|Food, Nutrition and Agriculture - 12 - Food Composition Data (FAO - FPND - FAO, 1994)|
|Food composition databases: Current problems and solutions|
Gustaaf Sevenhuysen is Associate Professor in the Department of Food and Nutrition, University of Manitoba, Canada.
Food composition databases are among the very important resources used by professionals working in areas such as food trade, food control, nutrition research and health promotion, The data in food composition tables are needed to make many health-related decisions and are assumed to be accurate. Unfortunately, this is not always the case, and serious attention is needed to ensure that high-quality food composition data will become available, However, work on food composition appears to receive very low priority compared to other activities when resources for development are limited.
INADEQUACY OF FOOD COMPOSITION DATA
Food composition tables and databases are available in most countries, yet the data they contain are invariably criticized as being too inaccurate for many purposes. For example, the use of food composition tables to calculate nutrient intakes for individuals is considered to be too unreliable for clinical and health-related research since nutrient contents of foods vary greatly. Similarly, food manufacturers have not been able to rely on the existing composition data to provide the accuracy required for regulator/work.
Users of food composition data need information beyond simple nutrient or component values, They demand more precise food descriptions that include the origin of data. Data on micronutrients and the different biologically active forms of nutrients are requested, as well as additional information on the representativeness and quality of existing data, Since such complete data and descriptions are not available to users at present, the tables and databases can be considered to be inadequate.
When food composition tables and databases were created in the United States and Europe during the 1950s and 1960s, considerable sharing of data took place so that listings could be compiled to interpret national nutrition surveys. In developing countries, the extent to which data were borrowed to compile tables was even greater. Thus, much of the food composition information used worldwide is based on outmoded technology and on analytical techniques that have been improved since the data were collected years ago.
When these databases were created, developers provided single value results for the nutrient composition of a food item. Users were not made sufficiently aware of the natural variation in the composition of a food or the compositional differences among foods from different areas, Descriptions and variations were not recorded by most institutions. Therefore, it is impossible to determine whether the reported values were influenced by factors now known to be important, Furthermore, users cannot be confident about the range of foods represented by these average values. Thus food composition data are inadequate for several important purposes, including food trade, clinical research and international epidemiology.
CONSEQUENCES OF POOR DATA
Current compositional data are used for many purposes in spite of the inherent uncertainties about the values. However, these uncertainties threaten to cause higher costs and inefficient use of resources.
Costs to industry
The lack of precise food composition data can be detrimental to food manufacturers in a number of ways. In complying with regulations on product content, the food industry may use too much of an ingredient that is known to contribute the component being regulated, A manufacturer faces higher costs when it selects alternative ingredients for which there are no data, Expenditures rise with more frequent chemical analysis for product development and quality control. Companies that have access to reliable data are able to modify product formulations in response to consumer concerns, while those that lack data lose competitive power.
These costs are more serious for small businesses, and the development of food composition tables will help the growth of this segment of the industry. Larger industries have solved their problems by generating new analytical data for their own manufacturing processes. The data obtained in this way are reliable and meet the specific needs of the production processes. Such data are, however, obtained at a higher cost than would be incurred by the use of a locally generated food composition table.
Costs to industry also rise with each additional government regulation placed on food products or their manufacture. Whether the food regulations are aimed at health promotion among the national population or at the promotion of trade, they often increase the need for food composition data.
Ultimately, officially accepted databases will reduce the cost of food to consumers because food manufacturers add the cost of privately obtained data to their prices.
Costs to government
Governments require precise food composition data to lessen the frequency of chemical analysis needed in the enforcement of food inspection and food labelling regulations. Food composition data can lower the costs of estimating the exposure of specific population groups to food components that are associated with chronic disease, In screenings for potential deficiency or chronic diseases, these data can reduce the expense of testing the nutrient intakes of individuals, Finally, the availability of data can lower the cost of estimating nutrient and food component intakes from imported food and reduce the use of composition data from other locations.
Costs to government are more difficult to quantify than those for industry, The effects of policies from various sectors on the health of the population are not well known, particularly in the case of chronic conditions, Therefore, it is difficult to determine the extent to which poor food composition data have contributed to uncertainty in research results, In spite of the fact that large population groups may be affected by such policy decisions, generating new food composition data is difficult because the process is expensive and time consuming.
Many developing countries lack good data, yet their populations are too small to justify the expenditure of large resources for food composition activities, However, a combined effort to generate data for a group of countries, which would benefit several populations at the same time, could be justified.
BENEFITS OF BETTER DATA
The values chosen as representative will be used to determine whether the intakes of individuals meet recommended intakes. As a result, these values affect decisions taken by food distributors and food manufacturers. Yet, the professionals who calculate individual nutrient intakes will often be unable to decide which database entries represent the actual food eaten because the information does not include details on variable factors that affect the foods composition, such as growing conditions, stage of ripeness or product formulation. Such uncertainty can be minimized by better descriptions of the source of data and the growing, storage and processing conditions. Database developers can promote better decision-making by providing the details that explain the food component values.
Recent changes in analytical methods have led to new nutrient values, For example, data on the vitamin contents of many foods were revised in national tables following the wider use of high-pressure liquid chromatography (HPLC) equipment. The reanalysis of b-carotene values in East Africa (West and Poortvliet, 1993) shows that previous methods overestimated the amounts of this nutrient in foods, The current estimates of amounts of b-carotene available to the population are half those used in the past, These results can affect food policies in East African countries, In this case, assumptions about the vitamin A activity in the diets of children will change dramatically as a result of better food composition data. Not only are nutritional problems defined more accurately, but subsequent interventions will be more effective in saving children from permanent damage, Finally, the costs of interventions can be calculated more accurately.
As newer methods are used to reanalyse commonly eaten foods in developing countries, it is reasonable to expect that similar advantages will be documented in other cases, As a result, interventions promise to be more efficient in improving the nutritional status of populations.
USE OF EXISTING DATA
Food composition databases are used regularly in research and policy planning with little quality control because descriptions of the data which would permit informed judgements are not provided. For example, when the separate values for several foods are not available, it is assumed that their nutrient contents are the same and that the content of one food can represent all the others, This acceptance of methods that involve extensive simplifications is unusual in science, and the compromise on precision and accuracy may be greater than would normally be acceptable for scientific methods, In any one table there are good data, but it is not possible to know what proportion of the data are good, As time passes, the proportion of data that are considered to be good decreases because the foods in the market change in composition and newer analysis techniques raise the quality standards.
Nevertheless, it is possible to use current composition data, provided adequate care is taken to explain the limitations in results. The user needs to remember to ask for and check the description of the data and to check the extent of the possible compromises on data quality and representativeness, In addition, the user needs to clarify the level of precision required for the purpose. While a very high level of precision is needed for clinical applications, a much lower level is acceptable for food aid calculations involving some nutrients.
It is most important that users indicate the extent of uncertainty in the results of calculations based on the data, either with straightforward descriptions or as a set of confidence intervals. Such explanations will ensure that readers clearly understand the reasons for the limitations of the results given, The explanations permit more reliable interpretation of nutrition survey and policy formulation data and remind the decision-makers who rely on the interpretations of the uncertainties in food composition data and the reasons for actively maintaining such data.
At present, more description of data is required, including the application of universally recognized food naming systems and more complete explanations of data quality, In addition, better understanding is required to determine which data are interchangeable and which are not, particularly because food composition data are used for many different purposes.
MODERN FOOD DATABASES
The essential part of a table or database is the set of values of chemical analysis. Each value is associated with a food name and a nutrient name. All other features of a database are chosen by users and are designed to facilitate entry, retrieval or modification of values (Greenfield and Southgate, 1992).
Retrieving nutrient and component values is the first function of any system for managing the data. Two basic elements of management are the list of foods and the list of nutrients, However, users require more than names in selecting the same foods from a long list because names may have slightly different meanings for individuals whose training and experience differ. The problem is that a single, short food name cannot describe all the different attributes of a food item.
The food should be labelled with a name from the local culture that is fully understood by users in the area; this name can be a word from the local language or English, or both, Variations in common names of food may be unrelated to differences in composition. Each name shows the characteristics of the food that a user group views as important; it can reflect the way the food is used in the kitchen, at meal time or during social occasions. The first task therefore is to choose names that individuals can use within their culture but that also separate foods according to differences in composition. Hence, this descriptive name should imply details about the food, such as maturity of the sample analysed, storage conditions or any processing or refinement steps. Local users are familiar with the common forms of the foods, but other users need to have these descriptions in order to choose appropriate nutrient values.
The ways of characterizing any food item are too numerous to capture in a food name, In order to make the many possible descriptions available in a practical way, foods need to be coded with more than local names. Several systems have been developed to help in this task. Eurocode is a system of codes developed cooperatively by a number of European institutions for the food composition databases of European countries (Kohlmeier, 1992). The codes are made up of a number of separate elements which refer to different attributes such as the group to which the food belongs and the country where the food is commonly eaten, Eurocode is designed to integrate databases from European countries. The International Network of Food Data Systems (INFOODS), established by the United Nations University, has designed a system with simplified names for international use (Truswell et al., 1991), Langual is a comprehensive coding system developed by the United States Food and Drug Administration (1989) to bridge the needs of many different user groups, The Langual system uses diverse codes for each food, chosen from different hierarchies of codes, each one describing a major characteristic of the food.
Any coding system that creates food group structures in the database is potentially complex and requires significant time and effort for the addition of codes or descriptors to the existing data. This problem is particularly pronounced in the case of Langual because of its high level of complexity and the large amount of data used to describe foods, However, such problems are minimized with the use of computers; database management software reduces the amount of arduous work and allows customized applications of the same coding system for user groups with different needs.
Nutrient names or labels also need to be standardized because users demand that labels refer not only to the nutritionally important component of the food, but also to the analytical method used to generate the value in the database, The solution proposed is a system of nutrient labels called tag names, which are nine-digit letter combinations that can be incorporated into any food composition database. With a central registry of tag names it is possible for users in many areas to label the nutrient entries in their own databases in a way that is compatible with other databases.
The process of retrieval is complicated by the need for descriptions of source and quality to be associated with each nutrient or component value. The source descriptions allow users to decide whether values are likely to represent the composition of foods of interest. At present, many databases or tables imply that all the values are relevant for any geographical areas mentioned in the title. For example, the food composition database for Thailand is intended to represent foods anywhere in the country, as is Indonesias database, or Canadas. However, foods from different areas of these large countries may show important variations in content, and only average figures are included in the respective databases. Source codes can provide users with more detail on which to base judgements about the composition of foods eaten.
Indications of data quality
Most databases do not include information on the quality of data. The values included are those deemed to be reliable by the analysts who provide the data; there is rarely any assurance that judgements are comparable or explanation of the judgements for users.
Users must rely on analysts explanations to know the effect of analytical methods on composition data and to judge whether the data are precise enough for their purposes. A code with three categories, ranging from unreliable to as accurate as current technology and methods allow, has been developed to score data quality (Schubert, Holden and Wolf, 1987), although to date it has been applied for only a few nutrients.
Other descriptions can be included in databases, especially with new electronic capabilities, For example, images of foods offer advantages in representing many food characteristics, such as physical state, colour, maturity, part of plant or animal, preparation and processing, Also, pictures present the food to the user in its familiar form; this is very important in the case of unknown, imported foods, Another use of pictures is to show the sample that was actually analysed; this image is particularly useful for analysts and database developers who need to compare data on the same foods from different sources.
To improve food composition data, several priorities were established in 1994 at an international meeting held in Tunis, Tunisia (FAO, 1994). The first priority is to generate new data that provide better estimates of the real composition of food.
A second priority is to describe both the foods and the composition data more fully. In this way, users who need additional detail can decide whether the conditions of the food under consideration match those of the samples in the database and whether the values were obtained from appropriate locations.
The third priority is to express data through distributions of values, instead of mean or median values, in order to capture the uncertainties in the numbers. Until now, data have not been available to quantify the uncertainty, and the results as presented often imply greater accuracy than could actually be ensured.
Since professionals cannot work without food composition data, they frequently say that poor data are better than no data, In the short term, professionals must continue to use the existing food composition data. The situation can only change in the long term, because the effort required to improve data is large in comparison to the resources available to most institutions or governments for carrying out the work. Hence, it is difficult to change available food composition data within the time allowed for planning specific policies or designing particular interventions that depend on such data.
In each context, local, national or international, the costs associated with using poor data need to be weighed separately, In choosing foods and nutrients for analysis, priorities also depend on the relative importance of the foods in terms of local food consumption and trade.
Data will be more effectively generated and managed if professionals and institutions cooperate rather than working independently. Regional cooperative programmes promoted by FAO and INFOODS are good examples of collaboration that increases the effectiveness of separate efforts, As a result of these efforts, more and better databases for local objectives will become available in the next few years.
FAO. 1994. Proceedings of discussions on food composition. Tunis, Tunisia. Unpublished Food Policy and Nutrition Division Report. Rome.
Greenfield, H. & Southgate, D.A.T. 1992. Food composition data: production, management and use. London. Elsevier Science Publishers,
Kohlmeier, L. 1992, The Eurocode 2 food coding system, Eur. J. Clin. Nutr..46 (Suppl. 5): 526-534.
Schubert, A., Holden, J.M. & Wolf, W.R. 1987. Selenium content of a core group of foods based on a critical evaluation of published analytical data, J, Am. Diet, Assoc., 87: 285-299.
Truswell, A.S., Bateson, D.J., Madafiglio, K.C., Pennington, J.A.T., Rand, W.M. & Klensin, J.C. 1991, Committee report: INFOODS guidelines for describing foods to facilitate international exchange of food composition data. J. Food Compos. Anal., A; 18-38.
United States Food and Drug Administration, 1989, Langual users manual. Washington, DC, Center for Food Safety and Applied Nutrition, Division of Information Resource Management, Food and Drug Administration.
West, C.E. & Poortvliet, E.J. 1993. The carotenoid content of foods with special reference to developing countries. Arlington, Virginia, USA, International Science and Technology Institute, Vitamin A Field Support Project (VITAL).