Early food composition studies were carried out to identify and determine the chemical nature of the principles in foods that affect human health. These studies were also concerned with the mechanisms whereby chemical constituents exert their influence and provided the basis for the early development of the science of nutrition (McCollum, 1957), and they continue to be central to the development of the nutritional sciences. Current knowledge of nutrition is still incomplete, and studies are still required, often at an ever-increasing level of sophistication, into the composition of foods and the role of these components and their interactions in health and disease.
Somogyi (1974) reproduced a page of the earliest known food composition table, dated 1818. Ever since, it has been customary to record food composition data in printed tables for use by both specialists and non-specialists. While printed tables will continue to be produced, computerized data systems have replaced them in some settings because of the ease with which data can be stored, and the facility with which the large amounts of data can be accessed and processed.
These systems are increasingly used to generate printed and computerized food composition tables and data files. Computerized and printed tables generally contain a subset of nutrients and foods and often no further documentation. A single computerized data system can generate a variety of tables and files, each containing specific subsets of numeric, descriptive and graphical information. Examples are the different user databases released by New Zealand (Burlingame, 1996).
Studies of the relationship between diet and health have led to increased interest in the range of biologically active constituents present in foods that accompany the nutrients, and data for these constituents are often required, as are data for additives and contaminants. A well-designed data system can accommodate non-nutrient data, although this should not detract from the primary objective of the database programme – the provision of data on the nutrient content of foods.
Early food composition tables were based on analyses carried out in the laboratories of researchers such as Von Voit in Germany, Atwater in the United States of America and Plimmer in the United Kingdom (UK) (Somogyi, 1974; Atwater and Woods, 1896; Widdowson, 1974). Later, the United States moved towards compiling tables from scrutinized data produced by a number of laboratories. An element of this procedure was introduced into the UK tables, where the third edition of McCance and Widdowson (1940) included vitamin and amino acid values drawn from the literature. Southgate (1974) distinguished these two methods as the direct and indirect method of compiling tables. These methods, and other procedures for compiling food composition data, were described by INFOODS (Rand et al., 1991).
The advantage of the direct method, in which all of the values are the results of analyses carried out specifically for the database being compiled, is that close control of the sampling, analysis and quality control procedures yields highly reliable data. Early UK food composition workers analysed different purchases of the same food separately, but without duplicate determinations, with the intention of gaining some limited information on nutrient variation in each food (McCance and Shipp, 1933). In subsequent versions of the UK tables, however, the various purchases of the food were combined, reducing costs and increasing the number of foods that could be analysed in a given period of time (McCance, Widdowson and Shackleton, 1936). Even with this procedure, the direct method remains costly and time-consuming, and imposes pressure on the analytical resources available in many parts of the world.
The indirect method uses data taken from published literature or unpublished laboratory reports. There is consequently less control over the quality of the data, which may be uneven. Great care must therefore be taken in their appraisal for inclusion in the database. In some cases, values are imputed, calculated (see below), or taken from other tables or databases, and it may be impossible to refer back to the original source; these values carry a lower degree of confidence. The indirect method is most commonly employed when analytical resources are limited, or the food supply is largely drawn from food imported from other countries where compositional data are available. Although the indirect method is clearly less demanding of analytical resources than the direct method, the level of scrutiny required often makes it time-consuming and costly.
Most food composition databases nowadays are prepared by a combination of the direct and indirect methods, containing original analytical values together with values taken from the literature and from other databases as well as imputed and calculated values. This combination method is the most cost-effective and is particularly successful when staple foods are analysed directly, and data for less important foods are taken from the literature (including that from other countries, if necessary). However, minimization of the amount of imputed and calculated values in principle increases the reliability and representativeness of the database.
Food composition databases currently available contain compositional values of differing quality, reflecting the different ways in which they were obtained. If data are to be used internationally they must be of consistent and compatible quality so that they can be used in combination for collaboration between individuals and countries in nutritional research, nutrition education, food regulation, and food production and processing. Data types and sources can be identified in food composition databases by codes (USDA, 2003a; Burlingame et al., 1995a), as is done in many countries, and by reference (Wu Leung, Butrum and Cheng, 1972). In general order of preference, the sources of data are:
These are values taken from the published literature or unpublished laboratory reports, whether or not they were from analyses carried out explicitly for the purpose of compiling the database. They may be assimilated into the database unmodified, or as a selection or average of analytical values, or as combinations weighted to ensure that the final values are representative. Original calculated values are included in this category (e.g. protein values calculated by multiplying the nitrogen content by the appropriate factor, or fatty acids per 100 g food calculated from fatty acid values per 100 g total fatty acids).
These data are estimates derived from analytical values obtained for a similar food (e.g. values for peas used for green beans) or for another form of the same food (e.g. values for “boiled” used for “steamed”). They may also be derived by calculation from incomplete or partial analyses of a food (e.g. carbohydrate or moisture by difference, sodium derived from chloride values or, more commonly, chloride calculated from the value for sodium). Similar calculations can be made by comparing data for different forms of the same food (e.g. “dried” versus “fresh” or “defatted” versus “fresh”).
These are values derived from recipes, calculated from the nutrient contents of the ingredients and corrected for preparation factors: loss or gain in weight, usually referred to as yields, and micronutrient changes, usually referred to as retention factors. Such values are only rough estimates, because the preparation conditions for recipes vary dramatically, such as cooking temperature and duration, which will significantly affect yield and retention. Another calculation method is the calculation of the nutrient values of cooked foods based on those of raw foods or foods cooked in a different way, using specific algorithms, retention and yield factors.
Figure 1.1 The integration of nutritional analyses of foods into food and nutrition research
These are values taken from other tables and databases where reference back to the original source may or may not be possible. Adequate reference to original sources is necessary to justify a borrowed value. In some cases, the borrowed values should be adapted to the different water and/or fat contents.
These are values presumed as being at a certain level or as zero, according to regulations.
Foods are chemically analysed for a variety of purposes. Food composition databases rely on nutritional and toxicological analyses conducted by government, academia and industry to determine the potential contributions of foods to the diet, and to determine compliance with regulations concerning composition, quality, safety and labelling. Foods may also be analysed for the purpose of ongoing monitoring of the food supply (e.g. Bilde and Leth, 1990). All of these compositional studies produce data that can be considered for entry into a food composition database.
In human nutrition studies, the composition of foods is investigated, ideally, in a research setting interacting with one or more other areas of nutrition research (Figure 1.1). The data are most useful when they represent foods in the forms generally consumed (see Chapter 5, Sampling).
In agriculture, factors such as disease resistance and yield, rather than nutritional value, have tended to dominate decision-making regarding policies and programmes. Similarly, in food technology economic considerations such as consumer appeal and profitability have been the major influences on product development. Attitudes are changing, however, and nutritional quality is now one of the factors considered in cultivar selection and the development of processed foods.
The production, handling, processing and preparation of foods profoundly affect their nutritional quality. Extensive literature covers agricultural practices (climate, geochemistry, husbandry, post-harvest treatments); processing methods (freezing, canning, drying, extrusion); and stages in food preparation (holding, cutting, cooking). Most nutritional studies in these areas, however, cover a limited range of nutrients (notably labile vitamins); very little information is provided on the broad range of nutrients (Henry and Chapman, 2002; Harris and Karmas, 1988; Bender, 1978; Rechigl, 1982). Nevertheless, data from these types of studies can often be useful in food composition databases, either as data per se, or in establishing relevant yield and retention factors for calculations (see Chapter 9).
Levels of certain nutrients, additives and contaminants in foods are monitored for several reasons. Some nutrients, for example, may react adversely under particular processing conditions, producing poor sensory quality or affecting the safety of the food (e.g. trans fatty acids). Labelling regulations also require prescribed levels of nutrients in specific foods (e.g. vitamins and minerals in fortified foods, polyunsaturated fat levels in margarine). Certain toxic substances are limited to prescribed levels and are monitored by government, industry and other laboratories. The nutrient content of manufactured foods is rarely made available in electronic format to compilers, and care must be exercised when compiling databases using information provided on food labels.
Food composition tables were, in the early development of nutrition, the major resource of food composition data; they are, however, constrained physically by the growing volume of compositional data, and their attendant documentation, or metadata. They are also expensive to update and thus older data can remain in use for longer than is desirable. The most significant disadvantage of food composition tables is that calculations made using the data they contain can only be made with considerable additional work. Computerized compositional databases do not suffer from these disadvantages and are often used instead of the printed tables as the primary sources of compositional data for foods. A comprehensive food composition database should be the repository of all numeric, descriptive and graphical information on the food samples.
This book is concerned with the production and assessment of food composition data intended for entry into a computerized database, but it is equally applicable to data intended for printed food composition tables, because the principles involved are virtually identical.
Food composition data can be managed at four different levels, which together provide an effective way of handling the data (Table 1.1). This approach has advantages for assessing the quality of the data. They form a sequence of stages.
These are the published research papers and unpublished laboratory and other reports containing analytical data, together with their bibliographic references. Normally, the data sources are part of the reference database.
These records (written or computerized) hold all data in the units in which they were originally published or recorded, and are scrutinized only for consistency as would be normal in the refereeing of scientific papers prior to publication. Foods should be coded or annotated to assist in identification, and values should be annotated to indicate unit, calculation, mode of sampling, numbers of food samples analysed, the analytical methods used and any quality assurance procedures in place. Any bibliographic references relevant to the data source are noted. At this stage it is possible to make a preliminary assessment of the data quality (see Chapter 8).
Such records should make it unnecessary to refer back to the original data sources whenever a query arises. Normally, the archival data are used in the preparation of the reference database.
Table 1.1 Stages in food composition data management | ||||
|
Stages |
Description |
Format |
||
Data source |
Public and private technical literature containing analytical data, including published and unpublished papers or laboratory reports |
As presented by original authors |
||
Archival record |
Original data transposed to data record without amalgamation or modification; scrutinized for consistency |
One data set per original source to include details of origin and number of food samples, food and analytical sample handling, edible portion, waste, analytical methods and quality-control methods |
||
Reference database |
Data from all records for one food brought together to form the total pool of available data
|
Common format
|
||
User database |
Data selected or combined to give base mean values with estimates of variance for each food item Format
|
Common format
|
||
The reference database is the complete pool of rigorously scrutinized data in which all values have been converted into standard units and nutrients are expressed uniformly, but in which data for individual analyses are held separately. This database should include all foods and nutrients for which data are available, and provides links to sampling procedures and analytical methods, laboratory of origin, date of insertion and other relevant information, including bibliographic references to the data sources. The data will usually be expressed according to the conventions, units and bases adopted for the user databases (see Chapter 9).
The reference database will usually be part of a computer database management system, with computer programs or written protocols developed to calculate, edit, query, combine, average and weight values for any given food. It is from this database and its programs that the user databases can be prepared.
The database will be linked to records on analytical methods and records for other constituents, for example non-nutrient constituents such as biologically active constituents, additives and contaminants. Records of physical characteristics such as pH, density, non-edible portion or viscosity that are often collected in food technology papers should also be linked to the reference database. Conversion factors, calculations and recipes should also be stored.
In general, the user database is a subset of the reference database, and the printed form often contains less information than the computerized form. Many professional users of food composition data would require the information recorded in the reference database, but most require only a database containing evaluated food composition data that, in some cases, have been weighted or averaged to ensure that the values are representative of the foods in terms of the use intended. Moreover, values for nutrients in each food may, if appropriate, be amalgamated (e.g. total sugars, ratios of the different classes of fatty acids) rather than shown as individual constituents. These databases may contain indications of data quality based on assessment of the sampling and analytical procedures.
These databases should include as many foods and nutrients as possible, with preference being given to complete data sets. Methods, sampling procedures and literature sources should be coded at nutrient level so the user can perform an independent evaluation or comparison with other databases. The data, of course, must be expressed in uniform, standard units (see Chapter 9). The defining feature of a user database may be considered as a database that gives one series of data per food item.
Simplified databases or tables can be produced from the main user database. In these, fewer nutrients are covered, and some reductions of food categories may be possible (e.g. for meat cuts data may appear only for “medium cooked,” omitting “rare” and “well cooked”). Values can appear as units per 100 g of food or per average serving, expressed in household units or portion sizes. Modified versions of the database can also be produced to assist manufacturers in food labelling. Various types of database or printed table can be produced from the same comprehensive database, ranging from a fairly extensive version for the professional user to a smaller version for consumers or for users involved in large-scale food preparation.
Tables and databases restricted to selected nutrients can be produced for people with special dietary needs or interests (e.g. for diabetics, or for people with kidney disorders for whom a diet controlling protein, sodium and potassium is required, or for nutrition educators, or for people wishing to lose weight). Data may be presented per 100 g of food, or per portion size or common household measures. Such tables and databases might be produced showing foods with ranges of nutrients – high, medium and low levels, for example. Data could also be given in other useful units (e.g. sodium and potassium in millimoles for renal patients).
Ideally, each country should have an established programme to manage its own food composition data, the data being considered an important national resource, as important as any other national collection of data.
While the level of certain nutrients in some foodstuffs will vary little between countries (e.g. the amino acid composition of lean meats), other nutrients, even in foods that are available worldwide, will vary greatly because of differing cultivars, soils, climates and agricultural practices. Recipes for composite dishes with the same name vary between countries. Different technological practices are also used; flour, for example, is produced and used at different extraction rates and may be fortified to different levels with different nutrients (Greenfield and Wills, 1979). Some countries have unique foods, food products or processing procedures (Somogyi, 1974). For these and other reasons, it is essential to develop a national food composition database programme, and to ensure that such a programme draws on data from other countries only when those values are considered applicable to nationally consumed foods.
Although attempts are being made to develop common food standards (e.g. the Joint FAO–World Health Organization [FAO/WHO] Food Standards Programme, Codex Alimentarius (FAO/WHO, 2003a,b), differences in food descriptions will continue to occur between countries.
The preparation of regional food composition databases is of great importance. Many countries, particularly in the developing world, lack the resources needed for a full-scale national food composition programme, but share a similar food supply to that of neighbouring countries. Cooperation between United States government departments, the Institute of Nutrition of Central America and Panama (INCAP) and FAO has produced some early regional food composition tables for Latin America (Wu Leung and Flores, 1961), Africa (Wu Leung, Busson and Jardin, 1968), East Asia (Wu Leung, Butrum and Cheng, 1972) and the Near East (FAO, 1982). More recently, this cooperation with FAO/UNU/INFOODS has led to the publication of regional tables for Pacific island countries (Dignan et al., 1994), Latin America (LATINFOODS, 2000) and Southeast Asia (Puwastien et al., 2000).
Some countries are collaborating on food composition analyses among themselves – for example, those in the North European region and those in the South Pacific region (Becker, 2002; South Pacific Commission, 1982). Other regional programmes may be those serving participating countries in multicountry epidemiological studies (Slimani et al., 2000). Simplified national programmes can be derived from such international or regional programmes.
The current high level of interest in nutrition requires that food composition databases meet the following criteria:
1. Data should be representative
Values should represent the best available estimate of the usual composition of foods in the forms most commonly obtained or consumed. Ideally, some measure of variability in the composition of the food should be given.
2. Data should be of sound analytical quality
Original analytical data from rigorously scrutinized sources are the ideal. Values from other databases, and imputed or calculated data should be included only when original analytical data are not available or are known not to be of sufficient quality. High-quality analytical data are those produced by methods that have been shown to be reliable and appropriate to the food matrix and nutrient in question. These methods must be applied proficiently, and evidence of this proficiency is required to assure data quality. It is also desirable that the analyst and the laboratory satisfy criteria of good laboratory practice. Further, evidence is required to show that the food sample was representative and was collected and handled properly. However, for existing data, documentation on sampling, source or analytical method is often not available, at least electronically.
Chapters 5, 6, 7 and 8 contain more specific guidelines for sampling procedures, methods of analysis and assurance of data quality; these three areas should always be considered in determining the quality of analytical food composition data.
3. Coverage of foods should be comprehensive
The database should include all foods that form a major part of the food supply and as many as possible of the less frequently consumed foods. The selection of foods for inclusion in a database is discussed in Chapter 3.
4. Coverage of nutrients should be comprehensive
Values should be included for all of the nutrients and other components known or believed to be important to human health. National priorities regarding health will have a major role in deciding which nutrients should be included. The criteria for selecting nutrients to be covered are discussed in Chapter 4.
5. Food descriptions should be clear
To be easily identified, foods must be unambiguously named and described. (Food nomenclature is discussed by McCann et al. [1988]; Truswell et al. [1991]; Møller and Ireland [2000a,b]; and Unwin and Møller [2003].)
6. Data should be consistently and unambiguously expressed
The data should be unambiguous in mode of expression and consistent in the use of units, factors used in calculation, and procedures used in rounding values.
7. Origins of data should be provided at nutrient value level
Information should be given on the sources of the data, noting whether data are analytical, calculated or imputed, and, as appropriate, on the procedures of any calculation and imputation, and the methods of sampling and analysis. Confidence or quality codes for the values should also be supplied.
8. Tables and databases should be easy to use
In addition to having clear terminology and systematic expression, databases and computerized tables must be easily accessible and readily understood. Printed tables should be of clear legibility and manageable size and weight.
9. The content of different databases should be compatible
The descriptions of foods, modes of expression and derivations of values should conform as closely as possible to existing international standards (e.g. the INFOODS tagnames) and to other major comprehensive food composition databases. Scientific needs require computerized databases and tables to be constructed with a view to using them in combination with other such systems.
10. Database should have few missing data
It follows from the above that any food composition database or table should aim to have as few gaps as possible because missing data can significantly distort the resultant nutrient intake estimations. It may be better to include imputed or borrowed data, always clearly identified as such, than no data at all. On the other hand, practical considerations often dictate that an incomplete database or table be produced to meet immediate needs. Information besides nutrient data (e.g. data on toxic substances or additives), though useful, is not essential at this stage.
Food composition data are used primarily for the assessment and the planning of human energy and nutrient intakes. In both cases, the approach is most useful when applied to groups rather than individuals. Assessment and planning can be divided into several subcategories for which the precise requirements of the database differ and for which additional information is required.
When the weights of consumed foods are known, food composition data permit the intake of each nutrient to be calculated by multiplying the weight of each food by the concentration of the nutrient in that food and then adding the results, according to the equation:
I =Σ (W1C1 + W2C2 + W3C3 + ........WnCn)
where: I = intake of the nutrient, W1 = weight consumed of food 1, C1 = concentration of the nutrient in food 1, etc.
Knowledge of nutrient intakes is required at several levels, as outlined below.
A person's nutrient intake can be calculated by the use of food composition data and food intake data (estimated from a dietary history or dietary recall or measured in a weighed intake study) (Cameron and van Staveren, 1988; Nelson, 2000). This information can show gross dietary adequacy or inadequacy, or dietary imbalance, and is important in the determination of dietary advice or in prescription of a therapeutic diet. The user must be aware, however, that due to the natural variability of foodstuffs, food composition data may not predict the composition of a single portion of any particular food with accuracy.
Foods consumed by populations can be measured by various techniques (Marr, 1971) and translated, by means of food composition data, into nutrients consumed. The results give one indication of the nutritional status of the group (Jelliffe and Jelliffe, 1989; Gibson, 1990) and may be used to explore the relationship of a diet to a variety of health indices – sickness and death patterns, growth rate, birth weight, measures of clinical nutritional status, physical performance, etc. Examples of groups usually studied in this way are:
Data drawn from studies of groups are used not only for identification of nutritional problems and planning of nutrition interventions to counteract them; they can also be employed in research that seeks to identify nutrient intakes desirable for good health. The results of such studies may feed back into food and nutrition policy in the form of food supplement programmes for children, food stamps for low-income groups, dietary advice to pregnant women, preventive diets for reducing heart disease rates, etc.
National statistics for agricultural production, adjusted for exports, imports, non-food use and gross wastage, are multiplied by nutrient composition data and divided by the total population to produce estimates of gross nutrient availability per capita. These data permit an assessment of the gross adequacy or inadequacy of the national food supply and indicate shortfalls or excesses. Food monitoring systems (e.g. Bilde and Leth, 1990) can follow the consumption of desirable and undesirable substances over a period of years.
Data from individual nations can be assembled to give cross-national or worldwide pictures of food and nutrient availability; such data are used in formulating food and nutrition policy, in setting goals for agricultural production, in formulating guidelines for consumption and particular policies such as food fortification or food supplementation (Buss, 1981).
Internationally, this information has implications for trade and for the development of assistance policies. In research, comparisons of nutrient intakes of different countries, together with other epidemiological data, enable further elucidation of the role of dietary constituents in health and disease. At present, long-term changes in the food supply can only be monitored adequately by the use of up-to-date food composition tables and databases. For example, the fat and iron content of meat have been altered in Western countries by changes in methods of husbandry and butchering. Comparison of today's cuts with those of ten years ago can be made by reference to past food composition tables (Vanderveen and Pennington, 1983).
Similar calculations can be made to provide estimates of the distribution of nutrients within a country. These findings can indicate actual or potential nutritional problems. Such studies are often critically important for developing countries that have diverse geographical regions. Periodic surveys, as part of a full system of nutritional surveillance, can monitor nutritional change and the effectiveness of food and nutrition policies.
The physiological requirements or recommended intakes of most nutrients have been estimated (e.g. FAO/WHO/UNU, 1985), and it is the job of the nutritionist to translate these requirements or recommendations into desirable food intakes, at varying levels of cost. Again, this task can be performed at several levels, as outlined below.
A therapeutic diet must be nutritionally balanced and adequate while at the same time controlling the intake of one or more specified nutrients. The prescription of therapeutic diets, therefore, requires professional training and a detailed understanding of the composition of foods. Table 1.2 lists types of disorder that require therapeutic diets, together with the dietary components that must be controlled. Unfortunately, most available food composition tables and databases do not hold information on all of the components listed in Table 1.2, and primary data sources may have to be consulted to obtain the required information.
Food composition data are used to translate recommended nutrient intakes into cost-limited foods and menus. Large sectors of the population (e.g. military establishments, workplace cafeterias, hospitals, prisons, schools, day-care centres and hotels) are provided with meals in this way.
Table 1.2 Examples of clinical conditions that require food composition information for the planning of therapeutic diets | ||||
|
Clinical condition |
Composition information required |
|||
|
Requiring general dietary control |
|
|||
|
Diabetes mellitus |
Energy value, available carbohydrate, fat, protein, dietary fibre |
|||
|
Obesity |
Energy value, fat |
|||
|
Hypertension |
Energy value, sodium, potassium, protein |
|||
|
Renal disease |
Protein, sodium, potassium |
|||
|
Deficiency states |
|
|||
|
Anaemia |
Iron, folate, vitamin B12 |
|||
|
Vitamin deficiencies |
Specific vitamin contents |
|||
|
Metabolic disorders |
|
|||
|
Haemochromatosis |
Iron |
|||
|
Hyperlipidaemias |
Fat, fatty acids, cholesterol |
|||
|
Inborn errors of amino acid metabolism |
Amino acids |
|||
|
Gout, xanthinuria |
Purines |
|||
|
Gall bladder disease |
Fat, calcium, cholesterol, dietary fibre |
|||
|
Wilson's disease |
Copper |
|||
|
Intolerances |
|
|||
|
Disaccharides, monosaccharides |
Individual sugars, especially sucrose, lactose, fructose, galactose |
|||
|
Gluten (and other specific proteins) |
Gluten, specific proteins |
|||
|
Migraine |
Monoamines |
|||
|
Allergies |
Specific proteins |
|||
| Note:This list is not intended to be inclusive. | ||||
A national food and nutrition policy will often define goals for the intake of certain nutrients. These goals must be translated into food production targets for the agriculture sector or into food consumption targets for the market or the public health sector (e.g. through increased subsidy or promotion of certain foods).
Food regulators use nutritional data on primary foods or “traditional” food products as a reference point for desirable nutrient levels for processed and newly introduced foods. For example, consumers should be able to rely on a traditional dairy product having certain levels of calcium and riboflavin; new processing techniques should not significantly alter the essential nutritional quality of the well-recognized product. Similarly, a manufactured or fabricated substitute should provide the same nutritional value as the food it is intended to replace (Vanderveen and Pennington, 1983).
A food composition database can also provide a preliminary check on label information or claims. For example, a food may be advertised as high in nutrient X, and information on the composition of its listed ingredients will indicate whether that food product could be high in nutrient X without fortification (for which special regulations may exist). Further, data on “new” cultivars being evaluated for widespread commercial introduction can be compared with data for traditional cultivars.
Some countries permit the nutrition data used in labelling certain composite foods to be calculated from nutrient data for ingredients taken from food composition tables and databases. In such cases, it must be ensured that nutrient values from the food composition tables and databases are comparable with those of the food regulations concerning food labelling.
Nutrition interventions, such as food aid programmes, supplementation schemes and disease prevention programmes, require the use of food composition data in order to translate specific nutrient needs into food requirements. Note that such programmes may require confirmation by direct analysis, particularly at the research level.
The limitations of food composition tables or databases are often not sufficiently understood by many users. Foods, being biological materials, exhibit variations in composition; therefore a database cannot accurately predict the composition of any given single sample of a food. Hence, although food composition tables and databases can be used to devise a diet, meal or supplement, the levels of nutrients are essentially estimates. For metabolic studies a direct analysis is usually necessary to obtain the required accuracy in the measured intake of the nutrients being studied.
Further, food composition databases and tables are limited in their usefulness for regulatory as well as scientific purposes. They cannot predict accurately the nutrient levels in any food; this is especially true for labile nutrients (e.g. vitamin C and folates) or constituents added or removed during food preparation (fat, moisture). Furthermore, the composition of a given food may change with time (e.g. a manufacturer's formulation may change) invalidating the use of the values in the database. Predictive accuracy is also constrained by the ways in which data are maintained in a database (as averages, for example).
Food composition databases frequently cannot be used as literature sources for comparison with values obtained for the food elsewhere. Values from one country should be compared with values obtained in other countries by reference to the original literature. Food composition databases can be used more confidently when the values are known to be based on original analytical values. Any imputations, calculations, weightings or averaging must be clearly documented and, most important, food items must be adequately described to enable comparisons to be made.
It seems that, despite major efforts during the past 20 years on harmonizing food descriptions, nutrient terminology, analytical methods, calculation and compilation methods, values from existing food composition tables and databases are not readily comparable across countries. In addition, users may not always be aware of the difference in nutrient values between raw and cooked foods and might erroneously use the values for raw foods in place of those for cooked ones. This is often the case in countries using food composition tables that contain mainly raw foods.
Finally, there has been an increase in the consumption of manufactured foods and mineral and vitamin supplements, accounting for up to 60 percent of the total food intake, but these are rarely listed in food composition tables and databases (Charrondiere et al., 2002). As a result, it can be assumed that nutrient intake estimations are increasingly unrepresentative of the actual nutrient intake.
The users of food composition tables and databases vary greatly: economists, agricultural planners, nutritionists, dietitians, food service managers, food and agricultural scientists, manufacturers, food technologists, home economists, teachers, epidemiologists, physicians, dentists, public health scientists, non-specialist consumers and journalists. Access to different types of computerized tables and databases is required to suit these differing needs; this is now achievable due to the availability of computers.
