method requires calibrating with a large number of samples with moisture values measured by conventional methods to develop the analytical equations. Nuclear magnetic resonance (NMR), gas–liquid chromatography (GLC) and gas–solid chromatography (GSC) methods also require detailed calibration and are of greatest value in measuring the distribution of water in foods and identifying the forms of water in meats.

Nitrogen and nitrogenous constituents

Lakin's (1978) review still provides a comprehensive account of the analysis of nitrogen and nitrogenous constituents, and the methods are discussed briefly by Sullivan (1993) when reviewing the AOAC Official Methods, by Chang (1998) and by Southgate (1999). The range of methods is summarized in Table 7.2.

Total nitrogen

The proximate system, where “protein” is measured as total nitrogen multiplied by a specific factor, continues to dominate food composition studies. Most cited values for “protein” in food composition databases are in fact derived from total nitrogen or total organic nitrogen values. In the majority of cases, total nitrogen is measured using some version of the Kjeldahl (1883) method (which measures total organic nitrogen). In this method the organic matter is digested with hot concentrated sulphuric acid. A “catalyst mixture” is added to the acid to raise its boiling point, usually containing a true catalytic agent (mercury, copper or selenium) together with potassium sulphate. All organic nitrogen is converted to ammonia, which is usually measured by titration or, more rarely, colorimetrically. In the original method, a relatively large analytical portion (1–2 g) was used, but this requires large amounts of acid. Micro-Kjeldahl methods are much more commonly used as they produce a reduced amount of acid fumes and also require less acid and catalyst mixture. Environmental considerations exert considerable pressure to ensure the safe disposal of mercury and, especially, to minimize acid usage.

The micro methods can be automated at several levels (Egan, Kirk and Sawyer, 1987; Chang, 1998). Automation of the distillation and titration stages works well but automation of the digestion has proved quite difficult.

The Dumas method measures the total nitrogen as nitrogen gas after complete combustion of the food. Comparison of the results obtained with those obtained using the Kjeldahl method shows good agreement (King-Brink and Sebranek, 1993). The method has been successfully automated and, although the instrumentation is expensive, a high throughput of samples is possible, with good precision. The equipment uses very small analytical portions, and a finely divided analytical portion is essential.

NIR can also be used to measure nitrogen in some foods, although a large number of calibration samples is required.

Protein

Since the development of the proximate system of analysis, “crude protein” values have been calculated by multiplying the total nitrogen (N) by a certain factor. This factor was originally 6.25, based on the assumption that proteins contained 16 percent of N. It has been known for a considerable time that proteins of plant origin (and gelatin) contain more N and therefore require a lower factor. Jones, Munsey and Walker (1942) measured the nitrogen content of a wide range of isolated proteins and proposed a series of specific factors for different categories of food. These factors have been widely adopted and were used in the FAO/WHO (1973) review of protein requirements. These are listed in Table 7.3. Several authors have criticized the use of these traditional factors for individual foods (e.g. Tkachuk, 1969). Heidelbaugh et al. (1975) evaluated three different methods of calculation (use of the 6.25 factor, use of traditional factors and summation of amino acid data) and found variations of up to 40 percent. Sosulski and Imafidon (1990) produced a mean factor of 5.68 based on the study of the amino acid data and recommended the use of 5.70 as a factor for mixed foods.

In principle, it would be more appropriate to base estimates of protein on amino acid data (Southgate, 1974; Greenfield and Southgate, 1992; Salo-Väänänen and Koivistoinen, 1996) and these were incorporated in the consensus document from the Second International Food Data Base Conference held in Lahti, Finland, in 1995, on the definition of nutrients in food composition databases (Koivistoinen et al., 1996).

If these recommendations are to be adopted, the amino acid data should include values for free amino acids in addition to those for protein amino acids because they are nutritionally equivalent. The calculations require very sound amino acid values (measured on the food) as discussed below, and involve certain assumptions concerning the proportions of aspartic and glutamic acids present as the amides and correction for the water gained during hydrolysis. Clearly, this approach would not be very cost-effective when compared with the current approach.

At the present time it is probably reasonable to retain the current calculation method, recognizing that this gives conventional values for protein and that the values are not for true protein in the biochemical sense. However, it is important to recognize also that this method is not suitable for some foods that are rich in non-amino non-protein nitrogen, for example cartilaginous fish, many shellfish and crustaceans and, most notably, human breastmilk, which contains a substantial concentration of urea.

A number of direct methods for protein analysis have been developed for specific foods based on reactions involving specific functional groups of the amino acids present; these are thus not applicable to the measurement of proteins in general. Such methods include formol titration (Taylor, 1957) and the biuret reaction (Noll, Simmonds and Bushuk, 1974). A widely used group of colorimetric methods is based on reaction with Folin's reagent, one of the most widely used biochemically in the dairy industry (Lowry et al., 1951; Huang et al., 1976). These methods are most commonly calibrated with bovine serum albumin, which is available at high purity.

Dye-binding methods have been widely applied in the dairy industry (Udy, 1971); dye-binding can be made more sensitive by extracting the dye (McKnight, 1977), and the methods have been included in the AOAC Official Methods. Most of these methods depend on calibration against the Kjeldahl method. Pomeranz, Moore and Lai (1977) have published a comparison of biuret, NIR, dye-binding and alkaline distillation in the measurement of protein in barley and malt. Ribadeau-Dumas and Grappin (1989) have published a review of protein measurements in milk. In general, dye-binding methods have their widest application in the routine quality control of analysis of large numbers of similar types of sample (Van Camp and Huyghebaert, 1996).

Amino acids

Before the development of ion-exchange chromatography (IEC) individual amino acids were measured by colorimetric methods or by microbiological assay. Although these methods yielded acceptable results they have been almost completely superseded by chromatography procedures (Moore and Stein, 1948). These use automated systems that give complete analyses rapidly and with reasonable levels of precision.

The amino acids in the protein must first be released by hydrolysis and this constitutes the most critical stage of the analysis. Acid hydrolysis, usually with 6M HCl in an oxygen-free solution, gives complete release of most amino acids. Tryptophan is completely degraded in acid conditions and threonine, serine and the sulphur amino acids are partially degraded. Alternative hydrolysis conditions must therefore be used to measure tryptophan. Cystine and methionine are usually protected by specific oxidation before hydrolysis. Losses of threonine and serine are time-dependent and it is necessary to carry out serial hydrolyses to estimate the rate of degradation and correct the values accordingly. Conversely, the branched-chain amino acids are slowly released on hydrolysis, and serial hydrolyses are necessary to estimate complete release (Neitz, A., personal communication). Williams (1982) reviewed the development of IEC techniques and discusses the use of high-performance liquid chromato­graphy (HPLC) as an alternative.

The conditions for acid hydrolysis require pure acid and a high ratio of acid to analytical portions of the food. Even so, high-carbohydrate foods often react with the amino acids during hydrolysis, leading to losses that are difficult to quantify (Silvestre, 1997). Vapour phase hydrolysis has been suggested as an approach that minimizes the degradative losses. In this method the dried food (or protein) sample is hydrolysed by condensing acid. 6M HCl corresponds to the constant boiling mixture for the acid (De Geeter and Huyghebaert, 1992).

Sulphur amino acids are usually oxidized with performic acid before hydrolysis. Some chlorination of tyrosine can occur and the addition of phenol to the acid is often used to reduce this. The hydrolysis should be carried out under nitrogen or, preferably, in sealed tubes.

Hydrolysis must be carried out for three different time periods – 24, 38 and 48 hours – to allow correction for slow release and degradation losses. If pure bovine serum albumin is hydrolysed as a standard, this should also be hydrolysed for the same time periods.

Tryptophan is measured after alkaline hydrolysis (KOH, Ba(OH)2 or LiOH) (Landry and Delhave, 1993). It is usual to measure the leucine in the hydrolysate to adjust the values to be consistent with the acid hydrolysis. A number of alternative reagents and pre- and post-column derivatives have been used, but ninhydrin, despite its instability, is probably the most widely employed. Most other reagents vary in their sensitivity. Capillary gas chromatography has also been used, but most of the reagents vary in their rates of reaction with different amino acids.

In calculating the results of amino acid analyses it is important to express the amino acid values as mg amino acid per g nitrogen applied to the column. As a check on the analyses it is also important to calculate the recovery of nitrogen as amino acids and ammonia from the measured amino acids. There will usually be some losses during hydrolysis and the chromatography. If the losses are found to exceed 10 percent, repeating the hydrolysis should be considered.

Since 1990, HPLC methods of derivatized amino acids have replaced IEC for the analysis of protein hydrolysates in most laboratories as they offer reduced analysis time and improved limits of detection of about 1 picomole (pmol) (Cohen and Strydom 1988; Davey and Ersser 1990; Sarwar and Botting, 1993).

HPLC may be used to separate amino acids on ion-exchange columns with postcolumn derivatization with ninhydrin or OPA (o-phthaldialdehyde) (Ashworth, 1987) or by precolumn derivatization followed by separation on reversed-phase octyl- or octadecyl silica (Cohen and Strydom, 1988). For the analysis of amino acids in protein hydrolysates, reversed-phase HPLC with precolumn derivatization with PITC (phenylisothiocyanate) is becoming established as a cheaper alternative to commercial amino acid analyses using IEC. The PITC derivatization method enables the accurate determination of all nutritionally important amino acids except tryptophan in 12 minutes, while a liquid chromatographic method requiring no derivatization enables the determination of tryptophan in about eight minutes (Sarwar and Botting, 1993).

The range of methods is summarized in Table 7.4.

Available lysine 

Lysine can become nutritionally unavailable under certain conditions that lead to the e-amino group reacting with carbohydrate. This reaction reduces the biological value of the protein. Using the Carpenter method (1960) available lysine can be measured by its reaction with 2,4-fluorodinitrobenzene. This method has been the subject of many modifications (Williams, 1982). HPLC separation of e-DNP lysine is described by Peterson and Warthesen (1979).

Other nitrogenous substances

Several groups of foods, fish and other marine foods, meats, fungi and vegetables contain a range of nitrogenous materials, amines (Steadman, 1999) and nucleic acids. Many of these react with ninhydrin and can be separated by IEC. Methods for nucleic acids were reviewed by Munro and Fleck (1966). They may also be separated by HPLC and detected by their strong ultraviolet (UV) absorption.

Lipid constituents

FAO/WHO (1994) recommended that adequate food composition data on fats should be widely accessible and that standard methods and reference materials should be used for the analysis of fatty acids and preparation of nutrient databases. The report provides good coverage of the compounds and nutritional issues of interest. Christie (2003) is a key reference for lipid analysis.

In the proximate system of analysis, `fat' is measured as the fraction of the food that is soluble in lipid solvents. The extracted material contains a range of different classes of substances. For nutritional purposes the measurement of `total fat' has limited value; nevertheless, it still is widely reported and is retained in many requirements for food labelling and the regulation of food composition.

The range of methods is summarized in Table 7.5.

Total fat

The values obtained for total fat or total material soluble in lipid solvents are very method-dependent. Carpenter, Ngeh-Ngwainbi and Lee (1993), in their review for the AOAC of methods for nutritional labelling, set out the nature of the problems encountered. Gurr (1992) and Gurr, Harwood and Frayn (2002) discuss in detail the methods available for separating the different classes of lipids.

The classical method is based on continuous extraction performed on dried samples of food in a Soxhlet extractor, sometimes preceded by acid hydrolysis. This technique is time-consuming and subjects the extracted lipids to long periods at high temperatures. Its main drawback, however, is that it yields incomplete lipid extractions from many foods, especially baked products or those containing a considerable amount of structural fat. The extractant used is often petroleum spirit (which is less flammable that diethyl ether and less likely to form peroxides), which requires completely dry analytical portions and the removal of mono-and disaccharides. Values obtained using this method require close scrutiny before their inclusion in a database and their continued use is not recommended.

Other solvents, for example, trichloroethylene, are used in a number of automated systems of the `Foss-Let' type; these appear to give more complete extractions (Pettinati and Swift,1977).

The use of mixed polar and non-polar solvents has been shown to extract virtually all the lipids from most foods. In the case of baked (cereal) products, however, incomplete extraction of fat may occur. Chloroform–methanol extraction is well known (Folch, Lees and Stanley, 1957; Bligh and Dyer, 1959); this combines the tissue-penetrating capacity of alcohol with the fat-dissolving power of chloroform. The resultant extracts are complete but may also contain non-lipid materials and require re-extraction to eliminate these. This extraction method is preferred when the extract is to be subsequently measured for fatty acids and sterols (Shepherd, Hubbard and Prosser, 1974). The method is effective for composite foods and is included in the AOAC Official Methods. It has been shown to be useful for foods such as brain and egg that are rich in phospholipids (Hubbard et al., 1977). The measurement of lipids after acid (Weibull and Schmid methods) or alkaline (Röse-Gottlieb method) treatment also provides good extraction from many foods. These techniques are recognized as regulated methods by the AOAC and the European Union. Alkaline methods are almost exclusively used for dairy foods and are the approved method for such foods. The extracts from acid and alkaline treatments are not suitable for fatty acid analysis because some oxidation and losses analysis because some oxidation and losses due to (acid) hydrolysis of fats may occur. The AOAC has adopted methods for determining total fat (also saturated, unsaturated and monounsaturated fats) in foods using acid hydrolysis and capillary gas chromatography (Ngeh-Ngwainbi, Lin and Chandler, 1997; House, 1997) to comply with the Nutrition Labeling and Education Act (NLEA) definition of fat.

Lipid classes show strong carbonyl absorption bands in the infrared region. NIR has been used for legumes (Hunt et al., 1977a) and various other foodstuffs (Cronin and McKenzie, 1990). The effective use of the method depends on extensive calibration against comparable matrices using another approved method; for this reason the technique is most commonly applied in routine analyses of large numbers of very similar samples, for foods such as cereals and dairy products.

Triacylglycerols

Although it is probable that the composition of triacylglycerols (triglycerides) has nutritional significance, few databases contain compositional information. Methods for separating the individual components have not been extensively developed (Gurr, Harwood and Frayn, 2002). Thin-layer chromatography in combination with chromatography has been used. Total values can be found by separating the free fatty acids from the total lipid and can be used to give a “by difference” value. HPLC techniques have been proposed for the complete fraction of triacylglycerols (Patton, Fasulo and Robbins, 1990a,b; Gonzalez et al., 2001).

Fatty acids

Separation by GLC of the methyl esters of the fatty acids prepared by transmethylation of the lipid extracts from foods is the method of choice. The development of column packing materials, capillary techniques and detector amplification systems has extended to application of the method for the separation of isotopic forms and longer-chain fatty acids. The technique published by the International Union of Pure and Applied Chemistry (IUPAC) (Paquot and Hautfenne, 1987) forms the basic procedure. The exact method chosen will depend on the food to be analysed and the fatty acids of particular interest. Many users will be particularly interested in n-3 and n-6 fatty acids, trans acids and levels of long-chain fatty acids such as eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA). Automation of sample injection and the computerization of the chromatographs have added to the costs of the analytical apparatus but greatly improve accuracy, precision and analytical throughput. The American Oil Chemists' Society (1998) methods are: Method No. Ce 1-62 (packed column method for methyl esters of C9–C24 acids, and animal fats), Method No. Ce 1b-89 (capillary method for marine oils and for ethyl or methyl esters of C14–C24 acids (percentage relative values and mg/g levels of EPA and DHA), Method No. Ce 1c-89 (capillary method for fatty acids, trans isomers and cis, cis methylene-interrupted isomers in vegetable oils), Method No. Ce 1e-91 (capillary method for C4–C24 fatty acids), and Method No. Ce 1f-96 (capillary method for cis- and trans fatty acids in hydrogenated and refined oils and fats).

Infrared detectors are useful in the measurement of trans fatty acids (AOAC International, 2002). The major difficulty is the assignment of unequivocal identity to isomers. This requires good standards or combining the GLC separation with mass spectrometry (Beare-Rogers and Dieffenbacher, 1990), which may make it impractical for some developing countries.

Infrared absorption is currently the preferred method for the measurement of trans fatty acids in hydrogenated fish oils. GLC measurement of trans fatty acids in partially hydrogenated vegetable oils using a flame ionization detector (FID) often underestimates the trans fatty acid content, even on very long, highly polar, capillary columns (Aro et al., 1998).

Food composition laboratories lacking GLC instrumentation do not usually undertake fatty acid measurements but may seek cooperation with a laboratory with the necessary capital resources. The samples may be transferred to the laboratory as fat (which requires cold storage during transit and the addition of an antioxidant) or methyl esters (which also need to be protected against oxidation). It is important to verify these arrangements with the analysing laboratory to avoid interference by the antioxidants during chromatography.

The unsaturation of a fat can be estimated by iodine value determination (IUPAC, 1979; AOAC International, 2002); this remains a useful technique when full fatty acid analyses are not undertaken.

Sterols

In the past, nutritional analyses emphasized the measurement of cholesterol, but there is an increasing focus on the measurement of other sterols, especially phytosterols.

Cholesterol. The older techniques, using gravimetric and colorimetric methods, are now regarded as obsolete and are no longer used. The preferred methods are chromatographic, with widespread use of GLC of a range of derivatives separated on low-polarity columns (Punwar, 1975; Hubbard et al., 1977). One problem with sterol analysis in general is that the greater proportion of other lipids in most foods limits the application of the methods to the lipid extract directly.

Saponification is required before the preparation of derivatives. The use of trimethysilyl (TMS) derivatives met the standards required by the AOAC (Carpenter, Ngeh-Ngwainbi and Lee, 1993) for use with mixed foods. The procedures are somewhat complex and simplified methods have been proposed that require shorter sample preparation times (Thompson and Merola, 1993).

Improvements in the development of capillary GLC have provided the basis for developing procedures that do not require derivatization and that meet the appropriate standards (Jekel, Vaessen and Schothorst, 1998).

Other sterols. The method described above can also be used for the separation and measurement of the range of phytosterols found in the diet (Jonker et al., 1985), as can derivatization with TMS (Phillips, Tarrogo-Trani and Stewart, 1999).

Phospholipids

A comprehensive review of phospholipids published in 1973 (Ansell, Hawthorne and Dawkins.) summarized the analytical procedures available. Subsequently, HPLC techniques were developed (Hammond, 1982; Patton, Fasulo and Robbins, 1990a,b) and are now the methods of choice. Gunstone, Harwood and Padley (1994) provide an overview of methods for measuring the range of phospholipids.

Carbohydrates

The range of carbohydrates found in the human diet (see Table 4.3) illustrates the nature of the task facing the analyst who wishes to follow the recommendations published by FAO/WHO (1998) for measuring the carbohydrates in foods separately. Not all types of carbohydrates are, of course, present in all types of foods.

The distinctive metabolic and physiological properties of the different carbohydrates emphasize the fact that for nutritional purposes it is inadequate to consider the carbohydrates as a single component of foods.

The calculation of “carbohydrate by difference” using the Weende proximate system of analysis described at the beginning of the chapter was a reflection of the state of knowledge of carbohydrate chemistry at the time. Moreover, the system was designed for animal feedstuffs, especially for ruminants, and most of the carbohydrates (except lignin-cellulose of which crude fibre was an approximate measure) would therefore be digested in the rumen.

For nutritional purposes carbohydrates can be considered as falling into three groups based on the degree of polymerization:

• sugars (mono- and disaccharides);
• oligosaccharides (polymers of three to nine monosaccharide or uronic acid units);
• polysaccharides (polymers containing more that nine units), which fall into two broad categories: a-glucans (starches, starch hydrolysis products and glycogen) and a much more diverse group of non-a-glucans (non-starch polysaccharides [NSPs], which are the major constituents of dietary fibre).

These broad chemical groupings do not correspond precisely with physiological properties or with analytical fractions. The analyst faced with the analysis of carbohydrates, particularly NSPs, is “bound to make a compromise between the ideal of separating the many components and measuring them or a scheme that is entirely empirically based” (Southgate, 1969). In many cases, a food contains a limited range of carbohydrates and simpler procedures can be used for its analysis (Southgate, 1991).

The range of methods is summarized in Tables 7.6 to 7.8.

Sugars

A range of methods can be used for the analysis of the free sugars in foods; the choice depends primarily on the qualitative composition of the free sugars present in the food. Where a single carbohydrate species is present virtually any procedure can be used, but most foods contain a mixture of three or more components and separation of the components is required to produce accurate results. Specific enzymatic methods are available for the analysis of certain common mixtures without separation.

The methods for free sugars (and uronic acids) provide the end-analytical methods for most of the higher carbohydrate polymers after hydrolysis and separation of the components.

The evolution of the methods closely parallels the development of analytical techniques coupled with the pressures of the demands for analytical results. Thus the physical techniques were initially developed for the analysis of sucrose solutions in the sugar-refining industry. The reducing sugar methods were also developed for this industry and the methods were refined and their protocols codified under the auspices of the International Commission for Unified Methods of Sugar Analysis (ICUMSA, 1982). These methods still give satisfactory results providing the protocols are followed closely.

Colorimetric techniques were developed later, with the advent of improved methods for assessing optical density (although early measurement involved the visual matching of solutions). The range of chromogenic reagents for different monosaccharide classes and uronic acids mostly involve reactions in concentrated acids although colorimetric methods are based on reductiometric methods and a few on other reactions (Hudson et al., 1976). The methods are not especially robust, but on simple sugar mixtures with proper quality control they give sound values. The methods are not truly specific and this limits their use for the analysis of mixtures (Hudson and Bailey, 1980).

Specific enzyme methods have been developed, the most notable being the glucose-oxidase method, which has a colorimetric end-point. A series of coupled reactions with NADPH–NADP using specific enzymes permits the analysis of mixtures of glucose/fructose and glucose/fructose/sucrose and maltose/galactose (Southgate, 1991).

Chromatography, initially on paper or silica plates, provided good separations and semi-quantitative methods, but ion-exchange techniques were difficult to develop.

Gas chromatographic analysis depended on the preparation of suitable volatile derivatives. Initially trimethyl-silylation provided suitable derivatives for the analysis of sugar mixtures, although the chromatograms were very complex. The most widely used and powerful method for the analysis of mixtures involves the reduction of the monosaccharides to the alditols and acetylation.

HPLC columns are now available that give good separation of sugar mixtures without the need for the preparation of derivatives. The first detectors used refractive indices to measure the eluted peaks, but these are relatively insensitive and have been superseded by the pulsed amperometric detector, which has improved sensitivity.

Polyols (sugar alcohols)

Polyols are not widely found in foods. Some can be measured by specific enzyme methods although HPLC methods are more commonly used.

Oligosaccharides

These are widely distributed, especially in vegetables, and the malto- series is found particularly in foods that have partial starch hydrolysates and glucose-syrup preparations as ingredients. The malto-oligosaccharides are hydrolysed by brush-border enzymes and are “glycemic carbohydrates” that need to be measured separately.

Fructo-oligosaccharides are increasingly used as ingredients and should be measured after hydrolysis with specific fructan hydrolases. The remaining galacto-oligosaccharides should also be measured after specific enzymatic hydrolysis. GLC and, particularly HPLC separation techniques also offer powerful methods for the analysis of these oligosaccharides (Quigley, Hudson an Englyst, 1997).

Polysaccharides

These are best considered, for nutritional purposes, under two headings – starch and non-starch polysaccharides (NSPs).

Starch. This category includes all the a-glucans, starches, partially hydrolysed starches and glycogen. The latter is a minor component of most animal products; it is found in significant concentrations in fresh liver and horse flesh and as traces in lean muscle.

Polarimetric methods are limited to some cereals, but with proper calibration and standardization can give satisfactory results (Fraser, Brendon-Bravo and Holmes, 1956; Southgate, 1991).

Dilute acid hydrolysis can be used for highly refined foods with low concentrations of NSPs, and virtually any monosaccharide method can be used to measure the glucose produced.

The use of a glucose-specific method such as glucose-oxidase extends the range of foods for which this method is useful (Dean, 1978; Southgate, 1991).

Enzymatic hydrolysis with specific amylolytic enzymes, followed by precipitation of the residual NSPs with ethanol, and measurement of the glucose produced, is the most satisfactory and widely applicable method. The choice of enzymes and the conditions for hydrolysis are critically important. If values for total starch are required, any enzymatically resistant starch must be treated with alkali or dimethyl sulphoxide (DMSO) before hydrolysis (Southgate, 1991).

Resistant starch. Although enzymatically resistant starch was first observed analytically, the current view is that it should be defined as resistant physiologically, that is, resistant to hydrolysis in the human gastrointestinal tract (Gudmand-Hoyer, 1991). Englyst, Kingman and Cummings (1992) have distinguished three types of resistance, due to physical enclosure of starch, starch granule structure, and retrogradation. The latter type is more common in processed foods. The most common approach is to measure starch before and after treatment with DMSO.

Rate of digestion. Englyst and his coworkers (1999) have proposed that the rate of digestion of starch is the major determinant of variations in the glycemic responses to food and proposed that the starch can be considered to fall within three classes: rapidly digestible starch, slowly digestible starch and resistant starch. While the rate can be distinguished in vivo, simulation analytically is quite difficult. Collaborative studies have shown that reasonable precision can be obtained (Champ, 1992).

Glycemic index. There has been great interest in including glycemic index (GI) values in food composition databases and a set of tables of GI values has been published (Foster-Powell and Miller, 1995). The GI values (strictly speaking, a ranking of the carbohydrates in foods) are based on their glycemic effect compared with that of a standard food. The GI is defined as “the incremental area under the blood glucose response curve expressed as a percentage of the response to the same amount of carbohydrates from a standard food taken by the same subject” (FAO/WHO, 1998). The standard food is usually white bread or glucose. FAO/WHO (1998) have published a protocol using six or more subjects and define carbohydrate as “glycemic (available) carbohydrate”. A working definition used by the main Australian laboratory measuring GI defines carbohydrate as “total carbohydrate by difference minus the sum of dietary fibre plus resistant starch (if known) or the sum of starch plus sugars, including polyols and other slowly absorbable sugar derivatives” (Brand-Miller and Holt, personal communication).

In Australia, the use of a GI symbol on food labels is permitted and a Web site is available for consultation (http://www.glycemicindex.com). The GI of meals can be calculated but not of cooked recipe foods because the GI of a food is affected by cooking and processing.

Estimates of the different rates of digestion of starch in foods show some correlation with glycemic indices measured in vivo. These require a number of human subjects to have blood glucose levels measured at intervals for three hours after consumption of a fixed amount (50 g) of glycemic carbohydrates. The area under the curve is compared with the area under the curve for a 50 g glucose load or, better, 50 g of glycemic carbohydrates from white bread. White bread is preferred because glucose loads can be emptied slowly from the stomach because of osmotic effects. An interlaboratory study (Wolever et al., 2003) showed that within-subject variation in glycemic response needs to be reduced to improve precision of the method.

An in vitro method for rapidly available glucose published by Englyst et al. (1999) showed a high correlation with glycemic response.

Non-starch polysaccharides. Methods for NSP analysis involve treatment of the sample to remove free sugars and starch by enzymatic hydrolysis. The unchanged NSPs are recovered by precipitation with ethanol (80 percent v/v), then washed and dried. The NSPs are hydrolysed using one of two methods: sequentially with dilute acid, which hydrolyses most of the non-cellulosic polysaccharides (NCPs), and with 12M H2SO4, which hydrolyses the cellulose; or the NSPs are hydrolysed completely using 12M acid (see “Measurement of NSPs” below for further details).

The monosaccharides are analysed by GLC after derivatization (as the alditol acetates [Englyst, Wiggins and Cummings, 1982]) or by HPLC, or as a total colorimetrically (Englyst, Quigley and Hudson, 1994). The methods are not very robust (Southgate, 1995), although collaborative trials have shown that when careful attention is given to the protocol the methods give reasonable precision.

Choice of method for carbohydrates

There is no single method that meets the recommendations of the FAO/WHO (1998) review. Ideally, when planning to measure the carbohydrates in foods one should aim to measure the different carbohydrate species in a food sequentially using one analytical portion; this approach avoids the possibility of double measurement of an overlapping fraction.

The basic principles of such an approach are set out schematically in Figure 7.1.

Figure 7.1 Principles of measuring carbohydrates and dietary fibre

Note: v/v =by volume; NSP= non-starch polysaccharides.

Extraction of free sugars, polyols and oligosaccharides. This could be done with an aqueous extraction, but this procedure will extract proteins with the result that subsequent analysis is more complex. The removal of fat is desirable for technical reasons, as this facilitates a more complete extraction of sugars. Extraction with aqueous alcohol is the most common approach: 80 percent v/v aqueous ethanol is most commonly used, but 85 percent v/v methanol is also useful, as is isopropanol. The extractions are usually made with a boiling solvent; care should therefore be taken to protect the analysts from solvent fumes. If the extract is likely to be acid it is important to neutralize the acid to avoid hydrolysis of di- and higher saccharides.

The aqueous alcohols will also extract some lower polysaccharides – short-chain polysaccharides as defined by Englyst and Hudson (1996). These should preferably be measured after selective enzymatic hydrolysis. Modern enzyme technologies have produced a wide range of very specific enzymes with high activity; many companies specialize in this area, for example, Boehringer Mannheim, Germany; Megazyme, Ireland; Nova, Denmark; and Sigma, the United States. Several of these companies prepare enzyme method “kits”. The rate of development of enzyme technology is such that it is expected that selective enzymatic hydrolysis will become increasingly important analytically because of the specificity offered (McCleary and Prosky, 2001).

Starch hydrolysis. The next stage is to remove starch using selective enzymatic hydrolysis. A number of enzymes can be used for this purpose. A mixture of amylase and pullulanase has been used to give complete hydrolysis to glucose but many glucoamylases give virtually complete hydrolysis to glucose. The conditions for enzymatic hydrolysis are critical, both to ensure complete and rapid starch hydrolysis and to minimize hydrolysis of NSPs, especially b-glucans. Unhydrolysed NSPs are recovered by precipitation with ethanol to 80 percent v/v.

Measurement of NSPs. The precipitated NSPs are washed and dried gently and then hydrolysed. This may be done in boiling 1M H2SO4 followed by hydrolysis in 12M acid at ambient temperature. This produces, first, a hydrolysate containing the monosaccharides from the NCPs and, second, the monosaccharides from a cellulosic fraction. Alternatively, the NSPs may be hydrolysed in 12M acid followed by the dilute acid, which produces a hydrolysate containing the monosaccharides from the NSPs as a whole. Uronic acids are not hydrolysed completely by these methods, and colorimetric analysis is widely used (Englyst, Quigley and Hudson, 1994). Specific enzymatic hydrolysis of the uronic acid containing polymers is now possible (Quigley and Englyst, 1994).

Dietary fibre

Dietary fibre should be considered as part of the carbohydrates in foods. The major problem in the choice of method lies in the definition of dietary fibre and its interpretation in an analytical context. The term was first used in 1953, by Hipsley, to describe the sum of the

hemicelluloses, cellulose and lignin in food, in other words the components of plant cell walls in foods. Trowell, in 1972, took up the term for “the indigestible components of the plant cell wall in foods”. Both these terms were too vague to use as a basis for an analytical method and in 1976 Trowell et al. (1976) proposed that it be defined as “the sum of the plant polysaccharides and lignin that are not digested by the enzymes of the gastrointestinal tract”. This was closely analogous to the “unavailable carbohydrates” as defined by McCance and Lawrence (1929) and measurable by the procedures proposed by Southgate (1969).

In this method the aim was to measure the carbohydrates specifically using colorimetric techniques. Englyst developed this approach using the more specific GLC methods, which gave values for the non-starch polysaccharides and incorporated a stage to convert resistant starch to non-enzymatically resistant starch. The procedure was developed in a series of collaborative studies and the most recent protocols are described by Englyst, Quigley and Hudson (1994) and Southgate (1995). This method measures only the NSPs and does not include lignin.

In other parts of Europe, especially Sweden and Switzerland, and in the United States, the focus was directed at the “indigestibility of the polysaccharides and lignin”. A gravimetric method was developed where the residue after starch removal is weighed to give a measure of total dietary fibre (TDF); this has evolved into the Official AOAC Method No. 982.29 (Prosky et al., 1992). The method requires correction of the residue for undigested protein and for mineral contamination; total nitrogen and ash in the residue are measured and deducted to give the TDF values. These include lignin, resistant starch and all other indigestible carbohydrates (Guillon et al., 1998). A modification has been introduced to include the measurement of indigestible oligosaccharides.

The Englyst NSP and the AOAC TDF procedures are not very robust, especially where low levels are present (Southgate, 1995). The NSP method uses analytical portions of 100–200 mg and the preparation and homogeneity of these portions is absolutely critical. The mixing procedures also require close attention during the execution of the method.

The AOAC gravimetric procedure requires great skill when measuring low levels but gives good precision with high-fibre foods such as bran and wholemeal products. The residue also includes heat-induced artefacts.

In many countries, the choice of method for nutrition labelling will be defined by legislation. Nutritionally specific measurement of the different carbohydrate fractions is the preferred approach. The measurement of soluble and insoluble fractions is highly method-dependent and the FAO/WHO (1998) review concluded that there was no physiological justification for recording separate values based on solubility.

It is important to recognize that the hypothesis concerning the protective effects of dietary fibre was based on differences between diets (Burkitt and Trowell, 1975), i.e. it was a statement about the protective effects of diets that were rich in foods containing plant cell walls in a relatively unprocessed state. These diets are rich in many other components in addition to dietary fibre.

Alcohol

The classical method for measuring the alcoholic content of beverages is distillation of the de-gassed beverage and measurement of the specific gravity of the distillate. While this is still a valid and precise approach, measurement by GLC (which is simpler and quicker) or, alternatively, a specific enzyme procedure using alcohol dehydrogenase (Bergmeyer, 1974) are preferred methods as the distillation methods can be interfered with by other volatile constituents.

Organic acids

A variety of specific enzyme methods for different organic acids (Bergmeyer, 1974) remain valid, but these approaches have been superseded by HPLC methods (Wills et al., 1983). In a food product that contains acetic acid, simple acid-base titration can be used (Sadler and Murphy, 1998).

Inorganic constituents

The majority of methods for inorganic constituents require the organic matrix of the foods to be removed, or extraction and concentration, before they can be applied. Destruction of the food matrix removes a large number of potential sources of interference and provides the inorganic material in a concentrated form. In classical food analysis the organic matrix was incinerated (usually in a muffle furnace at a controlled temperature) and the resultant inorganic residue was weighed to give a value for ash in the proximate system of analysis. The organic matrix can also be destroyed by being heated in concentrated acids. This procedure minimizes losses during the oxidation and avoids any reaction between the inorganic constituents and the vessel used for dry incinerations.

Once the organic matrix has been removed the inorganic constituents can be measured using a variety of techniques. These include classical gravimetric or volumetric methods, polarimetry, ion-selective electrodes, colorimetric procedures (which may or may not be highly specific) and instrumental methods (which offer an increase in speed of analysis, automation and good precision). Many of the instrumental methods can be used for analysis of a number of constituents. In using these methods it is important to ensure that interference from other constituents is eliminated and it is essential to use standard reference (or in-house reference) materials with a similar matrix and apply other quality control measures. This approach is of fundamental importance in the measurement of trace inorganic constituents.

Total ash

Nutritionally, there is little value in recording ash values other than to provide an approximate estimate of the total inorganic material and to check for replication in the destruction of the matrix. A value for total ash is, of course, essential when it is necessary to calculate carbohydrate “by difference”.

In dry ashing, the food is incinerated in a crucible, usually made of silica, although porcelain (can be used but less suitable) or platinum (very expensive but the least reactive) can be used. The food matrix must be destroyed by heating gently at first to char the sample and then at 500 °C in a muffle furnace (Wills, Balmer and Greenfield, 1980) to prevent foaming of lipids (and sugars) until a white (or light grey) residue is produced. Heating above 500 °C can result in the loss of alkali metals. The general procedure is described by Osborne and Voogt (1978) and in the AOAC Official Methods (see Sullivan and Carpenter, 1993).

In the case of “wet ashing” acid digestion, the food sample is heated with acid – usually a mixture of nitric and sulphuric acids. Perchloric acid is often included in the digesting acid mixture although this introduces the risk of explosion and the procedure must be carried out in a fume hood designed for the use of perchloric acid. Wet ashing offers the advantage that no reactions with the crucible can occur that can lead to the formation of insoluble silicates. Digestion can be carried out in a Kjeldahl flask but this requires a larger quantity of acid. Particularly for trace element analysis, digestion is best carried out in a sealed container. Tubes designed for this purpose are available from most laboratory suppliers. They are made from resistant glass and have a cap with a plastic insert to provide an inert gas-tight seal. The analytical portion and the acid are placed in the tube, which is then capped and may be heated in a conventional or microwave oven. The tube is then allowed to cool completely before the gases are released with care.

For trace element analyses, the acids used must be of the highest analytical quality; blanks should be run as a matter of course and digestion of the reference materials should be included.

The most widely used instruments are atomic absorption spectrophotometers, which are suitable for the analysis of most cations of nutritional interest. The more simple flame photometers can be used for the analysis of Na and K.

Plasma emission instruments such as inductively coupled plasma spectrometers are available that permit the analysis of a wide range of elements and have the capacity to handle a large number of samples and analytes (McKinstry, Indyl and Kim, 1999). They do, however, require high initial capital expenditure and routine maintenance. Ihnat (1982;1984) provides a detailed review of the application of these methods to foods. Sullivan (1993) discusses the use of these techniques in the AOAC's Methods of analysis for nutrition labeling (Sullivan and Carpenter, 1993).

Preparation of analytical portion. The residues from dry ashing are usually dissolved in dilute acid and made to volume before analysis. The solutions from wet ashing usually need dilution to a suitable volume before analysis.

Tables 7.9 and 7.10 show methods of analysis for cations and anions, respectively, in foods.

Cations

Sodium and potassium. Flame photometry and atomic absorption spectrophotometry (AAS) are the preferred techniques. Mutual interference can occur and interference from phosphorus has been observed. These can usually be overcome by the application of appropriate standards.

Calcium. Flame photometry and AAS techniques have similar sensitivities. Interference from phosphorus can occur but this can be suppressed by the addition of lanthanum salts or by the use of N2O flames. Compleximetric titrimetric methods have been used and classical gravimetric methods can be used with foods rich in calcium.

Magnesium. AAS is the method of choice as this offers greater sensitivity than other procedures, with the exception of activation analysis.

Iron. This can be measured by AAS or inductively coupled plasma spectroscopy (ICP) instrumentally. There are, however, sound colorimetric methods available.

Zinc. While colorimetric methods are available, AAS or ICP are the better techniques to use.

Selenium. Hydride generation AAS has been widely used and is probably the method of choice at the present time (Foster and Sumar, 1996; Murphy and Cashman, 2001). Cathodic stripping voltammetry has also been proposed as a method (Inam and Somer, 2000).

Copper and other trace elements. These can be measured satisfactorily by AAS but may require the use of special conditions. ICP, when available, is a satisfactory technique (Coni et al., 1994). Colorimetric methods for copper are quite sound (Sullivan and Carpenter, 1993).

Anions

Phosphorus. This can be measured by ICP but a well-established colorimetric method is the preferred method when applied to wet-digested samples (Fiske and Subbarow, 1925). If dry­ashed samples are used, the pyrophosphates formed during ashing must be hydrolysed.

Chloride. A range of methods can be used. Ion-specific electrode analysis represents the simplest approach, but the classical reaction by titration is also satisfactory (Cotlove, Trantham and Bowman, 1958). Procedures using automated conductimetry also seem to perform well (Silva et al., 1999).

Iodine. This is regarded as one of the most difficult inorganic elements to measure. Dry ashing followed by titration or GLC has been used by the AOAC (Sullivan and Carpenter, 1993). Ion-specific electrodes offer some potential.

Fluorine. Polarographic methods have been developed that produce a very good sensitivity (Guanghan et al., 1999). Methods using ion-selective electrodes also seem to perform well (Kjellevold-Malde, Bjorvatn and Julshamn, 2001).

Sulphur. Sulphur may be measured via conversion to barium sulphate (Paul and Southgate, 1978) or by x-ray fluorescence (Isherwood and King, 1976).

Nitrate and nitrite. Methods include colorimetry (AOAC, 1980), HPLC (Wooton et al., 1985) and capillary ion electrophoresis. Ion-specific electrodes can also be used (Marshall and Trenerry, 1996).

Vitamins

“Vitamin” is a physiological term rather than a chemical term, expressing a certain physiological activity that is related to the chemical substances responsible for this activity. Vitamin activity may be due to a group of chemical compounds, usually related structurally to one another (vitamers).

The analysis of vitamins presents a number of challenges to the analyst and considerable analytical activity has been, and still is, directed at achieving the ideal analytical method for providing chemical values that predict the physiological vitamin activity for human beings in the current context. The ideal method would measure the different vitamers separately so that a value could be calculated for the total vitamin activity (Brubacher, Müller-Mulot and Southgate, 1985). This ideal is rarely possible, in part because of the presence of interfering substances without vitamin activity.

The discussion of methods for individual vitamins will emphasize the handling and preparation of samples for analysis; these are crucial factors because of the lability of some vitamins. Many vitamins are sensitive to light and some can be oxidized very rapidly. Heating can increase the rate of oxidation and may also lead to isomerization to inactive forms; unnecessary heating should therefore be avoided.

A number of detailed reviews on the analysis of vitamins in foods are available (Bates, 2000; Eitenmiller and Landen 1998; Machlin, 1984; Christie and Wiggins, 1978; Van Niekirk, 1982). Brubacher, Müller-Mulot and Southgate (1985) was the result of a collaborative European project which tried to establish a handbook of tested methods. A review of the AOAC Official Methods of vitamins is given by Sullivan and Carpenter (1993). Table 7.11 summarizes the methods for lipid(fat)-soluble vitamins and Table 7.12 summarizes those for the water-soluble vitamins.

Lipid-soluble vitamins

These are the vitamins A, D, E and K, and the carotenoids with provitamin A activity. As nutrition interest is now also focused on the non-provitamin A carotenoids, it is also desirable to cover more of these carotenoids.

Figure 7.2 Structures of the main vitamin A-active retinoids

to light and all preparations of analytical portions must be carried out in subdued lighting, preferably gold lighting. The food samples are saponified in alcoholic potassium hydroxide with the addition of an antioxidant, ascorbic acid, butylated hydroxytoluene (BHT) or pyrogallol. The vitamins are extracted into a suitable organic solvent. The extract is evaporated with additional BHT at a controlled temperature. Both normal-phase and reversed-phase HPLC can be used for the separation. In normal-phase separations measurement is usually by fluorescence; in reversed-phase separations UV detection and measurement is preferred. Standards should be followed throughout the entire sample preparation and analysis and must be controlled regularly for purity (Brubacher, Müller-Mulot and Southgate, 1985).

Nutritional interest originally focused on the carotenoids that demonstrated provitamin A activity, that is, were converted in the body to vitamin A. These are b-carotene, g-carotene, a-carotene and b-cryptoxanthin (Figure 7.3). During the 1990s it was recognized that many other carotenes are biologically active as antioxidants and therefore this review is concerned with methods that permit the measurement of a wider range of carotenoids. There are some 600 carotenoid isomers (Bauemefeind, 1972), but many of these have restricted occurrence or are present in minor amounts in most common foods. Debate about how to present different carotenes and their relative activity in databases continues.

The classical method was to perform a simple chromatographic separation of the carotenes as a group, and measure spectrophotometrically against a common b-carotene standard (Brubacher, Müller-Mulot and Southgate, 1985). This has been replaced by more detailed separation using ion-exchange columns and HPLC. The conditions applied in saponification are critical and need to be carefully controlled using standard mixtures. If this is done, then comparable values can be obtained (Mangels et al., 1993) with sufficient confidence to construct a database for the provitamin carotenoids (Chug-Ahuja et al., 1993).

HPLC is now the most widely used and preferred method. Scott (1992) and his colleagues (Scott and Hart, 1993; Scott et al., 1996), as part of an EU project to develop a SRM mixture of carotenoids, made an extensive series of studies on the various stages of the saponification extraction and HPLC analyses. Other analysts have also carried out detailed studies of the method (Wills and Rangga, 1996; Taungbodhitham et al., 1998). These studies provide the basis for obtaining sound analytical values for the most important carotenoids. A revised system for evaluating published carotene values taking into account these studies has been proposed and the production of quality codes is now being evaluated.

Vitamin D. Two forms of vitamin D are found in foods, cholecalciferol (D3) and ergocalciferol (D2). One IU is equivalent to 0.025 m g of cholecalciferol or ergocalciferol. Vitamin D3 is the more widely distributed (e.g. in fish oils, many fatty fish tissues, eggs, butter and cream cheese), and D2 occurs naturally in low concentrations in fish oils and mushrooms, and is the form used in fortification. Some meats contain 25-hydroxy-cholecalciferol in concentrations that contribute to vitamin D activity and need to be considered. Figure 7.4 summarizes the structures of vitamin D. Estimates of the relative activities of cholecalciferol, ergocalciferol and their metabolites vary. The convention appears to be to attribute a factor of five times the activity of cholecalciferol to 25-hydroxycholecalciferol (Chan et al., 1995, 1996). Therefore values for different forms should always be presented separately in analytical reports and reference databases.

Figure 7.3 Structures of the main vitamin A-active carotenoids

Vitamin D in foods is found at a very low concentration, which makes its analysis difficult. The original methods were biological using chicks or young rats (e.g. Method No. 936.14 [AOAC International, 1995]). These methods are difficult to perform and had generally low precision. The major problem with vitamin D analysis is that most food sources contain other lipids that tend to interfere (Ball, 1998).

Figure 7.4 Structures of the main compounds in foods with vitamin D activity

Gas chromatography is discussed by Koshy (1982), but HPLC is now the preferred technique and several methods have been published (cholecalciferol and 25-hydroxy-cholecalciferol in egg yolk [Mattila et al., 1993], ergocalciferol and 25-hydroxyergocalciferol in edible mushrooms [Mattila et al., 1994], and cholecalciferol, ergocalciferol and their 25 hydroxy metabolites in milk and meats [Mattila et al., 1995]). Similar methods (unpublished) were used for meats in the United Kingdom food composition tables (Chan et al., 1995, 1996) (V. Grace, UK Food Standards Agency, personal communication). The most useful method available involves a preliminary semi-preparative HPLC stage that eliminates much of the interference from other lipids. The food sample is saponified in alcoholic potassium hydroxide under nitrogen, with an antioxidant, ascorbic acid, hydroquinone, pyrogallol or BHT having been added before the saponification solution. The unsaponified lipids are extracted with a suitable organic solvent. An internal standard of the form of vitamin D not present in the sample is used. The unsaponified lipids are concentrated by rotary evaporation at low temperature. The extract is dissolved in the mobile phase of the semi-preparative HPLC. The conditions are carefully controlled to give a precise collection of the vitamin D.

The analytical separation may be carried out on normal or reversed-phase HPLC with UV detection. Reversed-phase is recommended for the analytical separation after normal-phase for the semi-preparation stage.

25-hydroxycholecalciferol can be measured by HPLC, as mentioned above (MAFF, 1997), but radio-immunoassay is probably the best choice at the present time where the necessary funds and equipment are available (Bates, 2000).

Vitamin E. Vitamin E activity is exhibited naturally by eight substances structurally based on tocopherols and tocotrienols (see Figure 7.5). Each vitamer has a different vitamin activity compared with a-tocopherol, which is seen as the primary structure. The preferred analytical method is therefore one that separates and measures all the different vitamers.

Figure 7.5 Structures of the main compounds with vitamin E activity

The food samples are saponified using alcoholic potassium hydroxide. The vitamin E vitamers are susceptible to oxidation at higher temperatures in alkaline conditions and should be protected by saponifying under nitrogen with the addition of antioxidants. The saponification conditions are similar to those used for vitamins A and D.

A colorimetric method, the Emmerie–Engel reaction with the reduction of ferric chloride and reaction with a, a'-dipyridine or 4,7-diphenanthroline, is also available. The complexes are rather unstable and give a total tocopherol value. The colorimetric method has been superseded by, first, GLC and, then, HPLC, which is now the preferred method.

Both normal-phase and reversed-phase HPLC can be used, although the normal-phase represents the better approach and separates all the vitamers. Detection uses fluorescence (Piironen et al., 1984, 1987). External standards are used and these need to be checked spectrophotometrically.

Vitamin K. Vitamin K activity is possessed by phylloquinone (K1), the menaquinones (K2 group) and menadione (synthetic K3). The structures are shown in Figure 7.6.

Figure 7.6 Structures of the main natural compounds with vitamin K activity

Vitamin K is sensitive to alkali and UV radiation and the appropriate precautions need to be taken during analytical operations. Colorimetric procedures are available, but these lack specificity and have been replaced as the methods of choice. Most analytical attention has been given to the measurement of vitamin K1. One major problem in the analysis is the presence of lipid, which must be removed by digestion with lipase before extraction with hexane (Indyk and Woollard, 1997). The solvent is evaporated under a stream of nitrogen and the residue dissolved in methanol, which is applied to a reversed-phase HPLC column. The eluate is reduced post-column with zinc and the fluorescence is then measured.

Semi-preparative separations have been used after digestions (Cook et al., 1999) and dual electrode detection systems have also been proposed (Piironen and Koivu, 2000). Most authors comment on the great variability of the values obtained and emphasize the need for proper repeat sampling and replication of analyses (Piironen et al., 1997; Jakob and Elmadfa, 1996).

Water-soluble vitamins

These include vitamin C and a number of vitamins of the B-group. The study of vitamin C has a long history (Carpenter, 1986) and this vitamin is discussed first.

Vitamin C. Two substances show vitamin C activity, L-ascorbic acid and the first product of its oxidation – L-dehydroascorbic acid (Figure 7.7). The D-isomer (erythorbic acid), which is used as an antioxidant food additive, is not active. Ascorbic acid is a powerful reducing agent which is oxidized very quickly, especially at raised temperatures and in alkaline solutions. During the preparation of food samples for analysis it is especially important to minimize the losses due to oxidation (Brubacher, Müller-Mulot and Southgate, 1985).

In most fresh foods the amounts of dehydroascorbic acid are very low and for many purposes the measurement of ascorbic acid alone may be adequate. Thus, the reduction of 2,6-dichlorophenolindophenol is the simplest and most reliable method (AOAC Method Nos 967.21 and 985.33 [Sullivan and Carpenter, 1993]).

Figure 7.7 Structures of the common compounds with vitamin C activity

The colorimetric method of Roe and Kuether (1943) involving the reaction with 2,4-dinitrophenyl hydrazine measures both ascorbic and dehydroascorbic acid.

The method of Deutsch and Weeks (1965) also measures both active forms fluorimetrically, after oxidation, and is recognized as an Official Method by the AOAC, both as originally described, and in a semi-automated version (Method Nos 984.26 and 967.22 [Sullivan and Carpenter, 1993]). Where the presence of erythorbic acid is not suspected, the fluorimetric method is probably the preferred method. HPLC techniques developed in the 1980s (Finley and Duang, 1981; Rose and Nahrwold, 1981; Keating and Haddad, 1982; Wimalasiri and Wills, 1983) for the separate measurement of ascorbic, dehydroascorbic and erythorbic acids are now widely used and give satisfactory performance (Schüep and Keck, 1990).

B-vitamins. This group includes a number of structurally distinct vitamins that were initially grouped together because they were water-soluble. The initial approach to the measurement of these vitamins, some of which are present at very low concentrations, was selective microbiological methods (Bell, 1974; Ball, 1994), and for some vitamins, total folates and vitamin B12, microbiological assays remain the only practicable methods. For the remaining B-vitamins, more specific chemical procedures, especially HPLC, have been developed and collaboratively tested.

Thiamin. The structures of the substances showing thiamin activity (B1) are shown in Figure 7.8. Thiamin is sensitive to heat and alkaline conditions and appropriate precautions must be undertaken during its analysis. Thiamin can be measured microbiologically using Lactobacillus viridescens or L. fermentum, but most analyses are based on its oxidation to thiochrome, which can be measured directly fluorimetrically. This is most conveniently carried out in conjunction with HPLC separation of interfering compounds. Thiamin, riboflavin and vitamin B6 are present in foods as enzyme cofactors combined with phosphate and must therefore be hydrolysed and treated with phosphatase before analysis. In early descriptions of the methods for these vitamins different conditions were used, but a number of collaborative studies (van den Berg et al., 1996; Ndaw et al., 2000) have shown that a common method for preparing the food samples can be used.

Figure 7.8 Structures of thiamin (vitamin B1)

The food sample is hydrolysed with acid and then treated with takadiastase or a phosphatase. Some authors use a ion-exchange pre-column (Bognar, 1981). The extract is then oxidized with potassium ferricyanate to form the thiochrome; it is then analysed using a reversed-phase HPLC column and the thiochrome is measured fluorimetrically. The analyses are controlled using an external standard. A post-column oxidation can also be used. In the large collaborative study reported by van den Berg et al. (1996) variations between the different practices in a range of laboratories did not affect overall performance of the method. Microbiological results also showed good agreement with the results from the HPLC methods.

Riboflavin. The structure of riboflavin (vitamin B2) is shown in Figure 7.9. It is found in foods as the free riboflavin or riboflavin-5'-phosphate (FMN) and as flavin adenine dinucleotide (FAD). The vitamin is very sensitive to light and UV radiation but relatively stable to heat and atmospheric oxygen. The analytical operations must therefore be carried out under conditions that minimize the exposure to light. The vitamin must be extracted from foods by treatment with acid and a suitable phosphatase enzyme. The riboflavin can be measured directly using fluorimetric methods, although many foods contain interfering substances and separation from these by HPLC is the preferred approach (Wimalasiri and Wills, 1985; Schüep and Steiner, 1988; Arella et al., 1996). Reversed-phase HPLC separation using fluorescence detection is the method most commonly used. In the collaborative study reported by van den Berg et al. (1996) minor variations in local methods did not affect performance. Microbiological assay using Saccharomyces carlsbergensis and S. uvarum tended to give slightly higher results than the HPLC, as observed previously by Hollman et al. (1993).

Niacin. Niacin activity is due to nicotinic acid and nicotinamide (Figure 7.10). Both forms are stable to atmospheric oxygen, light and heat in the dry state and in aqueous solution. A number of bound forms have been found in cereals that are extractable by alkali but these are probably not bioavailable. Tryptophan is also metabolized to niacin and the total niacin activity must include the contribution from tryptophan (Paul, 1969).

Niacin can be measured microbiologically with Lactobacillus plantarum (AOAC Method Nos 960.46, 944.13 and 985.34 [Sullivan and Carpenter, 1993]). Colorimetric methods based on the Konig reaction using oxidation with cyanogen bromide and reaction with p-amino-benzoyl-diethylaminoethanol have also been used (AOAC Method Nos 961.14, 981.16 and 975.41 [Sullivan and Carpenter, 1993]), but the toxic nature of cyanogen bromide makes it difficult to recommend these for routine use.

An HPLC method has been proposed and seems to perform reasonably well (Finglas and Faulks, 1987). After acid hydrolysis the food sample is filtered, treated with alkali, autoclaved and microfiltered before reversed-phase HPLC and fluorescence detection. A simplified extraction protocol has been proposed (Lahély, Bergaentzlé and Hasselmann, 1999) and has been shown to perform well with a range of foods (Rose-Sallin et al., 2001).

Figure 7.9 Structures of riboflavin (vitamin B2)

Figure 7.10 Structures of niacin and niacinamide {vitamin B3)

Vitamin B6. There are five compounds showing vitamin B6 activity whose structures are shown in Figure 7.11: pyridoxamine, pyridoxine, pyridoxal and the corresponding phosphate esters.

Vitamin B6 activity cannot therefore be measured using a method for a single substance. Microbiological assay using Saccharomyces carlsbergensis provides a measure of total activity (AOAC Method Nos 960.46, 961.15 and 985.32 [Sullivan and Carpenter, 1993]). The assay is carried out after an acid hydrolysis and hydrolysis of the phosphates enzymatically, and the same extraction procedures as for thiamin and riboflavin can be used (van den Berg et al., 1996; Ndaw et al., 2000). The acid hydrolysis also hydrolyses glycosides, which are present in plant foods and which may or may not be bioavailable to humans.

Comparison of HPLC and microbiological assay has indicated that further work is required (van den Berg et al., 1996; Bergaentzlé et al., 1995). Ndaw et al. (2000) used an extraction procedure without the acid hydrolysis stage and the HPLC method of Schüep and Steiner (1988) and the procedure performed well with standard materials.

Vitamin B12. A group of complex structures possesses vitamin B12 activity (Figure 7.12). Classically it has been measured microbiologically with Lactobacillus leichmanii.

The levels of vitamin B12 in foods are very low and it is extracted with hot water or a buffer in the presence of potassium cyanide, which converts the vitamin into the cyano form (AOAC Method Nos 960.46, 952.20 and 986.23 [Sullivan and Carpenter, 1993]).

A number of sensitive methods have been developed for clinical use (Bates, 1997; 2000) using competitive protein binding and a range of radio-immunoassays, but these have not been evaluated in a range of foods.

Folates. The folates comprise a group of compounds related to folic acid (pteroyl-glutamic acid). Folic acid does not occur naturally in foods but is widely used in food fortification or as a supplement. Most of the naturally occurring folates are derivatives of 5,6,7,8-tetrahydrofolic acids and exist in the monoglutamate or polyglutamate forms. Their structures are summarized in Figure 7.13.

The biological activity of the forms differs and the ideal analytical nutritional procedure therefore should involve the measurement of the different vitamers.

Total folate values are best measured by microbiological assay using Lactobacillus rhamnosis (caseii). Most organisms cannot use the polyglutamate forms, and deconjugation with a suitable enzyme (hog kidney, chicken pancreas, human plasma) is a preliminary stage in the analysis. The extraction is carried out in the presence of ascorbic acid to minimize oxidation. The extract is treated with a combination of protease, lipase and amylolytic enzymes, which improve the efficiency of extraction. The different conjugase enzymes give similar performances. At one time it was assumed that the measurement of folate before and after deconjugation would give values for “free” folate and total folates. The organisms respond to varying extents to the glutamate derivatives and the concept is flawed. The conditions for the microbiological assay were studied by Phillips and Wright (1982, 1983), Wright and Phillips (1985) and Shrestha, Arcot and Paterson (2000); these procedures give satisfactory quantitation.

Figure 7.11 Structures of the most common compounds with vitamin B6 activity

Figure 7.12 Structures of vitamin B₁₂ and analogues

Source: Modified, with oermission, form Brown, G.M. & Reynold, J.J., Annual Review of Biochemistry, 32: 419-62. © 1963 by Annual Reviews Inc.; reproduced with permission from Shils, M.E. & Young, V. (1988) Modern nutrition in health and disease. 7th ed. Philadelphia, PA, USA, Lea & Febiger.

Figure 7.13 Structures of folacin (folates)

Separation of the different folate vitamers using HPLC techniques is now widely used (Finglas et al., 1999) and some databases give values. Intercomparison studies have shown that values for 5-methyl tetra-hydrofolate showed reasonable agreement, but the agreement with other vitamers was not satisfactory (Vahteristo et al., 1996). Subsequent studies on the standardization of the HPLC methods have shown that while it is possible to measure the 5-methyl form with reasonable confidence, the other vitamers are still not measured properly by existing methods that use fluorimetric detection. A kit is available for folic acid and an evaluation has been published by Arcot, Shrestha and Gusanov (2002).

Pantothenic acid. The structure of pantothenic acid is given in Figure 7.14. Pantothenic acid in the free form is unstable and extremely hygroscopic. It is usually present bound to proteins or in the form of salts. Only the dextro- form is active. The classical method is microbiological using Lactobacillus plantarum as the test organism (Bell, 1974; AOAC Method Nos 960.46 and 945.74 [Sullivan and Carpenter, 1993]). The food is extracted with water and where the food is rich in fats these are best removed before analysis. The aqueous extract is usually autoclaved and the pH adjusted with acid and alkali to around 6.8. The mixture, after incubation overnight, is heat-treated to stop growth and growth is measured turbidometrically.

Biotin. Biotin is found in foods as the free vitamin and bound to protein. Figure 7.15 shows the structure of the vitamin. The classical method is microbiological using Lactobacillus plantarum (Bell, 1974; AOAC Method No. 960.46 [Sullivan and Carpenter, 1993]). An HPLC method has also been described (Lahély et al., 1999). Preliminary extraction with acid followed by papain treatment is required to extract the vitamin from the food. The HPLC

Figure 7.14 Structure of pantothenic acid

Figure 7.15 Structure of biotin

method uses a reversed-phase separation, post-column derivatization with avidin-fluorescence 5-isocyanate and fluorescence detection.

Radio-assays using the specific binding protein have also been described (Bates, 2000).

Bioactive food components

Pennington (2002) has published a comprehensive review of food composition databases for bioactive food components, including flavonoids, tannins, allyl sulphides, capsaicin, indoles, lignans, monoterpenes, phenolic acids, plant sterols and probiotics, categorized by food and by compound, and available as an annotated bibliography of over 400 pages on individual components (Pennington, 2001). Given the number and diversity of these components, it is not possible to review the methods for all of them (Speijers and van Egmond, 1999). This section therefore focuses on methods for measuring flavonoids, isoflavonoids, lignans and total antioxidant activity in view of the fact that these have been the subject of much interest in recent years. Methods for plant sterols were reviewed earlier in this chapter.

Flavonoids. A rapid method based on reversed-phase HPLC with UV detection was developed by Hertog, Hollmann and Venema (1992) for the quantitative determination of five major flavonoid aglycones (quercetin, kaempferol, myricetin, luteolin and apigenin) in freeze-dried vegetables and fruits, after acid hydrolysis of the parent glycosides. More recently Merken and Beecher (2000) published a gradient HPLC method with photodiode array detection for 17 prominent monomeric flavonoid aglycones representing all of the five common classes of flavonoids.

Phytoestrogens. The main plant compounds with known or suspected estrogenic activity are lignans, isoflavones, coumestans and resorcyclic acid lactones (Price and Fenwick, 1985). The modes of estrogenic action are discussed by Clarke et al. (1996). The major isoflavonoids are genistein, daidzein, formononetin, biochanin A and glycitein. Genistein, daidzein and glycitein occur in foods as their glycosides, all of which are biologically inactive. The free aglycones are formed by metabolic action of the human gut microflora, although this hydrolysis varies considerably from person to person (Xu et al., 1994). The total bioactivity is represented by the analysis of aglycones; however, this potential activity is represented by analysis of the conjugates and aglycones separately. The most active plant estrogen known is coumestrol (a coumestan); zearalenone is a potent resorcyclic acid lactone formed as a secondary metabolite of fungal species, mainly Fusarium (and is thus regarded as a contaminant). The lignans matairesinol, secoisolariciresinol, pinoresinol and isolariciresinol are potent phytoestrogens and are precursors of the mammalian lignans, enterolactone and enterodiol.

Given the very large number of plant compounds with estrogenic activity and the question of whether to analyse both the conjugates and the free forms or only the aglycones (after hydrolysis), many methods of analysis are in existence and there is little agreement on which method is best. No method is available to separate and quantify all bound and free compounds of interest in this category. Probably the most comprehensive method for the aglycones is the isotope dilution gas-chromatographic–mass spectrometric method of Adlercreutz and coworkers (Mazur et al., 1996), which analyses daidzein, genistein, biochanin A, formononetin, coumestrol, secoisolariciresinol and matairesinol, but not glycitein, as silyl derivatives. The method is expensive and needs access to mass spectrometry (MS). Another comprehensive method for foods that analyses daidzein, genistein, biochanin A, formononetin, coumestrol, secoisolariciresinol and matairesinol, but not glycitein, uses an HPLC-MS method originally developed for plasma and urine (Horn-Ross et al., 2000; Coward et al., 1996; Horn-Ross et al,. 1997; Barnes et al., 1998).

Isoflavones and coumestrol. For the USDA–Iowa State University Isoflavones Database (2002), the reference method adopted was the linear gradient method of Murphy et al. (1997), which separates daidzein, genistein, glycitein and their conjugates in soy-based infant formulas. Hutabarat, Greenfield and Mulholland (2000) have published a rigorously validated isocratic HPLC method for genistein, daidzein, formononetin, biochanin A and coumestrol (but not glycitein), while King and Bignell (2000) have published an HPLC method for daidzin, genistin, glycitin and their aglycones. A collaborative trial published by Klump et al. (2001) led to a recommendation to adopt as first action AOAC Method No. 2001.10 for the determination of isoflavones in soy and selected foods containing soy. This method uses reversed-phase liquid chromatography to separate and measure genistein, glycitein and daidzein and their glucosides, and also produces values for total isoflavones expressed as aglycones.

Lignans. Meagher et al. (1999) measured isolariciresinol, pinoresinol, secoisolariciresinol and matairesinol using HPLC with photodiode array detection, and Liggins, Grimwood and Bingham (2000) have published a GC-MS method for the determination of matairesinol, secoisolariciresinol and shonanin in foods as trimethylsylyl derivatives.

Total antioxidant activity. There is growing interest in ways to represent the total antioxidant activity of foods. A number of methods have been used but no standards exist and at this stage the inclusion of values for total antioxidant activity in foods in databases is not recommended. The topic is fully reviewed by Frankel and Meyer (2000).

Food composition data

Energy

The gross energy content of a food may be determined experimentally with a bomb calorimeter (Brown, Faulks and Livesey, 1993). An adiabatic bomb calorimeter is preferred for precise measurements, but the ballistic bomb calorimeter (Miller and Payne, 1959) gives a precision that is adequate for most nutritional studies. The values obtained using an adiabatic bomb calorimeter are corrected for the heat generated from the oxidation of nitrogen and sulphur in the food. The calorimeters are usually calibrated using benzoic acid as a thermo-chemical standard.

The values obtained are the gross heats of combustion and are not the values used in nutritional sciences and food composition databases; for these purposes, metabolizable energy is used. This is the energy that is available for use in metabolism by the body. Metabolizable energy values are calculated using energy conversion factors (Atwater and Bryant, 1900; Southgate and Durnin, 1970; Merrill and Watt, 1973; Allison and Senti, 1983) for the protein, fat, carbohydrate and alcohol contents. Recently, Livesey (2001) has argued that a better system for calculating the energy values of food would be the net metabolizable energy system (Blaxter, 1989).

Recently, the contributions from dietary fibre, polyols and oligosaccharides have been widely discussed (Livesey, 2001; FAO/WHO, 1998), but most databases do not yet use the energy conversion factors for these components.

In many countries, Le Système International d'Unités (or International System of Units [SI]) (BIPM, 1998, 2003) is used to express the energy values of foods and diets, using the Joule (J) (work): 1 kcal is equivalent to 4.184 kJ (thermochemical equivalent) (Royal Society, 1972). When expressing the energy value of foods, no more than three significant figures should be used. Whichever system of calculation is chosen for energy, it should be clearly indicated.