Since the first edition of this book was published (Greenfield and Southgate, 1992) there have been many dramatic changes in the world that are of great significance to the areas of food production, management and use. These are outlined below. First, it is especially noteworthy that the final report of the International Conference on Nutrition (ICN) produced a World Declaration and Action Plan for Nutrition (FAO/WHO, 1992) that contained references throughout to the need for data on the nutrient composition of foods, especially in the sections under Section IV, “Strategies and actions”. Specifically under Section IV, 9, j, we read “Support and encourage ... the development and use of local food composition information”. The ICN was followed up by means of countries developing their Plans of Action for Nutrition and, later, Implementation Reports on these plans. New Zealand's National Plan of Action for Nutrition (MOH, 1996) emphasized, “Food composition provides essential information for effective monitoring of food and nutrition. In order to remain current food composition needs to continue to be updated and expanded to include new local and appropriate international food composition values.”
Second, the activities of INFOODS have now moved centre stage with the centralization of this important function within FAO. FAO had earlier reduced its involvement in food composition work after the publication of the food composition tables for the Near East (FAO, 1982) but, in 1994, FAO renewed its commitment to improving the quality and availability of food composition data in developing countries. As part of this new effort, FAO joined UNU in the coordination of INFOODS. With the amalgamation of UNU and FAO interests, INFOODS began to operate out of the FAO Rome headquarters in 1998 (see Chapter 1).
Third, in 1993 the First International Food Data Base Conference was held in Sydney, Australia (as an official satellite to the International Union of Nutritional Sciences [IUNS] International Congress on Nutrition) and proceedings were published (Greenfield, 1995). The series of international food data conferences has continued, each with published proceedings: Finland in 1995 (Finglas, 1996), Rome in 1999 (Burlingame, 2000), Slovakia in 2001 (Burlingame, 2002), and the United States in 2003 (Pennington and Stumbo, 2004). In 1997, IUNS established a Task Force for the International Food Data Conference, to oversee the arrangements by selecting convenors and venues and assisting with publicity and provision of resources (IUNS, 2003). These conferences and their published proceedings have done much to promote food composition research internationally. INFOODS (2003) hosts the Web site for the International Food Data Conference.
Fourth, access to personal computers became almost universal, with the wide accessibility of the Internet in the early 1990s creating infinite possibilities in terms of making information about food composition available worldwide. The first edition of this book (which was in preparation from 1983 to 1992) was mostly prepared by sending messages through telex and posting drafts back and forth through airmail, while the current edition has almost all been prepared by means of e-mail correspondence and attachments. Many food composition data or databases can now be accessed or downloaded from the Internet. Some are free, e.g. Nutrition Society of Malaysia (2003), LATINFOODS (2003) and USDA (2003a). Others are for purchase online, for example the German database (Souci-Fachmann-Kraut, 2003). In Australia and New Zealand, a free simplified food composition database of local foods that enables calculation for nutrition-labelling purposes is available via the Internet (FSANZ, 2003; Crop & Food Research, 2003), and many food composition programmes have their own Web sites, for example the Danish Food Composition Databank (Danish Veterinary and Food Administration, 2003). User software can also be purchased via the Web, with direct downloading of the software a possibility. Databases and software can even be downloaded on to palmtop devices.
While there is always a danger that the availability of food composition data over the Internet could lead to the downloading of inappropriate data, or posting of poor quality data (or data with no source identified), the Internet nevertheless represents an unlimited potential force for good in the area of food composition.
Burlingame et al. (1995b) were the first to document the tremendous potential of food images as tools to support the development of food composition data. It would be particularly useful to see more data linked to images of both foods and labels on the Web. An exemplary site is the United States Food and Drug Administration's online Regulatory fish encyclopedia (FDA, 2003), which shows photographs of aquatic foods in their raw state and as prepared raw for retailing, together with taxonomic information and images of isoelectric focusing gel strips for unique identification. Although this site is not linked to compositional data on fish it does demonstrate the exciting possibilities available for foods in general.
The development of online training courses in food composition data production, management and use would appear to be an essential next step.
Fifth, increasing interest in nutritional epidemiology continues to be a driving force in the demand for more and improved food composition data. Major prospective epidemiological studies are beginning to yield results that demonstrate the importance of this approach to analysing the relationships between food and health. Epidemiological studies need to produce tailored databases (Slimani, Riboli and Greenfield, 1995). Multicentre multinational studies need specific food composition databases that will produce comparable results (i.e. not attributable to artificial differences among the different national databases) (Deharveng et al., 1999; Charrondiere et al., 2002)
Sixth, international standards development and the increasing tendency to harmonize food regulations worldwide have been major forces in developing improved methods of analysis and quality assurance programmes to ensure that data from all parts of the world are more reliable and compatible. The joint FAO/WHO Codex Alimentarius has become the global reference point for all countries in formulating and harmonizing food standards and ensuring their global implementation (FAO/WHO, 1999). Other events, for example, the merging of food markets in Europe, have created the need to harmonize food laws and compliance with them (Goenaga, 1994). Buss et al. (1998) identified the food composition priorities and resources in relation to the European Union, while a group in Europe has been very active in comparing analytical methods and developing certified reference materials (Finglas, 1996; Vahteristo et al., 1996; van den Berg et al., 1996), as has another in Asia (Puwastien, 2000). These developments have achieved much to improve laboratory performance and data quality.
The principal objective of the INFOODS initiative (under whose aegis this book has been produced) is the development of an international network of food data systems, dependent on the development and potential integration of compatible local, national and regional collections of food composition data. A list of INFOODS Regional Data Centres is provided in Appendix 1.
Compatibility does not require the adoption of the same format or the development of one database system that meets all present and future needs; it merely means that the data can be used together (Southgate, 1985) and interchanged and interpreted without ambiguity or loss of information (Klensin, 1992). Although some essential features such as the modes of expression and nomenclature of foods and nutrients must be the same, one of the most important requirements for compatibility is that the data be of high quality – a user must have confidence that the data are fit for the task at hand.
The thesis developed in the first edition of this book was that the production of sound compositional data depended on an integrated series of activities involving the data users, the analysts who generate the data and the database compilers. Sound data quality must be built into the programme from its inception. As described throughout this book, there have been some considerable movements to advance this aim.
The preparation of the revised edition of these guidelines revealed a number of topics whose further study would advance the development of compositional databases. They are discussed below, following the order of their emergence in these guidelines.
It is essential to recognize that a sound compositional database that is both comprehensive and representative of available foods is an essential basic tool for virtually all quantitative nutrition research, dietary evaluation and development of food and nutrition policies.
The validity of nutritional epidemiological studies depends on accurate food consumption and food composition data. Failure to understand the relationships between diet and health or disease are often due to inadequacies in food composition or food consumption data. Thus, a food composition database programme should be an integral part of any national nutrition research programme as it is, for example, in the USDA's Human Nutrition Program (USDA, 2003d), which states:
The mission of the Human Nutrition Program is to conduct basic and applied research to identify and understand how nutrients and other bioactive food components affect health. The ultimate goal of this food-based agricultural research is to identify foods and diets, coupled with genetics and physical activity, that sustain and promote health throughout the life cycle. The research components of this program include: nutrition requirements; diet, genetics, lifestyle, and the prevention of obesity and disease; nutrition monitoring; composition of foods; health promoting intervention strategies for targeted populations; health promoting properties of plant and animal foods; bioavailability of nutrients and food components (e.g. phytonutrients and phytochemicals).
Programmes for the collection of food composition data used to vary widely among countries, the differences often reflecting historical differences in how nutrition developed within individual communities. The international need for this large body of information demanded a degree of harmonization and the development of compatible standards of data quality, which in turn required that some common principles for the organization of nutrient composition studies of foods be developed.
In the course of the reading and consultations for the revision of these guidelines, it became clear that the most important organizational principle was still the integration of the efforts of the users (real and potential), of those involved in sampling and analysis, and of the compilers. The involvement of these three major elements in all stages of the programme is probably the most effective way to achieve high data quality. Data quality can be “grafted on” by the compilers at a later stage, but this approach invariably results in the rejection of work that would have met the desired standards had they been introduced earlier. Quality assurance programmes within the analytical laboratory are essential, but they need to be incorporated into the programme as a whole. This statement is as true today as it was when the first edition was written.
The coverage of foods in all existing databases is very limited, compared with the numbers of foods consumed. This situation is likely to persist for the foreseeable future because the resources required to prepare truly comprehensive databases are considerable. It is therefore vital that priorities are properly assessed when future analytical studies are planned and that reanalyses are undertaken only when good evidence indicates nutritionally significant changes in composition or when new information on nutrients is needed.
There are three broad groups of foods for which information is conspicuously limited, and for which analytical work would be worthwhile.
Uncultivated foods. These foods are prominent in many communities and can assume great importance in times of food shortage following the failure of cultivated crops. Systematic compositional studies of uncultivated foods now assist nutrition studies of populations consuming them (e.g. Brand-Miller et al., 1993; Kuhnlein et al., 1979; Kuhnlein et al., 2002). Such studies could also provide information on species that may be suitable for further development (e.g. Dawson, 1998).
Individual cultivars. Many studies have demonstrated that different cultivars of the same species can have very different nutrient contents (Huang, Tanudjaja and Lum, 1999). With the advances in food biotechnology, the documentation of the composition of the existing food biodiversity, cultivar by cultivar, should be a priority (Kennedy and Burlingame, 2003) and a prerequisite before embarking on the development of genetically modified cultivars, as was recently recommended by the International Rice Commission (FAO, 2002; Kennedy, Burlingame and Nguyen, 2003).
Cooked foods and composite dishes. Foods are most often consumed in this form. In most databases direct analytical information is limited, forcing reliance on calculations from recipes. While this approach has its uses, there is a need to supplement and, ideally, replace calculated values with analytical data. Such studies will require careful attention to the design of sampling protocols.
The generation of analytical values to fill the gaps found in most nutrient databases depends in part on the availability of suitable methods, which will be discussed later. Nutritional priorities determine which nutrients should be studied. Data for carbohydrates and dietary fibre in foods are now available worldwide, although gaps still remain for many foods in most countries. Methods for fatty acid analysis are now well established, and many new fatty acid data and data compilations have been produced (Quigley et al., 1995; Exler, Lemar and Smith, 2003; Mann et al., 2003). Data are still badly needed on the folate values of foods, especially given the recognition of the importance of folate in the neurological development of the foetus and the introduction of mandatory or voluntary fortification of foods with folic acid. More data on carotenoids (both provitamin A carotenoids and others that are not precursors of vitamin A) have been collated into databases (Chug-Ahuja et al., 1993), as have data on phytoestrogens, although such data are mainly from only a few sources. Other bioactive components of great interest and that need research have been summarized by Pennington (2002).
Research on bone mass and osteoporosis have revealed that data are badly needed on vitamin D in foods. Interest in this vitamin has resurged in recent years and some compilations of data are now recognized as being out of date, with new data only being produced slowly
(J.M. Holden, US Nutrient Data Laboratory, personal communication, 2002). More data are also needed for vitamin K in foods given the increasing awareness of this nutrient's significance in bone health (Buttriss, Bundy and Hughes, 2000; Bolton-Smith et al., 2000; Shearer and Bolton-Smith, 2000).
An experimental basis is required for the design of sampling protocols. Despite the importance of variability within foods, formal studies on the factors involved in the variability of food components and the magnitude of their effects have been restricted to a few major commodities and have rarely been performed for nutritional reasons. Such studies could be incorporated, with advantage, into studies of the factors affecting the nutrient composition of many important foodstuffs.
The effects of sample handling are frequently studied during the course of a food composition study, and it would be valuable if these investigations could be conducted in a more formal way, making the information suitable for publication. Such information would be useful to all engaged in similar work.
The detailed studies of food nomenclature undertaken by McCann et al. (1988) and Truswell et al. (1991) and the formal studies of food classification undertaken for the Eurocode system (Arab, 1985; Arab, Wittler and Schettler, 1987) were central to controlling a major source of error in the use of nutrient data, that is, the identification of food items. This work developed further with LanguaL (Pennington et al., 1995; Møller and Ireland, 2000b). These systems can develop a degree of “elegant complexity” that makes them difficult to use accurately and consistently. It is therefore important to devise some formal procedures for evaluating nomenclature systems as they evolve. Some authors speculate that a single, internationally acceptable system of food nomenclature may not be an achievable goal (Burlingame, 1998). Nevertheless, this important work continues through an INFOODS-convened international technical committee, with the task of overviewing and focusing the work done on food classification and description in order to harmonize as far as is possible (INFOODS, 2003).
Since the first edition of this book was published in 1992 there has been an explosion of analytical methods development, particularly stimulated by the now worldwide acceptance of compositional standards for foods, and of the requirements of nutrition labelling in many countries (Government of Canada, 2002; EC, 1990; United States Code of Federal Regulations, 2003; FAO/WHO, 2001). This explosion makes it more difficult for a single analyst to be expert in methods across the board, and has created an even more urgent need for analysts, compilers and users of nutrient databases to share their knowledge and information.
The validation of methods for vitamin analyses is urgently needed, especially for the carotenoids (both the vitamin A-active and the non-vitamin A-active), folates and vitamin
D. In all cases procedures that permit the separation and measurement of the different forms are required. This information, together with estimates of the biological activity of different vitamers, would provide better estimates of the vitamin activity of foods than are currently available. All methods for vitamin analysis are time-consuming and therefore costly; efforts to devise more rapid specific procedures should be given high priority.
For some inorganic nutrients speciation is an important determinant of bioavailability, and its measurement could be useful (e.g. haem and non-haem iron).
The methodology for determining dietary fibre is developing rapidly; indeed, significant progress has been made during the preparation of these guidelines. The stage has not been reached, however, where the methods can be applied routinely to a wide variety of matrices; this is a legitimate goal for research.
For many methods there is a need to extend the range of food matrices covered, not necessarily because the methods are inappropriate, but simply because their wider applicability has not been assessed. A food composition study frequently covers a wide range of foods, and for that reason it would be helpful if the applicability of some methods could be extended to a greater variety of matrices. Well-tested methods with broad applicability are needed. In the long run it is hoped that instrumental methods can be further developed that do not destroy or invade the food – methods such as NMR, NIR and so forth offer the main potential in this regard.
Nutritional analysis is a specialized branch of food analysis, and since the first edition of this book was published many new, comprehensive and extremely useful analytical textbooks and manuals have been produced (see Appendix 7).
The importance of a quality assurance programme in the analytical laboratory was explained in Chapter 8. Such programmes have benefited from more collaborative studies and from improved availability of standards and standard reference materials (SRMs) as outlined above, but more work is still required.
The range of SRMs discussed in Chapter 8 still requires expansion, particularly for the more labile nutrients, and for “new” components of interest such as phytochemicals.
Typewritten and spreadsheet-prepared food composition tables, with their two-dimensional formats allowing little or no documentation for each value, are being superseded. Relational database management systems provide facilities for the compilation of thoroughly documented food composition information, including analytical values down to the finest level of disaggregation. These systems can provide ¾exible user interfaces for selecting, viewing and editing the data and documentary information in formats that are convenient and meaningful for the users. The information is stored in data structures that are designed to minimize redundancy and to support extensions to documentary information as and when these are de½ned in data management guidelines. Similarly, facilities for calculating and manipulating component values will be extended according to user requirements, preferably de½ned as internationally accepted guidelines (Unwin and Becker, 2002). A large-scale acceptance of international standards for interchanging food composition data will also facilitate rapid and simple interchange of food composition data (Klensin, 1992).
The greatest need is for more food composition data to be published in the scientific literature and for the standard of published data to improve. This could be achieved by requiring more documentation of the food samples analysed and, specifically, heightened scrutiny of analytical methods in the refereeing process. Details of quality assurance steps taken should be provided also. At present the methods sections of many papers rarely meet even the basic criterion of providing sufficient detail for a competent worker to repeat the work described. It is important that this minimum standard be preserved and, preferably, improved.
Formal procedures for the scrutiny of analytical data, from both published and unpublished sources, require further development. Such research should produce more objective indices of data quality that estimate the probability that the data are sound. Currently, errors can arise if intuitive data quality indices are manipulated as if they were real numbers. Formal analysis of the value judgments applied in the compilation process should lead to more objective and consistent assessments of data quality. Some steps towards these goals have been made, for example, the development of multinutrient data evaluation systems (Holden, Bhagwat and Patterson, 2002)
The application to the compilation process of the requirements for good scientific practice will provide a basis for quality. These include: independent replication of data, maintaining professional standards, documenting data, best practice in data management, questioning one's own data and fully documenting and preserving all data sources (Office of Science and Technology, 1998; Office of Research Integrity, 1998).
A database can be interrogated in a number of ways; at a simple level the composition of a single food item can be selected for information or scrutiny, but in the majority of cases data are required for combinations of food items. The accuracy with which a database predicts the composition of such combinations of foods is currently an area requiring research. All databases have limits of predictive accuracy determined by the variations in food composition. Future research needs to define and go beyond these limits. In addition, large-scale epidemiological studies (Riboli, 1991) have particular needs in the use of food composition databases. For example, the need to analyse dietary intake data on the basis of individual ingredients rather than composite foods may require specialized applications.
At present, intuitive assessments suggest that the principal requirements are for better data on variations in the nutrient composition of major foodstuffs, for elimination of missing values and for inclusion of more food items in the database. Formal studies are required to estimate the importance of these three elements, however, before substantial resources are committed to their resolution.
In countries where nutritional labelling of foods is common, reliable data from the food industry could be a major factor in improving database accuracy in use.
Perhaps most importantly, the objectives of international harmonization of food composition data and data management can only be achieved by training and education. Education and training programmes will develop a network of workers with common goals and standards who will contribute to the development of common approaches to the organization of food composition programmes, to food nomenclature and nutrient analysis and expression, as well as to food sampling and data quality assurance programmes. Data will become more compatible as their quality improves.
Courses in nutrient analysis of foods are now becoming more common in the training of analytical chemists, food scientists, nutritionists and dietitians, including at undergraduate level. Further, the movement of INFOODS into the basic work programme of FAO has led to the development of international short courses in food analysis and, in collaboration with Wageningen University, in food composition data, production, management and use.
The next long-awaited development is to see nutrient analysis of foods adopted as an essential component of the core training of food professionals such as dietitians and nutritionists because they are often the compilers of databases, as well as major users. Online training would be a very desirable development for the future, made possible by the development of the Internet, the widespread availability of computers and the fact that computer literacy is now an integral part of school education.
Finally, there is still a need for a fundamental change in attitudes towards the place of food composition work within the nutritional sciences themselves. Quantitative data on the composition of foods form the basis for virtually all quantitative human nutrition research and for the development of food and nutrition policies at the national and international levels. Food composition databases represent the primary scientific resource from which all other studies flow. It is vital for the development of the nutritional sciences that this key resource be maintained and developed as part of the activity of nutrition research as a whole.