Quality and Accessibility of Food-Related Data

Proceedings of the First International Food Data Base Conference A satellite to the 15th International Congress of Nutrition

Edited by Heather Greenfield The Proceedings of the First International Food Data Base Conference held in Sydney, Australia, 22–24 September 1993.

Heather Greenfield and FAO First edition Copyright© 1995

by AOAC INTERNATIONAL 2200 Wilson Boulevard, Suite 400 Arlington, VA 22201-3301

This publication is reproduced with permission of AOAC International.

Library of Congress Catalog Card Number 95-76836 ISBN 0-935584-56-0 Proceedings Sponsor

The publication of these Proceedings was

generously assisted by the Nutrition Section,

Department of Human Services and Health,

Australia, in accordance with its educational role in

the implementation of the national Food and

Nutrition Policy. Sponsors

The Conference was sponsored by:

Department of Industry, Trade and Commerce

The Scientific Association Dedicated to Analytical Excellence® AOAC INTERNATIONAL Department of Food Science and Technology, University of New South Wales Food Industry Development Centre, University of New South Wales Eurofoods-Enfant Australian International Development Assistance Bureau

National Food Authority Australian Institute of Health and Welfare McDonald's Australia Ltd NSW Dairy Corporation Goodman Fielder Ltd Xyris Software (Australia) Pty Ltd

National Association of Testing Authorities First International Food Data Base Conference

Scientific Programme Committee

D.A.T. Southgate (UK) Chair

N-G. Asp (Sweden)

G.R. Beecher (USA)

R. Bressani (Guatemala)

R. English (Australia)

J.N. Thompson (Canada)

Aree Valyasevi (Thailand)

C.E. West (The Netherlands)

Organizing Committee

H. Greenfield

J. Barnes

B. Burlingame

K. Cashel Contents

Preface...... xi

Ŷ Section I. National and International Food Composition 1 Programs Developing Comprehensive Policies and Programs for Improved Food Supplies and Nutrition...... 3 J.R. Lupien

The Food Composition Program of Indonesia: Past, Present and Future ...... 11 D.S. Slamet CHILEFOODS: Food Composition Activities in Chile and Latin America ...... 17 L. Masson Nutrient Composition of Wild-Gathered Foods from Mali...... 25 M.B. Nordeide, A. Oshaug, H. Holm

Ŷ Section II. Methods and Conventions of Nutrient Analysis 29 AOAC INTERNATIONAL-Validated Methods for Nutrient Analysis — Method Availability and Method Needs...... 31 J.W. DeVries Analysis and Classification of Digestible and Undigestible Carbohydrates...... 41 N-G. Asp Recent Developments in the Determination of Water-Soluble Vitamins in Food — Impact on the Use of Food Composition Tables for the Calculation of Vitamin Intakes...... 51 P.M. Finglas Update on the Analysis of Total Lipids, Fatty Acids and Sterols in Foods ...... 67 A.J. Sinclair Conventions for the Expression of Analytical Data ...... 75 D.A.T. Southgate

Ŷ Section III. Quality Control of Food Composition Data 83 and Databases Food Classification and Terminology Systems ...... 85 J.A.T. Pennington Nutritional Metrology: The Role of Reference Materials in Improving Quality of Analytical Measurements and Data on Food Components...... 99 J.T. Tanner, W.R. Wolf, W. Horwitz Strategies for Sampling: The Assurance of Representative Values ...... 105 J.M. Holden, C.S. Davis Assuring Regional Data Quality in the Food Composition Program in China...... 119 G. Wang, X. Li Quality Control for Food Composition Data in Journals — A Primer ...... 129 K.K. Stewart, M.R. Stewart

Ŷ Section IV. Information Needs and Computer Systems 139 The Future Information Needs for Research at the Interface Between Food Science and Nutrition...... 141 C.E. West Food Database Management Systems — A Review ...... 153 W. Becker, I. Unwin Data Identification Considerations in International Interchange of Food Composition Data ...... 165 J.C. Klensin Food Data: Numbers, Words and Images...... 175 B. Burlingame, F. Cook, G. Duxfield, G. Milligan Computer Construction of Recipes to Meet Nutritional and Palatability Requirements ...... 183 L.R. Fletcher, P.M. Soden Requirements for Applications Software for Computerized Databases in Research Projects...... 189 D. Mackerras

Ŷ Section V. Food Composition Data and Population 193 Studies

Food Monitoring in Denmark...... 197 A. Møller Food Composition Data Requirements for Nutritional Epidemiology of Cancer and Chronic Diseases...... 209 N. Slimani, E. Riboli, H. Greenfield Developing a Food Composition Database for Studies in the Pacific Islands ...... 217 J.H. Hankin, L. Le Marchand, L.N. Kolonel, B.E. Henderson, G. Beecher The Effects of Australian, US and UK Food Composition Tables on Estimates of Food and Nutrient Availability in Australia...... 225 K.M. Cashel, H. Greenfield Quality Control in the Use of Food and Nutrient Databases for Epidemiologic Studies...... 241 I.M. Buzzard, S.F. Schakel, J. Ditter-Johnson Construction of a Database of Inherent Bioactive Compounds in Food Plants...... 253 A.D. Walker, R. Preece, J.A. Plumb, R. Fenwick, B.K. Heaney

Ŷ Section VI. Copyright, Food Industry and Food Safety 255 Considerations International and Australian Copyright Considerations in Data and Data Compilations...... 257 S. Ricketson Non-Nutrient Databases for Foods ...... 275 K. Louekari Food Composition Databases in the Food Industry...... 281 O. de Rham The Databases of the Australian National Food Authority ...... 289 J. Lewis, S. Brooke-Taylor, F. Stenhouse Use of Databases for Nutrition Labeling in the United States...... 297 J.T. Tanner Functional Foods for Specific Health Use — The Needs for Compositional Data...... 305 K. Shinohara

The Second Food Data Base Conference 311 Preface

This Conference arose from the charge of IUNS Committee II/4 Techniques for Measuring the Value of Foods for Man (Chair: D.A.T. Southgate (UK)) “… to review techniques for the measurement of nutrients and other constituents in food … to improve and expand existing food composition data banks …”

The Conference was designed to recognize all aspects of food composition data production, management and use, and included papers, posters and computer demonstrations. All papers were invited in order to represent the diverse range of activities in the field. Sessions covered national and international food composition programmes, methods and conventions of nutrient analysis, quality control of food composition data and databases, a workshop on computer systems, food composition data and population studies, copyright considerations, and food industry and food safety considerations. These Proceedings comprise the invited papers and selected posters. The authors of all but two of the invited presentations provided papers which were sent to two referees each and revised in accordance with the referees' comments prior to acceptance for publication. Posters were not refereed. Thanks are due to the following who kindly reviewed manuscripts: K. Baghurst, P. Baghurst, D.H. Buss, K.M. Cashel, I. Coles-Rutishauser, J. Craske, J. Cunningham, J. DeVries, M. Filadelfi-Keszi, G. Greenleaf, W. Horwitz, J. Jin, M. Lawrence, S. Lee, J. Lewis, D. Mackerras, J. Klensin, R. Richards, G. Sevenhuysen, N. Slimani, D.A.T. Southgate, K.K. Stewart, A.S. Truswell, I. Unwin, J. Vanderslice, A. Walker, C.E. West and R.B.H. Wills. Special thanks are due to Jane Barnes for assistance with editing, Sharon Debrezceni for word-processing and Michael Wyatt of Keyword Editorial Services for preparation of the camera-ready text. Krystyna McIver of AOAC INTERNATIONAL kindly provided editorial advice.

Heather Greenfield Sydney January 1995 Section I

National and International Food Composition Programs

he Conference was opened by Senator the Honorable Graham F. Richardson, Minister for Health. The Session was chaired by Professor A. Stewart Truswell, TBoden Professor of Human Nutrition, University of Sydney, in his capacity as Vice- President of the International Union of Nutritional Sciences.

A keynote address was delivered by J. Lupien of the Food Policy and Nutrition Division, Food and Agriculture Organization, entitled Developing Comprehensive Policies and Programs for Improved Food Supplies and Nutrition, and this address was followed by papers by D. Slamet on The Food Composition Program of Indonesia: Past, Present and Future and L. Masson on CHILEFOODS: Food Composition Activities in Chile and Latin America, which are published on the following pages along with a print version of the poster entitled Nutrient Composition of Wild-Gathered Foods from Mali by M.B. Nordeide, A. Oshaug and H. Holm.

Posters displayed after Session I also were:

ƒThe Composition of Indian Restaurant Foods in Sydney, Bishop, C.G., Pratap, S.W., & Arcot, J., Department of Food Science and Technology, University of New South Wales, Sydney NSW, Australia. ƒTables of Composition of Australian Aboriginal Bush Foods, Brand Miller, J.C., James, K.W., & Maggiore, P., Human Nutrition Unit, University of Sydney, Sydney NSW, DSTO, Scottsdale, TAS, and School of Public Health, Curtin University of Technology, Perth, WA, Australia. ƒThe Food Composition Program of Nepal, Karki, T., Central Food Research Laboratory, Kathmandu, Nepal. ƒThe New Zealand Food Composition Database and Analytical Program, Burlingame, B., Arthur, J., Cook, F., Duxfield, G., Gibson, J., Milligan, J., & Monro, J., Nutrition Programme, NZ Institute for Crop & Food Research, Palmerston North, New Zealand. ƒThe UK Nutrient Databank, Buss, D.H., Corkill, M.J., & Holland, B., MAFF, 17 Smith Square, London, and RSC, Thomas Graham House, Science Park, Milton Rd, Cambridge, UK. ƒIngredient Composition of Australian Manufactured Foods, Cassidy, S., Dietitians' Association of Australia, Canberra, ACT, Australia. ƒDetermination of the “Pseudo-vitamins” in Dairy Products, Indyk, H., & Woollard, D.C., Anchor Products, Waitoa, and Ministry of Agriculture and Fisheries, Auckland, New Zealand. ƒNutrient Composition of Australian Ration Packs, James, K.W., Hancock, A.T., Coad, R.A., Sheedy, C.J., Lichon M.J., & Walker, G.J., DSTO, Materials Research Laboratory, Scottsdale, TAS, Australia. ƒCholecalciferol and 25-hydroxycholecalciferol Contents of Some Fish Species, Mattila, P., & Piironen, V., Department of Applied Chemistry and Microbiology, University of Helsinki, Finland. ƒA Nutritional Survey of Sydney Bread, Mugford, D., Griffiths, P., Walker, R., McGuirk, M., & Tomlinson, D., Bread Research Institute Inc., North Ryde, NSW, Australia. ƒNutrient Composition of South African Beef, Schönfeldt, H.C., Irene Animal Production Institute, Irene, South Africa. ƒThe Cholesterol and Fatty Acid Composition of South African Beef Carcasses, Schönfeldt, H.C., & Benadé, A.J.S., Irene Animal Production Institute, Irene, and Nutritional Intervention Programme, Medical Research Council, Tygerberg, South Africa. ƒComposition of South African Beef Carcasses, Schönfeldt, H.C., Naudé, R.T., & De Bruyn, J.F., Irene Animal Production Institute, Irene, South Africa. National and International Food Composition Programs

Developing Comprehensive Policies and Programs for Improved Food Supplies and Nutrition

John R. Lupien

Food Policy and Nutrition Division, Food and Agriculture Organization of the United Nations, Viale delle Terme di Caracalla, Rome 00100, Italy

In developing comprehensive policies and programs for improved food supplies and nutrition, governments need accurate information on food composition. The International Conference on Nutrition (ICN) 1992 emphasized food composition in its consideration of ways to improve household food security, prevent and control micronutrient deficiencies, assess, analyze and monitor nutrition situations and enhance food quality and safety. This paper summarizes the leading role FAO plays in international food and nutrient composition databases: developing global guidelines and standards; introducing food composition into countries' development activities; setting guidelines for laboratory facilities; disseminating standard techniques and references to improve analytical quality; supporting Codex Alimentarius, USDA and AOAC INTERNATIONAL; training in analytical techniques and data compilation; and documenting legal aspects of food composition. The potential role of FAO in collaboration with INFOODS and the Flair Eurofoods-Enfant Project in evaluating data quality and use; food nomenclature and foods to be included in nutrient databases; and criteria for setting food and nutrient data priorities is highlighted.

he International Conference on Nutrition (ICN) held in Rome from 5-11, 1992, encompassed nine themes, three of which are related to food composition work (1). TThese three themes were improving household food security, preventing and controlling specific micronutri-ent deficiencies and assessing, analyzing and monitoring nutrition situations. A government cannot make informed decisions regarding the food supply and meeting the food and nutritional needs of the population through an improved food supply without an understanding of the food that is eaten, particularly in the area of combating micronutrient deficiencies. Ŷ International Work in Food work related to sampling, analytical Composition techniques, and data compilation, An initial step in planning food supplies through publications and programs at is to assess the energy and nutrient country level; providing documentation situation by determining the composition on the legal aspects of food composition of the diet. For this purpose the Food data and their use in various and Agriculture Organization (FAO) jurisdictions which would assist the developed its expertise and reputation in documentation required for food international food composition with the analysis and data quality. development of the Food Balance A number of areas exist in Sheets. This work led to the production international food composition work that of regional food composition tables, the require future input and attention. These last being the Near East tables, include: evaluation of data quality and published in by FAO in 1982 (2). the need to relate the quality of the data FAO is committed to maintaining its to the use for which they are intended; leadership role in the area of food nomenclature and foods to be international food and nutrient included in nutrient databases; criteria composition databases. To this end a for setting food and nutrient data two-day meeting was held from 14–15 priorities. December 1992, hosted by FAO in As part of the follow-up activity to the Rome, with additional funding ICN, FAO is initiating work at the assistance from INFOODS, a UNU developing country level in the areas project, and the Flair EUROFOODS- mentioned above. In addition a joint Enfant Project (funded by the European FAO and UNU meeting in Tunis in Community). The participants at this March, 1994, has evaluated the meeting and subsequent forums progress in international food welcomed FAO's renewed activity in composition activities since the Bellagio international food composition and meeting in 1983 (3) and formulated a recognized its comparative advantage in long-term program of work suitable for developing worldwide guidelines and attracting outside funding and standards as well as the opportunity to participation. introduce food composition into the Ŷ Policies and Programs to Improve overall development activities of the Nutrition country. Specifically FAO can be instrumental The title of this paper suggests that in setting guidelines for improving policies and programs in many countries planning for laboratory facilities; are not comprehensive enough to improving analytical quality through the improve nutrition. In fact better dissemination of standard techniques nutritional status depends on the kind of including inter-laboratory tests and coordination across agricultural, health, reference materials; supporting the work educational, trade and development of Codex Alimentarius, AOAC policies, that is rarely achieved. In part INTERNATIONAL and national this situation occurs because the link laboratories involved in determining the between these policies and improved composition of foods in identifying nutritional status is not clear. Of course protocols for suitable, yet cheaper all countries have policies that affect analytical techniques; supporting the nutrition in some way, and the challenge education and training for composition is to find ways that enhance the piecemeal effects of separate initiatives many and varying views. However, after into a larger sum. extended discussions it was clearly There is agreement on a number of agreed that proper nutritional status general policy orientations to improve depends on the effective preparation nutritional status. For example, and implementation of a wide range of agricultural policies must be oriented agricultural, health, educational, trade towards rational and effective and development policies, carried out by development of better food supplies, a variety of government agencies, including the production, processing and academia, industry and the public at effective marketing of all of the elements large. of an adequate and nutritionally During the ICN, discussions were balanced diet. Health policies must give organized according to the following specific attention to a wide range of nine themes: preventive activities, such as ƒincorporating nutritional objectives, immunization, vitamin and mineral considerations and components into supplementation of vulnerable groups development policies and programs such as infants and pregnant or nursing women, and effective treatment of ƒimproving household food security diarrhoeal diseases. Agricultural and ƒprotecting consumers through health policies must stress the improved food quality and safety assurance of adequate food quality and ƒpromoting breast-feeding safety from the point of production, through harvesting, storage, ƒpreventing and managing infectious processing/preservation and marketing. diseases Educational policies must assure ƒcaring for the socio-economically adequate basic education for all, and deprived and nutritionally vulnerable include appropriate elements of nutrition ƒpreventing and controlling specific education for all, in elementary and micronutrient deficiencies secondary schools, and for use in the mass media. ƒpromoting appropriate diets and healthy lifestyles Ŷ International Conference on assessing, analyzing and monitoring Nutrition ƒ nutrition situations. The extent to which these policies These themes form the basis for reinforce the separate impacts they vertical and potentially free-standing have will determine the kind of activities. They are in effect the improvement in nutritional status that essential elements of policy which must may be expected. In this context it is be considered, formulated and important to discuss the International implemented in each country. A number Conference on Nutrition (ICN) which of cross-cutting, or broad, issues was held in Rome during December discussed at the ICN provide important 1992. Well before the Conference integration of policies and activities. started, the two sponsoring agencies, These issues included education, the Food and Agriculture Organization environmentally sound and sustainable (FAO) and World Health Organization development, gender and population (WHO), met together to decide which concerns, and resource allocation in the were the basic policy elements and implementation of specific nutrition activities necessary to improve nutrition improvement activities. for all and to define the objectives of the In discussing these general and ICN in this regard. Needless to say, the specific themes it became clear that issues were complex and expressed in implementing effective nutrition improvement activities needs to be inter- and 1970's for use in regions such as sectoral and inter-disciplinary, if Latin America, Asia, the Near East, and sustained success is to be attained. Africa (4, 5, 6). The work was carried out in collaboration with the US Ŷ The Need for Food Composition Data Department of Agriculture, the US Department of Health and Human The fundamental importance of Services, and the Institute of Nutrition of knowledge about the composition of Central America and Panama (INCAP) food in the themes discussed at the ICN in Guatemala. Following the production was recognized by FAO, and the Food of these publications FAO and its Food Policy and Nutrition Division has Policy and Nutrition Division de- reactivated work on food composition as emphasized international activities on one part of a number of activities food composition for several years and initiated to follow-up on the ICN other agencies increased their activities resolutions. FAO is currently committed in this area. to working with all of the agencies and Important progress in food professionals active in food composition composition has been made over the in an effort to make more reliable data last ten years, in large part due to the available worldwide. International Food Data Systems Food composition work does not only Project, or INFOODS. Their publications depend on the global ICN objectives, contain the basis for world-wide but is in a real sense needed to support standards in the development of food projects and interventions to improve composition development (7, 8, 9, 10). nutrition. Information on the nutritional But perhaps an even more crucial composition of food is essential for the contribution by INFOODS is the regional provision of adequate and appropriate system of data generation and diets for individuals and populations. In management for food composition data. this context food composition data are a Relying on a few key institutions around major requirement for action in nutrition the world that are capable of education, interventions on micro- coordinating a variety of food nutrients, nutritional support for health composition activities on a regional care plans, food trade, food labeling and basis, INFOODS has greatly expanded regulations, and the integration of the support for food composition work in nutrition concerns in agricultural developing countries. FAO intends to policies. cooperate fully with INFOODS to Given that this variety of actions is participate in, and build on, such implemented by a wide range of regional networks in future activities. ministries and agencies, the uses and At the same time, the EUROFOODS limitations of data on food composition group has achieved exemplary need to be clear. A large number of improvements in food composition data users therefore require reliable and used in Europe as a result of improving unambiguously labeled data and FAO the quality and compatability of the data can make a particular contribution in this from many sources, cultural area. In addition, the common use of environments and jurisdictions. The data in different policy arenas will help in experience of EUROFOODS some ways to promote the kind of demonstrates the procedures which integration of action referred to earlier. may be used in promoting the Ŷ Collaboration in Food Composition collaboration of countries in other regions of the world. FAO published a number of international Particularly important are the food composition tables in the 1960's advances of EUROFOODS in data quality related to food nomenclature and media, and utilization of reliable data food coding systems that were in a wide range of nutrition developed to integrate the national food improvement activities, projects and composition databases of European programs countries. For example, the names of ƒto set standards for minimum foods in different European countries laboratory environments required in may be the same, but refer to products developing countries to ensure with a different formulation. Conversely, acceptable data quality and different food names may refer to the comparability, and promote regional same food, but express differences in efforts for such standards culturally determined food preparation. The achievement of the EUROFOODS ƒto provide a forum for exchange of group in maintaining data quality in the information between professionals face of such diversity has provided on new developments and solutions those who are faced with integrating to common problems food composition databases in their own ƒto create a world-wide directory of regions with important examples and sources, quality and availability of lessons. food composition data, where actual compositional data will be kept in a Ŷ FAO's Strengthened Food large number of national or regional Composition Program centers The overall purpose of FAO's program to monitor the regulatory and legal in food composition is to promote the ƒ ownership aspects of food generation and dissemination of reliable composition data and their use in food composition data that meets the food labeling and food standards, in needs of local users, or at least at order to assist the documentation national level. The work itself will be required for food analysis and data implemented in collaboration with other quality. agencies, institutions and groups, In order to achieve these objectives, according to the following program FAO has several ongoing activities and objectives: will start new ones. The most important ƒto promote and expand existing of these include: activities at international, regional Support for new analytical work in and national centres active in food ƒ developing countries in particular, composition work, such as data where conditions for generating data generation, data management, and are difficult. For example, assistance the distribution of data to users to the Ethiopian Nutrition Institute for ƒto publish standards on terminology laboratory upgrading, staff training required for the identification of food and the preparation of a new food and nutrients, sampling procedures composition database is currently for food, requirements for handling being organized. Such work is food samples, analytical carried out in close collaboration methodology, assessment criteria for with laboratories at Wageningen data quality Agricultural University in The ƒto support training of staff involved Netherlands. with all aspects of food composition ƒFormulation of criteria for the work, including the collection, assessment of data quality and data preparation and analysis of samples, sampling. The activity requires as well as the management of extensive communication with resulting data using electronic laboratories in many countries to find the optimum detail to be included. ƒLastly, FAO and the UNU organized FAO is working closely with USDA a meeting to discuss the progress of where staff are producing expert worldwide food composition systems for the evaluation of data activities over the last ten years and quality. identify a program of work for the ƒFor a better view of the active years to come. The meeting held in institutions in the field, and the type Tunis at the end of March 1994 is of compositional data they generate seen as the first major update on and manage, FAO is collecting activities since the 1983 meeting on details for a “Directory of institutions” food composition in Bellagio (3). The and a “Directory of databases”. Tunis meeting provided an Again, the FAO effort in this area opportunity to share issues related complements the work organized by to food composition work in many others and close collaboration with countries and identify ways to INFOODS and EUROFOODS is support the activities. Participants maintained. considered criteria that are ƒFAO recognizes that a major appropriate for selecting and revision is needed for the food managing food composition data composition data used to calculate (11). With the recent changes in the nutrient content of national food analytical techniques, knowledge of supplies for each of the countries for the sources of error in nutrient which FAO Food Balance Sheets values, and electronic means for are published. In cooperation with sharing data, the meeting came at the FAO Statistics Division, which an appropriate time to take stock publishes the Food Balance Sheet and plan innovations for the next databases, the Food Policy and decade. Nutrition Division is helping with a Ŷ Changes in Food Composition major update of the system and data Data values. The work will give us new It is interesting to note that over time and more reliable estimates of the food composition data have changed nutrient supply in the near future. not only because of improved analytical ƒThe FAO Food Policy and Nutrition techniques and better knowledge of Division actively supports training for representative sampling, but also food composition related activities, because of real changes in the such as learning new analytical formulation of foods. For example, the techniques or modifications of plethora of low fat/low energy products techniques for different product with sensory characteristics of full fat groups, as well as learning computer formulations, have the potential to lower based systems for the management the fat intakes of people in many and storage of compositional data. developed countries. Or, the dwindling For example, FAO support for choice in variety of potatoes in several training workshops at the developing countries since the last Wageningen, Agricultural University, century to leave only a few commonly where participants from developing used ones, which reduces the variability countries have the opportunity to in nutrient intake from potatoes and can increase their skills in analysis. The reduce the food security for several Division expects that more training population groups. opportunities will become part of the We know some of the factors that program of work. determined these changes, as well as others. These factors demonstrate the interrelationships between the (2) FAO (1982) Food Composition agriculture, health, education and trade Tables for the Near East, Rome that can potentially affect nutritional (3) Rand, W., & Young, V. (1983) Food status. The positive effects should be Nutr. Bull. 5, 15–76 the objectives of interventions at any (4) INCAP-ICNND (1961) Food level of government or international Composition Table for Use in Latin cooperation. The detrimental effects of America, Interdepartmental such changes need to be avoided or Committee on Nutrition for National minimized, especially for disadvantaged Defense, Washington, & Institute of groups. Nutrition of Central America and Panama, Guatemala (5) FAO (1970) Food Composition Tables for Use in Africa, Rome Ŷ Conclusion (6) FAO (1972) Food Composition The integrated nature of actions to Tables for Use in East Asia, Rome improve nutrition shows the wide range (7) Klensin, J.C., Feskanich, D., Lin, V., of coordinated work which needs to be Truswell, A.S., & Southgate, D.A.T. done and the importance of food (1989) Identification of Food composition activities to overall efforts. Components for INFOODS Data Nutrition problems cannot be resolved Interchange, UNU Press, Tokyo by nutritionists alone, or by the use of (8) Rand, W.M., Pennington, J.A.T., one policy strategy or one program Murphy, S.P., & Klensin, J.C. approach. Each initiative intended to (1992) Compiling Data for Food improve nutrition needs to be based on Composition Data Bases, UNU the theory and experience appropriate Press, Tokyo to it, and implemented with explicit (9) Klensin, J.C. (1992) INFOODS expectations for nutritional Food Composition Data improvement. The development of Interchange Handbook, UNU Press, reliable, internationally comparable food Tokyo composition data is one crucial element (10) Greenfield, H., & Southgate, D.A.T. in achieving this integration and FAO is (1992) Food Composition Data. fully committed to that goal. Production, Management and Use, Ŷ References Elsevier Applied Science, London (11) FAO & UNU (1994) Report of (1) FAO & WHO (1992) Final Report of Discussions on Food Composition the International Conference on for Developing Countries (in press) Nutrition, Rome National and International Food Composition Programs

The Food Composition Program of Indonesia: Past, Present and Future

Dewi S. Slamet

Nutrition Research and Development Center, Bogor, Indonesia

Indonesia is a country of diverse geography, culture and dietary patterns. The national food and nutrition policy aims to achieve food self-sufficiency, and diversification of food supplies and food consumption. For this reason the need for complete, accurate and up- to-date food composition tables has been recognized. This paper reports the progress towards this goal.

ndonesia is an archipelago of some 13,000 islands with an aggregate land area of 1,900,000 km2 (Figure 1). The population of Indonesia is about 180 million, of which Iabout 80 per cent live in rural areas. Thousands of kinds of foods are available in Indonesia. The staple foods are rice, maize, sago, cassava and sweet potato, and the diet can be classified according to three distinct consumption patterns, namely (a) rice pattern (West Java, Sumatra, Kalimantan); (b) rice and maize pattern (Central and East Java, Sulawesi, Bali, Nusa Tenggara Islands, with cassava being consumed especially by the low income groups in all these regions); (c) sago and sweet potato pattern in Maluku and Irian Jaya (1, 2).

The Food and Nutrition Policy of the 1988, was made to assist this aim. This Fifth Five Year Nutritional Development recommendation embraced the need to Plan (1989–1994) is aimed towards improve and develop information achieving food self-sufficiency and the available to nutritionists including food diversification of the food supply and composition tables, so that nutrition consumption, i.e. away from total education directed towards broadening reliance on rice. A recommendation food choice could be carried out. Thus arising from the Fourth National Indonesia fully recognized the need for Workshop on Nutrition in Jakarta, June complete, accurate and up-to-date food composition data as an important tool in nutrition research and activities in the developing the entire range of food and country.

Figure 1. Map of Indonesia several nutrients: moisture, protein, total Ŷ Historical Background of Food Composition Tables in Indonesia fat, carbohydrate (by difference), calcium, phosphorus, iron, vitamin A, The first food composition tables (FCT), carotenes, thiamin, ascorbic acid and in Indonesia were produced by the total energy. The 483 food items Institute of Volksvoedings (1930–1940), included 10 per cent snacks, 5 per cent during the Dutch occupation, from the fermented foods, and 1 per cent dried, data available at that time. In the period salted fish. Only 116 food items were between the liberation of Indonesia in analyzed for amino acids and niacin (3, 1945 up to 1967, the name of FCT of 4). Indonesia changed several times. Since 1967 the FCT of Indonesia were called Ŷ Recent Activities in Generation and Daftar Komposisi Banan Makanan Compilation of Indonesian Food (DKBM), compiled by the Directorate of Composition Data Nutrition Department of Health and Food analyses are also carried out at published by Bhratara Jakarta (3rd the following institutions such as the printing, 1981). Most of the analytical Indonesian Institute of Science, the data were generated by the Nutrition National Institute for Chemistry, the Institute since 1950 and some were Department of Agriculture, the taken from other international FCT. The Department of Trade, Department of tables contained 410 food items of Industry, the Atomic Energy Agency and which 283 were raw foods. private laboratories (e.g. food industry). From 1970 till the present time the The analyses carried out in these Nutrition Research and Development laboratories are limited by their functions Centre (NRDC), Department of Health, and their facilities. In many cases the Bogor, has been the only institution in analytical data produced cannot be used Indonesia conducting the analyses of for food composition tables since the the nutrient composition of foods, with data cannot easily be related to the food the specific objective of producing the as consumed. national Indonesian FCT. Analytical data To generate and compile the food for 483 food items, raw, processed and composition data to produce Indonesian prepared, obtained from various regions FCT, a meeting between all interested of Indonesia, have been produced for parties was needed. Therefore in 1984, to tackle the problem of food objectives were adopted such as to composition data, a meeting was exchange and share information on organized by the NRDC in Bogor, current trends in food composition data involving all the various government generation and compilation, and to institutions which had a special interest develop a plan to standardize sampling, in food composition data. The meeting methods of analyses and compilation of was attended by about 19 food composition data in the ASEAN representatives from 12 institutions. The countries. meeting agreed that: the available Further regional progress was made DKBM should be revised to fulfil the at the next ASEAN workshop on food increasing needs related to the Nutrition data systems held in Bangkok, 25–27 Program; the name of the DKBM should October 1989, funded by Japan. The be changed to Komposisi Zat Gizi Indonesian representative reported the Pangan Indonesia (KZGPI); and the current activities of the food composition tables should contain nutrients such as program in Indonesia (8). These proximate constituents, minerals and activities continued under national vitamins, analyzed per 100 g edible guidance, the most recent being a small portion. Priority should be given to those workshop on nutritive composition of foods most commonly consumed by the foods conducted by the Nutrition people. These data would be most Research and Development Centre at needed for the Food and Nutrition Bogor, March 22–24, 1990 (9). Program (5). Ŷ Current Status Similar meetings were organized also by the Research and Development A collaborative food composition Centre for Applied Chemistry (LIPI) in program has recently been carried out 1986 in Bandung, to review the existing between the NRDC and the Directorate problems which included updating the of Nutrition. The data obtained were existing food composition data, compiled together with the food development of food composition data analytical work previously done by systems, quality assurance programs for NRDC from 1970 to 1989 and published food analyses and a food composition in journals (10–17). This new set of network (6). KZGPI was published in 1990 by the Also of importance was a workshop Directorate of Nutrition and NRDC (18). held under the auspices of the These tables contain nutrients for raw, Association of South East Asian Nations processed, traditional and fast foods. (ASEAN) Sub-committee on Protein and The foods were analyzed for proximate Food Habits Research and composition, minerals and vitamins. The Development in Jakarta, 20–23 October, table is divided according to food 1986, and attended by representatives groups: Table A, the nutritional from Indonesia, Thailand, Malaysia and composition of cereals, tubers, nuts, the Philippines. This technical workshop legumes, vegetables, fruits, eggs, fish on food composition data was an initial and miscellaneous (total of 281 raw food activity towards the development of the items); Table B, the nutrient composition ASEAN Food Data Network (7). The of processed and fast foods, (total of workshop was an important step 153 food items); Table C, riboflavin and towards a united effort to systematize, niacin content of foods (total of 142 food standardize and update the generation items); Tables D and E, essential and and compilation of food composition non-essential amino acids of 71 food data in the ASEAN countries and to items (18). facilitate the interchange and use of the Advice on compiling FCT, preparing data throughout the region. Special a manual of food composition analyses, and equipping the laboratory for future employed men and women have their work was provided by Associate meals outside of the home, and the Professor Heather Greenfield, from quality of food consumed depends on Australia, who was invited by the the environment and the capacity to buy National Institute for Health Research food. Nutritional information about pre- and Development as a consultant. The prepared and street foods is in great revised and expanded manual was demand. published in 1990 by the Directorate of According to the Department of Nutrition and NRDC (19), and includes Health, Indonesia currently still has four information on food sampling, sections main nutritional problems, i.e. vitamin A on direct analysis of sugars, starch, total deficiency, protein energy malnutrition, dietary fibre, fatty acids and cholesterol, iron deficiency and iodine deficiency. as well as a chapter on quality These are still prevalent in the poorer assurance. areas of eastern Indonesia which Presently, there are three FCTs require more urgent attention. At the available in Indonesia: DKBM (20); same time as changes in environmental KZGPI (18); and DABM. The DABM and economic conditions occur, other (Daftar Analisa Bahan Makanan) has nutritional problems are also arising in been published by the Medical Faculty Indonesia such as degenerative disease of the University of Indonesia, Jakarta (coronary heart disease, diabetes, and is comprised of data compiled hypertension and cancer), possibly mainly from international FCT (21). related to the dietary changes caused by alterations in food habits. Processed Ŷ Future Program foods might be higher in animal fats, Indonesia has recognized that the past cholesterol, sugars, salt, and food and present available data of food contaminants (including potentially composition are inadequate both in carcinogenic substances) than terms of food items and nutrients and traditional foods. Knowledge of the other food constituents. Furthermore in nutrient composition of foods forms the the last decade, Indonesia has grown backbone of clinical therapeutic diets. rapidly in the realm of research and These are important in the management technology, with particular emphasis in of diseases such as hypertension, heart the field of food technology, resulting in disease and diabetes. Without a rapid development of the local food knowledge of nutrient composition data industry. These food manufacturers dietitians and nutritionists in government produce processed foods, fast foods, health agencies would be unable to and new food formulas. The assess the adequacy of patients' diets development of new techniques in the and the nutrient intake of the people. production and processing of foods and The FCT would be used to calculate the the availability of a demand for new food nutrients of a typical daily intake, and products have created dramatic based on these calculations nutrient changes in the food consumption intakes could be compared to the pattern of the population, particularly in Recommended Dietary Intakes (RDI). If urban areas. These situations may have there appears to be a tendency of low or a negative or positive effect on the excessively high intake of any nutrient, nutritional status of the Indonesian advice and diet plans could be supplied people. by the dietitian or the nutritionist. Because of the need to provide A new project of the Nutrition family income, men and women work Research and Development Centre will long hours and may not have enough analyze the macro-nutrients, micro- time to prepare food at home. Most nutrients and fatty acids of various food items (especially traditional foods, fast Data Systems, Bangkok, pp.104– foods and foods from marine resources) 108 in collaboration with other research (9) Slamet, D.S. (1990) in Workshop of institutions. This project will produce Indonesian Food Composition new information on Indonesian foods, Tables, Bogor e.g. sugars, dietary fiber and Ȧ-3 and Ȧ- (10) Slamet, D.S., & Purawisastra 6 fatty acids. The project will require (1979) in Proceedings of Food additional equipment (such as LC and Technology Meeting, Jakarta, pp. GLC) and more trained food analysts, 158–175 as well as a high degree of collaboration (11) Slamet, D.S., & Komari (1985) between laboratories. Media Teknologi Pangan 1, 56–60 In the future, as recommended at the (12) Slamet, D.S., & Komari (1986) 5th National Nutrition Conference held Penelitian Gizi dan Makanan 9, 63– in Jakarta, April 20–22, 1993, the 76 government has decided that Indonesia (13) Slamet, D.S., & Komari (1986) should possess an up-to-date Penelitian Gizi dan Makanan 9, 77– Indonesian national food composition 84 table incorporating all the latest data. In (14) Slamet, D.S., & Ubaidillah (1987) the last decade there has been Penelitian Gizi dan Makanan 10, increasing interest in food composition 77–81 data in relation to diets, food habits and (15) Slamet, D.S., & Ubaidillah (1988) degenerative diseases. The next FCT Penelitian Gizi dan Makanan 11, will be critical to the success of projects 59–73 in these areas as well as other aspects (16) Slamet, D.S., Komari, & Ubaidillah of the Food and Nutrition Program. (1988) Penelitian Gizi dan Makanan 12, 58–71 Ŷ References (17) Study on Nutritive Composition of (1) Karyadi, D., & Hermana (1985) in Foods (1990) Directorate of Proceedings of the First Asian Nutrition and Nutrition Research Foods Conference, Bangkok, pp. and Development Center, Bogor 56–58 (18) Mahmud, M.K., Slamet, D.S., (2) Lie, G.H., Hermana, Suwardi, & Apryantono, R.R., & Hermana Ismyati, S. (1978) Proceedings of (1990) The Indonesian Food First ASEAN Seminar - Workshop Composition Table, Directorate of on Food Habits, Manila, pp. 34–39 Nutrition and Nutrition Research (3) Greenfield, H. (1991) Study of and Development Center, Bogor Nutritive Composition of Foods in (19) (19 Slamet, D.S., Mahmud, M.K., Indonesia, World Health Muhilal, Fardiaz, P., & Sumarmata, Organization SEARO, New Delhi J.P. (1990) Manual of Food (4) Slamet, D.S., & Tarwotjo (1980) Analysis, Directorate of Nutrition Penelitian Gizi dan Makanan 4, 21– and Nutrition Research and 23 Development Center, Bogor (5) Workshop on Food Composition (20) Directorate of Nutrition, Department Data (1984) Bogor of Health (1987) Food Composition (6) Slamet, D.S. (1986) in Workshop on Table, Bhratara Karya Aksra, Food Composition Data, Bandung Jakarta (7) Workshop for the ASEAN Food (21) Oei, K.N. (1992) Nutrient Analysis Composition Table, (1986) Jakarta Tables for Food, University of (8) Sumardi (1989) in Proceedings of Indonesia, Jakarta the ASEAN Workshop on Food National and International Food Composition Programs

CHILEFOODS: Food Composition Activities in Chile and Latin America

Lilia Masson

University of Chile, Department of Food Science and Chemical Technology, Casilla 233, Santiago 1, Chile

This paper describes the history and coverage of the food composition tables of Chile The computerization of the most recent tables is discussed, together with the development of software packages to access and use the database. The links with food composition activities in other Latin American countries, via the LATINFOODS network, are also covered.

hile is a modern country situated in the South West extreme of South America. It can be considered the longest country in the world with more than 4,000 km of CPacific Ocean coastline. It is narrow with a mean width of about 180 km. The total area is estimated as 1,992,000 km2; 740,000 km2 represents the continental area plus different islands including Easter Island while 1,250,000 km2 corresponds to the Antarctic territory (Figure 1).

The total population is about 13 are high at 92 per cent. The Chilean million, most of whom are concentrated population is a mixture of people from in the central zone of the country. The Spain and other European Countries Metropolitan Region, where the capital together with people of indigenous Santiago is located, has about 5 million ancestry. The latter are mainly in the residents, representing about 40 per Araucania Region (about 200,000 cent of the total population. Araucanos or Mapuches) and the First Chile is mainly an urban country, with Region close to the border with Bolivia about 87 per cent of the total population and Perú (Aymará natives), while the living in cities and the remaining 13 per inhabitants of Easter Island are mainly cent living in rural areas. Literacy levels Polynesian. In South America the countries of Chile, Argentina and Uruguay have a mainly European ethnic influence, while in the other Latin American countries, a high predominance of people of indigenous or other ethnic origins is found. Latin America has supplied many of its different native foods to the rest of the world. The 18 native foods from Latin America which have found their way around the world are: maize, tomatoes, aji or chilli, palta or avocado, beans, potatoes which are from Perú and Chile, lupin, quinoa, tapioca, yuca, pineapple, cocoa, banana, coconut, lucuma, chirimoya, papaya (the last three from Ecuador, Perú and Chile) and strawberries from Chile. All of these foods are still very important in the Latin American diet and are an integral part of the Chilean diet, with the exception of yuca and tapioca which are confined to the Amazonian Region. In general the Chilean diet is homogeneous throughout the country, with typical meals being prepared from maize, beans, potatoes, tomatoes and chilli. The main cereal in the Chilean diet is wheat, with a high consumption of different kinds of bread, pasta and noodles. Beef, poultry, pork, fish and shellfish, eggs, and milk and derivatives are the most important sources of dietary protein. Raw and boiled vegetables are always present in the Chilean diet as are fresh fruits, both groups being good sources of vitamins and minerals. For more than 30 years, wheat flour has been enriched by law with thiamin, riboflavin, niacin and iron. Long-standing government policies have been in force in Chile to promote good nutrition for infants, pre-schoolers and school children. Different programs to cover the main nutritional requirements of these groups have been established and thanks to these actions, which have been maintained for many years, it has been possible to reduce levels of infant Figure 1. Chile regions malnutrition in the country to low levels of prevalence at about 5–6 per cent. These different published editions of The situation is not as favorable in other the tables have permitted Chilean Latin America countries, which still seek professionals not only in the health field, adequate solutions for this problem. from government and private agencies, In the Chilean food composition but also in the food technology and database system the description of the education areas, to obtain current foods and meals has received comprehensive national information considerable attention, including the about the nutritive value of the main main ingredients used, and the English foods normally included in the Chilean language name. This helps the diet. understanding of these local The different food items were preparations for other users in Latin organized in 15 groups: America or elsewhere. Of course, in ƒmilk and its derivatives other Latin countries there are many different meals based on their native ƒavian and wild animals foods that will be included in their own ƒfish and shellfish (53 different fish food databases. and shellfish species, including low fat fish with less 1 per cent fat, e.g., Ŷ Chilean Food Composition golden congrio (Genypterus Activities blacodes), hake or merluza Activities in the field of food composition (Merluccius gayi), corvina have been maintained in Chile for over (Micropogon furnieri); semi-fat fish 30 years, initiated by Dr Hermann with about 3 per cent fat, e.g., Schmidt-Hebbel and continued by the albacora (Xiphias gladius), pejerrey group of food chemists belonging to the (Odeteshes regia), reineta Department of Food Science and (Lepidotus australis); and fatty fish, Chemical Technology at the University e.g.., mero (Dissosticuus of Chile. This permanent work has eligenoides) with 20 per cent fat, permitted the publication of eight cojinoba (Seriolella caerulea) 13 per editions of the Chilean food composition cent fat; Spanish sardine (Sardinops tables (1). The most recent edition was sagax) 10 per cent fat published in 1990 (2), listing more than shell fish, mainly natives, e.g., piure, 400 different food items, including data ƒ (Pyura chilensis), macha for the proximate analysis, energy (Mesodesma donacium), loco content, and major minerals and (Concholepa concholepas), sea vitamins. Supplementary tables are urchin gonads (Lexoxhinus albus) included for the various fractions of dietary fiber in legumes and cereals; ƒfats and oils retinol and cholesterol in different foods; ƒcereals and derivatives (including fluoride in some beverages, seafoods quinoa (Quenopodium quinoa) a and fruits; some trace minerals in native with a higher protein content vegetables; and amino acids. Other at 13 per cent than wheat, maize or complementary tables are also included, rice; together with 17 different e.g. energy conversion factors; essential varieties of bread, some flours, amino acids in different foods compared pasta and spaghetti and ten varieties with the provisional amino acid of cookies combination; servings weights and their legumes, including beans, peas, equivalence in grams or millilitres; and ƒ lentils, and three varieties of lupin the US Recommended Dietary seeds (Lupinus albus) Allowances (3) ƒvegetables, including native about structure; physiological roles; seaweeds and mushrooms families derived from essential linoleic and linolenic acids; and their ƒfruits, including close to 50 different requirements. The recommended species, produced for local dietary relationship between the three consumption and export, comprising groups of fatty acids is also discussed introduced species, e.g., kiwi fruit as well as the potentially adverse effects (Actinia chinensis); native or that could be produced by excessive characteristic fruits, e.g., chirimoya consumption of polyunsaturated fatty (Annona cherimolia), lucuma acids, and additional recommendations (Lucuma abovata), sweet pepino for vitamin E intake according to the (Solanum muricatum), and Chilean amount of polyunsaturated fatty acids papaya (Carica papaya); and, wild (PUFA) in the diet. fruits and seeds e.g., rosa mosqueta Five groups of different fats and oil (Rosa moschata mill), arrayan are discussed in relation to the main (Mireengenella apiculata), maqui fatty acids present in their structure. (Aristotelia chilensis), murtilla (Ungi They are summarized in five tables, molina), Chilean hazelnut (Guevuina according to the following distribution: avellana), and piñon (Araucaria Table I, Vegetable fats and oil with less imbricata) 40 per cent linoleic acid; Table II, ƒsugar and derivatives including fruit Vegetable oils with more than 40 per jams, honeys, instant desserts, cent linoleic acid; Table III, Fats of juices, chocolates, cakes, baked animal origin; Table IV, Fats and oils of specialities, jellies marine origin; Table V, Hydrogenated ƒbeverages comprising non-alcoholic, fats and oils. e.g., carbonated beverages and In total 62 different fats and oils with natural fruit juices, and alcoholic their respective fatty acid compositions beverages, e.g., different varieties of are tabulated. The fatty acid white and red wine and beers composition of the fat extracted from different native seeds, is also included, ƒmiscellaneous, e.g., salad dressings e.g., Chilean hazelnut or avellana and sauces, spices, different ready- (Guevuina avellana), mayu (Soffora to-eat meals, and snacks macrocarpa), Rosa mosqueta (Rosa ƒspecial dietary foods, e.g., infant moschata mill), quinoa (Quenopodium formulas, breakfast cereals, baby quinoa), pelu (Sophora tetraptera sensu foods R), tamarugo (Prosopis tamarugo phil), ƒothers, e.g. yeast, seasoning tablets. maracuya (Passiflora edulis), cardo Another activity in the field of food (Cynara cardunculus), and the fruit of composition has been the production of the Chilean avocado (Persea a publication in Spanish entitled Fats gratissima). In general, the linoleic acid and Oils Habitually and Potentially content is high and special mention Consumed in Chile: Fatty Acid should be made of Rosa mosqueta seed Composition (7). This issue contains oil which has about 43 per cent linoleic data obtained by the Food Chemistry acid and 35 per cent linolenic acid. Laboratory, University of Chile, for the Among the Ȧ-3 fatty acids present in main fatty acids present in the different different seafoods of Chilean origin, fats and oils. General information about eicosapentaenoic acid C20:5-Ȧ-3 (EPA) saturated, monounsaturated and and docosahexaenoic acid C22:6-Ȧ-3 polyunsaturated fatty acids is given. The (DHA) are the most important. For publication also includes explanations example, in jurel (Trachurus murphyi) DHA is about 25 per cent of the total fatty acid methyl esters and EPA about First, the number of nutrients that 10 per cent, while piure (Pyura should be included in the system was chilensis) has EPA present at a higher decided. A total of 171 nutrients was percentage compared with DHA selected, then coded and classified into This kind of work has continued and nine groups: proximate analysis, amino in recent years new items of vegetable, acids, vitamins, minerals, fatty acids, animal and marine origin have been sterols, carbohydrates, dietary fiber and studied, which will be incorporated in special constituents. A procedure for forth coming editions of this publication food classification and coding was also and in the database system. studied, and 14 groups were designated: Ŷ Food Composition Data Network in Latin America 01 Milk and milk products 02 Eggs and derivatives Due to its strong national food 03 Meats and derivatives composition program, Chile was invited 04 Avian, wild animals and derivatives to participate in the LATINFOODS 05 Finfish and shellfish products organization (8,9), whose second 06 Fats, oils and derivatives meeting was held in Santiago in 1988 07 Cereals and derivatives with the participation of many Latin 08 Legumes, seeds, oily fruits and countries. derivatives The main recommendation from this 09 Vegetables, seaweeds and meeting was that each country should derivatives start with the design of a local database 10 Fruits and derivatives system. In the meantime it was decided 11 Sugar products to organize CHILEFOODS as a branch 12 Beverages of LATINFOODS and different groups of 13 Miscellaneous, soups, sauces and Chilean experts in food chemistry and derivatives nutrition, mainly from universities, 14 Special dietary foods. government agencies and scientific A range of parameters to identify the societies were invited to participate. The different foods was also considered. group started to create a computer Other variables were: identification of system able to support efficient handling the samples, the analytical procedures of information about the composition of employed and the source of information. food produced in Chile. A short code of eight digits was chosen, In order to fulfil these objectives, two digits to indicate the food group, two suggestions from LATINFOODS were digits for the sub-group and the other considered (10). Composition tables four digits to identify the food. from other countries and the Eurocode 2 In order to individualize a specific system from Eurofoods-Enfant for food food, four parameters were considered: classification were reviewed. In addition common name, synonyms, scientific through a series of CHILEFOODS name and English language name; meetings, the different groups of weight and type of individual serving Chilean experts in the field of food were also considered. composition contributed their opinions Printed forms were developed to and suggestions for improving the facilitate one of the main tasks of project. To design the system it was LATINFOODS and CHILEFOODS, the necessary to identify the output and compilation of data about food input requirements, and then the composition. These forms are being archival and procedural specifications distributed among information were identified. generators and compilers. A total of 14 tables will gather all the necessary information. Along with the forms, a user databases that does not present handbook was produced with learning difficulties. To run this program instructions for their correct use. it is possible to use any computer The program was created in compatible with PC-DOS or MS-DOS in CLIPPER (11), a language for handling 3.3 version or later. databases and screens for program The program was called display. To run this program it is PROARCAN, since the evaluation of possible to use any computer dietary quality is based on a comparison compatible with PC DOS or MS DOS in between the individual's nutritional 3.3 version or later. This program requirements and the actual nutrient improves the information processing intake. PROARCAN also adjusts the function that CHILEFOODS nutritional requirements for age, sex encourages. Additionally, it is able to etc., and also computes the nutritional develop input modifications, elimination contribution of a specific diet, or a operations, interactive consultations, particular food. nutrients for which information is Once established both the compiled, information sources, CHILEFOODS and PROARCAN analytical methods. Since the system programs were tested to validate their has the facility to generate reports, it performance. permits the distribution of information, The CHILEFOODS and PROARCAN another objective of LATINFOODS and programs (14, 15) now constitute an CHILEFOODS. important part of the Chemical The other project was to design a Information Center (CIQ) at the Faculty computing system to determine energy of Chemical and Pharmaceutical and nutritional requirements, and to Sciences at the University of Chile. This assess nutrient contributions and quality offers a fast and up-to-date information of the diet of an individual. service to users in the food field and We started with a comparative other areas of chemistry. analysis of the different methods for Other Latin American countries have dietary intake estimation, with the carried out different actions in the field purpose of selecting the most adequate of food composition, many publishing method for a computing system that their own national tables. When their could evaluate the diet quality of an tables are computerized, it should be individual; the dietary recall and the possible in the future to establish an dietary record were the methods international data network around Latin selected for the system. America connected by the For the estimation of dietary intake, LATINFOODS organization. the required energy and nutritional Ŷ References composition of the foods was supplied by the computerized Chilean food (1) Schmidt-Hebbel, H., Pennacchiotti, composition table (2), together with the I., Masson, L., et al. (1961– 85) Latin American and international dietary Table of Composition of Chilean recommendations (12, 13). Foods, 1st Ed. - 7th Ed., University The activities to design the system of Chile, Santiago began with the identification of the (2) Schmidt-Hebbel, H., Pennacchiotti, required outputs, then the input I., Masson, L., et al. (1990) Table of requirements were studied, and finally Composition of Chilean Foods, 8th the specifications for files and Ed., Cramer SACI, Santiago procedures were identified. (3) National Research Council (1989) The program was created in dBASE Recommended Dietary Allowances, III Plus, a package for handling 10th Ed., National Academy of (11) Straley, S.J. (1988) Programming in Sciences, Washington DC Clipper: the Definitive Guide to the (4) Pak, N., Ayala, C., Araya, H., Clipper dBASE Compiler, 2nd Ed., Pennacchiotti, M., & Vera, G. Addison-Wesley Publishing Co. (1989) Arch. Lat. Nutr. 60, 116–125 Inc., Reading MA (5) US Department of Agriculture (12) Latin American Society of Nutrition (1976-) Composition of Foods: (1988) Arch. Lat. Nutr. 38, 383 Raw, Processed Prepared, Agric. (13) FAO/WHO/UNU (1985) Energy and Handbook No.8 series, USDA, Protein Needs, Report of an Expert Washington, DC and Consultative Meeting, (6) Masson, L., Mella, M.A., & Cagalj, Technical Report No. 724, WHO, A. (1990) Rev. Chil. Nutr. 18, 257– Geneva 265 (14) Masson, L., Elías, P., & Chavez, H. (7) Masson, L., & Mella, M.A. (1985). (1990) Computing System for Fats and Oils Habitually and Chemical Food Composition Data Potentially Consumed in Chile, Management, Department of Food Fatty Acid Composition, University Science and Chemical Technology, of Chile, Santiago University of Chile, Santiago (8) Masson, L., Araya, H., & Mella, (15) Masson, L., Rousseau, I., Elías, P., M.A. (1987) Arch. Lat. Nutr. 37, & Chavez, H. (1992) Computing 683–690 System to Determine Caloric and (9) Bressani, R. (1987a) Arch. Lat. Nutritional Requirements, Dietary Nutr. 37, 591–602 Contribution and Diet Quality of an (10) Bressani, R. (1987b) Arch. Lat. Individual, Department of Food Nutr. 37, 793–802 Science and Chemical Technology, University of Chile, Santiago National and International Food Composition Programs

Nutrient Composition of Wild-Gathered Foods from Mali

Marit Beseth Nordeide, Arne Oshaug, Halvor Holm

Nordic School of Nutrition, University of Oslo, PO Box 1046, 0316 Oslo, Norway

he aim of this collaborative project between Mali and Norway is to initiate food analysis which will lead to a food composition table in Mali. The food composition Ttable presently used is the FAO food composition table for Africa which does not have data from Mali. The first foods selected for analysis were important staples used by nomads. Laboratories in Norway and in Sweden have been involved in both chemical and biological analysis. The analytical results are used to discuss improvement of food quality/utilization by combined use of locally-produced and gathered foods in Gourma.

Ŷ Methods washed and boiled with water for 6 Foods. Wild-gathered Cenchrus biflorus hours, with water replaced every hour. (grains), Panicum laetum (grains) and Chemical Analysis. Dry matter, crude Maerua crassifolia (leaves) were protein (N 6.25), ash, gross energy, collected in the dry season, May 1992. lipids, 12 minerals and amino acid The foods were studied as raw material patterns were determined. and as processed. The grains of Biological Analysis. The protein Cenchrus biflorus and Panicum laetum digestibility, biological value and net were boiled in water for 40 minutes. protein utilization of Cenchrus biflorus, Leaves from Maerua crassifolia were Panicum laetum and Maerua crassifolia were determined in N-balance experiments in young growing rats. Table I. Composition of wild-gathered foods from Gourma (per 100 g edible portion) Food Dry matter Protein Fat Energy Ash K Ca Fe Zn Name g g g kJ g mg mg mg mg Cenchrus biflorus 96.9 22.1 7.3 1880 6.4 382 43 234 6.5 whole grains (n=7) Panicum laetum 96.7 9.5 4.8 1580 11.5 340 51 211 3.8 whole grains (n=4) Panicum laetum 96.7 8.2 8.8 1430 23.3 603 78 310 4.7 bran (n=2) Panicum laetum 98.1 12.4 2.2 1630 1.4 178 13 24 3.0 dehusked (n=1) Maerua crassifolia 97.5 26.0 4.1 1500 13.2 2262 1978 130 1.5 dried leaves (n=4)

Ŷ Results Leaves of Maerua crassifolia had Table I shows the composition of high protein quantity, but low availability wildgathered foods from Gourma; and of protein in raw unprocessed material. Table II shows the protein quality. The Processing the green leaves of Maerua data are provisional and require crassifolia yielded an acceptable food, validation. which can be an important source of protein and energy in the dry seasons. Ŷ Conclusion The use of unprocessed grains and Food quality can be substantially leaves as foods is recommended. improved by processing and combined Cenchrus biflorus is used both as raw use of locally gathered wild foods. and as boiled grains. Using raw foods Wild-gathered grains of Cenchrus causes waste and reduced quality. biflorus and Panicum laetum were found Processing and combination are to have high energy content and the necessary to reach a high nutritive value minerals potassium, iron and zinc. of food resources in the area. These grains had a relatively high Ŷ Reference protein concentration, but low protein quality. The protein quality was (1) FAO/WHO/UNU (1985) Energy and increased by adding lysine to the grains Protein Requirements, WHO Tech. Rep. during preparation (initial CS 28 Ser. No. 724, WHO, Geneva. increased to 100). Table II. Protein qualitya of wild-gathered foods from Gourma (per 100 g edible portion) Food Protein Chemical Limiting Biological True Net scoreb amino value digestibility protein acid utilization Name g % % % % Cenchrus biflorus 22 28 Lys 41 89 36 whole grains Cenchrus biflorus 22 28 Lys 41 87 34 whole grains, boiled Cenchrus biflorus 22 103 Thr 65 80 52 whole grains boiled + lysine Panicum laetum 9.5 34 Lys 44 81 35 whole grains Panicum laetum 8.2 62 Lys 39 56 22 bran Panicum laetum 12.4 23 Lys 34 95 32 dehusked grains Panicum laetum 12.4 23 Lys 43 82 35 dehusked grains, boiled Maerua crassifolia 26 116 Lys - ca 60 - dried leaves Maerua crassifolia 28 102 Lys 65 70 46 dried, boiled a N-balance experiments in rats; b reference pattern for preschool children 2 to 5 years (FAO/WHO/UNU 1985) Section II

Methods and Conventions of Nutrient Analysis

Supported by AOAC INTERNATIONAL

his session was chaired by Dr Doreen Clark, Managing Director of Analchem Bioassay, Sydney. The keynote address AOAC INTERNATIONAL-Validated TMethods for Nutrient Analysis — Method Availability and Method Needs was given by J. DeVries. This was followed by papers on Analysis and Classification of Digestible and Undigestible Carbohydrates by N-G. Asp, Recent Developments in the Determination of Water-soluble Vitamins in Food—Impact on the Use of Food Composition Tables for the Calculation of Vitamin Intakes by P.M. Finglas, Update on the Analysis of Total Lipids, Fatty Acids and Sterols in Foods by A.J. Sinclair and Conventions for the Expression of Analytical Data by D.A.T. Southgate. All of these papers are published on the following pages.

G.R. Beecher, F. Khachik and J.T. Vanderslice did not provide a print version of their presentation Recent Advances in the Analysis of Fat-Soluble Vitamins in Foods. They can be contacted at the Nutrient Composition Laboratory, Beltsville Human Nutrition Research Center, ARS/USDA, Beltsville MD 20705, USA for further information. Methods and Conventions of Nutrient Analysis

AOAC INTERNATIONAL- Validated Methods for Nutrient Analysis — Method Availability and Method Needs

Jonathan W. DeVries

Medallion Laboratories, General Mills Inc., 9000 Plymouth Ave No., Minneapolis, MN 55427, USA

Adequate analytical methods for nutrients in foods, food ingredients, and food products are the basic first step in determining the nutritional adequacy of a food supply. Whatever the ultimate use of nutrition data, i.e. consumer education via the food label, or databases for nutrient and deficiency disease studies, the assay used to provide the data must determine the analyte of interest adequately. AOAC INTERNATIONAL (formerly the Association of Official Agricultural Chemists then the Association of Official Analytical Chemists) has been systematically validating methods for nutrition analysis for over 100 years. This validation includes a complete peer review system, with study of the method in multiple laboratories and multilevel review of the study results to assure adequacy of a proposed method for its intended purpose. With the passage in the USA of the Nutrition Labeling and Education Act, concern arose regarding the availability and adequacy of validated analytical methods to meet the requirements of the labeling act. A special AOAC task force with members drawn from regulatory agencies, the food industry, academia, and analytical suppliers was formed to address the concerns. In this paper, results of the task force assessment of adequacy of current Official Methods for nutrition analysis are presented. Method matrix combinations where updated methods or method modifications are needed are covered. In addition, a number of special issues addressed by the task force relating to the analysis of fat, moisture, and carbohydrate, reference materials (certified and in-house), and the methods validation process are discussed. dequate analytical methods for nutrients in foods, food ingredients, and food products are the basic first step in determining the nutritional adequacy of a food Asupply. Whether the nutrition data are ultimately used to inform consumers with information on the food label, or to build databases to study correlations between nutrient(s) and deficiency diseases, the assay used to provide the data must determine the analyte of interest adequately. AOAC INTERNATIONAL (formerly the Association of Official Agricultural Chemists and then the Association of Official Analytical Chemists) has been systematically validating methods for nutrition analysis for over 100 years. These validated methods provide competent laboratories with a means of supplying dependable data for nutrition labels of databases regarding the nutrition content of foods and food products. This paper covers three areas, first, In 1965, the name of the association the history of AOAC INTERNATIONAL, was changed to Association of Official and its processes and criteria for Analytical Chemists. As the validation and acceptance of a method Association's activities grew to include as an AOAC Official Method; second, microbiologists and other scientists, the recent activities of AOAC carried out in membership base became international response to the recently proposed in scope. As a result, the name was Nutrition Labeling and Education Act updated to AOAC INTERNATIONAL in (NLEA)(1) in the US; and third, ideas for 1991. Membership now includes a methods validation scheme that might scientists from many fields worldwide be used to improve the method interested in improving analytical validation process for foods, providing methodology and results. better comparative data and more Method validation under AOAC rugged methods for laboratory use. INTERNATIONAL auspices includes a complete peer review system, with study Ŷ History and Procedures of AOAC INTERNATIONAL of the method in multiple laboratories and multilevel review of the study results In 1884, a group of regulatory to assure adequacy of a proposed agricultural chemists formed the method for its intended purpose. The Association of Official Agricultural key to rugged effective validated Chemists to adopt uniform methods for methods in AOAC INTERNATIONAL the analysis of fertilizers. The lies with the Associate Referee. The collaborative study was adopted as a Associate Referee is appointed on a means of validating methods and recommendation of a methods evaluating their performance. One of the committee or a General Referee on the active participants during the early years basis of the Associate Referee's of the Association, Dr. Harvey W. Wiley expertise in an analytical area, i.e. carried out extensive studies on the active in methods development work, adulteration of foods and drugs using actively carrying out work assignments AOAC methods. This ultimately led to or projects relating to the analyte of the passage of the US Federal Pure interest, etc. Quite frequently the Food and Drug Act of 1906. By 1912, Associate Referee develops an the Association had begun publishing its analytical method to meet a need, or validated methods as Official Methods in through knowledge of the literature USDA bulletins. By 1920, the volume of selects an applicable method for study. validated methods had grown to the After a requisite number of laboratories point where it warranted its own volume, have been found to carry out a and the Official Methods of Analysis was collaborative study, the Associate established. It has been revised and Referee distributes the methodology updated every five years since then (2). and samples. After the collaborative or issues can be considered for “Final study is complete, the Associate Action Status”, a status achieved Referee collects the data, develops a through ballot of the entire AOAC study report and submits a INTERNATIONAL membership. There is recommendation for method adoption to no difference in the Official Status of the Association. Assisting the Associate Methods, whether “First Action” or “Final Referee is the General Referee, Action”. “Final Action” only indicates that appointed on the basis of expertise and a method has withstood some test of experience in broad analytical areas, time with no substantive questions who brings this broad knowledge base raised regarding its performance. Any to bear on the study and its results. method achieving Official Status through When the General Referee and the the AOAC process has had both Associate Referee agree that a method substantial performance testing in performs sufficiently well (see below for multiple laboratories and peer review by a discussion of criteria) to be considered scientists who are experts in the as an Official Method, the method is analytical area. In addition, it has had submitted to an AOAC statistician and a intense scrutiny by scientists in related safety advisor for review. Upon endeavors. completion of these reviews, the method Criteria for validation of a method for is sent to an appropriate Methods Official Status are well established (3). Committee for review and The method must be submitted to recommendation regarding Official participating laboratories written exactly Status. Methods committees are as it is intended to be run. Participating constituted of members chosen for their laboratories are expected to run the broad expertise in a given analytical method exactly as written. For a given area such as Food Nutrition, Food collaborative study, participation by no Toxins, or Drug Residues. fewer than eight laboratories analyzing a Recommendation to the Official minimum of five sample materials is Methods Board to adopt a method as required for quantitative methods. For Official First Action requires agreement qualitative methods, no fewer than 15 of two-thirds of the members of the laboratories analyzing a minimum of two Methods Committee. If members of the analyte levels per matrix, five samples Methods Committee raise significant per level, and five negative controls are questions with regard to the method or required. Obviously in both cases its performance, the method cannot be participation by more laboratories and recommended for Official Status until the inclusion of more samples is those questions have been addressed encouraged. In extenuating by the Associate Referee. Upon circumstances, for example a particular recommendation from the Methods method being considered has significant Committee, the method is considered regulatory or commercial importance, for Official Status by the Official but can only be carried out in five Methods Board. The Board reviews the laboratories anywhere because only actions taken on the method, the review they have key instrumentation special process, and assures consistency consideration is given. Obviously, such between methods and between circumstances are rare. methods committee reviews. If the After the collaborative study is method is given “First Action Official complete, statistical outliers Status”, it is published in the Official (laboratories and/or data points) are Methods of Analysis. After “First Action” removed (3). Rejection of data from status for two years, methods which more than two-ninths of the laboratories have no unresolved negative comments (without a valid explanation such as failure to follow the method) is basis for The NLEA will have a significant the rejection of the method as being impact on industry, consumers, and insufficiently rugged to be adequate for government agencies. It is estimated intended purpose. Method performance that it will cost industry upwards of $1.5 (in statistical terms) will vary depending billion for the relabeling required, an on analyte, matrix, and/or analyte estimated $1500 per product for small quantity. Ultimately, therefore, a method firms and $900 per product for large must be accepted based on its firms. Analytical costs will probably performance in collaborative study as range from $750 for the 40 per cent of judged by scientific peer review by US foods that need label changes to experts from government, academia, $1800 for the 60 per cent of foods that industry and other organizations. These had not been previously labeled. experts are cognizant of the ultimate Research and development costs for use of the methods being validated and products that will be modified somewhat judge adequacy for intended purpose. for marketing advantage under the Working together in concert through provisions of the act are hard to AOAC INTERNATIONAL, these experts estimate, but run anywhere from have produced high quality methods for $20,000 to $400,000 per product. the analytical community to use. Typically two to five months will be needed to redesign and print new Ŷ Methods Needs for Nutrition Labeling packages. For consumers the cost of relabeling will be passed along in higher With the recent passage in the USA of product prices. No money has been the Nutrition Labeling and Education allocated for “Education”, so it is Act, concern arose amongst food expected that significant consumer consumers, producers, regulators, and confusion will exist after the label laboratories providing nutrition analytical changes occur. Governmental agencies services, regarding the availability and will incur extra costs for interpretation, adequacy of validated analytical analysis, and enforcement of the act. methods to meet the requirements of The effective date for NLEA was May the labeling act. The act was passed by 8, 1994, however other aspects of the US Congress in November of 1990. labeling had different effective dates, i.e. The act required the US Food and Drug juice labeling in May, 1993, health Administration to promulgate proposed claims in May 1993, and metric weight regulations for nutrition labeling of declarations in February, 1994. The nearly all foods sold in the US. The US NLEA now mandates nutrition labeling Department of Agriculture, although not of most products and allows specified legally required to do so, initiated uses of nutrient descriptors and health activities to adopt labeling regulations claims related to nutrition. essentially equivalent to those of the Label format(s) is(are) rigidly USFDA. The proposed regulations of specified under the NLEA (e.g. Figure November 1991, were open for public 1). Mandatory declarations include comment for a number of months, with calories, calories from fat, total fat, final regulations due in November of saturated fat, cholesterol, sodium, total 1992. The final regulations were actually carbohydrate, dietary fiber, sugars, issued in January 1993, with an protein, vitamin A, vitamin C, calcium, effective date for new label and iron. Voluntary declaration is implementation of May 8, 1994 (July 8, allowed for calories from saturated fat, 1994 for products under USDA polyunsaturated fat, monounsaturated jurisdiction). fat, stearic acid (USDA products), potassium, soluble fiber, insoluble fiber, sugar alcohols, other carbohydrates, (typically for micronutrients) or against a thiamin (vitamin B1), riboflavin (vitamin Daily Reference Value (typically for B2), niacin, vitamin D, vitamin E, folate, macronutrients). For example, the daily cyanocobalamin (vitamin B12), reference value (based on 2000 phosphorus, iodine, magnesium, zinc, calories/day) for fat is 65 g, for saturated copper, biotin, and pantothenic acid. fat is 20 g, for cholesterol is 300 mg, and for dietary fiber is 25 g. To encourage consistency in reporting of Are You Ready daily values, reference amounts relating for New to serving sizes have been published for common food items, e.g. 30 g for ready Food Labels? to eat cereals and cookies, 55 g for Nutrition Facts cake. Label serving sizes are to be in Serving Size 1 cup (228g) common household units, e.g. cups, Servings Per Container 2 teaspoons etc. Amount Per Serving Nutrient claims can be made Calories 90 Calories from Fat 30 regarding the food product. However, if fat, saturated fat, cholesterol, or sodium % Daily Value* exceed certain levels, this must be Total Fat 3g 5% disclosed on the package label along Saturated Fat 0g 0% with the nutrient claim. Adequate Cholesterol 0mg 0% Sodium 300mg 13% analytical methods are obviously Total Carbohydrate 13g 4% needed to assure compliance both with Dietary Fiber 3g 12% the spirit of the nutrient claim, as well as Sugars 3g to monitor the disclosure level Protein 3g compliance. For added nutrients (referred to as Vitamin A 80% • Vitamin C 60% Class I nutrients), the nutrient must be Calcium 4% • Iron 4% present at 100 per cent or greater than * Percent Daily Values are based on a 2,000 calorie diet. Your daily values may be higher declared. For naturally occurring or lower depending on your calorie needs: nutrients, (Class II), the nutrient must be Calories: 2,000 2,500 present at a level at least 80 per cent or Total Fat Less than 65g 80g greater than declared, but less than or Sat Fat Less than 20g 25g equal to 120 per cent of declared. Cholesterol Less than 300mg 300mg Examples of nutrients that must be Sodium Less than 2,400mg 2,400mg greater than 80 per cent of declared are Total Carbohydrate 300g 375g dietary fiber and potassium. Examples Dietary Fiber 25g 30g of nutrients that must be less than 120 Calories per gram: per cent of declared are fat, saturated Fat 9 • Carbohydrate 4 • Protein 4 fat, and sugar. Analytical variability is NATIONAL FOOD PROCESSORS ASSOCIATION taken into account for enforcement, so in cooperation with FDA and FSIS well-characterized validated methods Figure 1. Example of food label are necessary for compliance conforming with NLEA requirement. monitoring. Labels will list the quantity of a given Ŷ AOAC Response to Nutrition nutrient, along with a percentage of a Labeling Needs daily recommended dietary intake value guideline for the consumer to use for To deal with concerns regarding comparison. The percentage of daily availability of adequate methods to meet value is determined against either a the needs of NLEA, a special task force Reference Daily Intake (RDI) value of the AOAC with members drawn from regulatory agencies, the food industry, assignments to review AOAC methods academia, and analytical suppliers was on an analyte-matrix basis. After this formed. The objectives of the task force preliminary review was done, the entire were to: determine which Official task force, along with aid solicited from Methods are adequate to meet current others, reviewed the assessments of the nutrition labeling analysis requirements; individual members. The analyte-matrix determine which Official Methods need grid of adequate methods began to fill revisions or modifications to meet in. As the task force progressed, the current nutrition labeling analysis information being generated was requirements; determine which nutrient- regularly reported in The Referee to matrix combinations require the keep the AOAC membership informed development and validation of Official of progress and to allow feedback. For Methods; propose means by which example the assessment of adequate AOAC INTERNATIONAL can supply methods under the proposed regulations needed methods and/or modifications; was published in the July 1992 issue of identify means by which reference The Referee (4). materials might be incorporated into Initial review of adequate methods AOAC Official Methods and into the under the proposed regulations, validation process for AOAC Official indicated that 947 of the 1080 possible Methods, further assuring the quality matrix-analyte combinations had and performance of those methods. adequate methods. This meant that 88 The task force began informally at per cent of the methods needs were the AOAC INTERNATIONAL Annual addressed. In some cases, the Official Meeting in 1991, and was formally Methods were deemed adequate for the appointed by the board of directors in need, but newer technologies can be December of that year. Efforts were brought to bear on the analyte-matrix initiated immediately to obtain feedback combination to provide better methods regarding the status of Official Methods at this point in time. An example might used for nutrition labeling. A survey was be vitamin A. The Carr-Price (5) method conducted of laboratories carrying out provides adequate results for labeling nutrition analysis and using AOAC purposes, however most laboratories methods. An information gathering today would rather use liquid session was also held in March of 1992. chromatography (e.g. (6)) and avoid A number of task force meetings were handling the corrosive antimony held in the succeeding months to carry trichloride. Therefore, although the task out the assigned objectives and fulfill the force accepted the adequacy of the task force's mission. Carr-Price method, it is recommending Under the proposed nutrition labeling that validation of liquid chromatography regulations, up to 54 nutrition-related methods be undertaken. items were either required or could be As the list of adequate methods was placed on the label, everything from A being generated, a complementary list (ash) to Z (zinc). To organize the task of of methods in need of validation or evaluating methods for these analytes, revision was also developed. This was the task force divided foods into 20 published in October 1992 (7) to alert different matrix groups that were felt at members of methods needs. the time to cover the scope of foods and Special Nutrition Labeling Issues food products. This resulted in 1080 analyte-matrix combinations to be As the task force evaluated methods for assessed regarding availability of nutrition analysis, a number of issues adequate methods. Individual committee were raised, in particular, methods for members took upon themselves fat, dietary fiber, moisture, carbohydrates, standards and reference appropriate methodology when a materials for Official Methods, and the definition is adopted. need for a clear-cut means of A complete listing of moisture determining if a particular Official methods, along with their characteristics Method is applicable to all foods. has been published by the moisture Subcommittees of the task force were subcommittee (9). As with complex formed to address each of these issues. carbohydrates, a clearer definition of Fat has traditionally been analyzed moisture will be helpful in validating by a variety of methods depending upon more concise methodology for this matrix, analyst carrying out the analysis, analyte. and intended use of the resulting data. The subcommittee on reference Typically, the result was dependent materials published a listing of upon determination of some solvent- commercially available reference soluble (solvents varied depending on materials for the nutrients requiring the method) fraction of the food being mandatory labeling in August, 1992 analyzed. The task force realized that a (10). The subcommittee further went on single concise definition for fat was to publish Guidelines for the Preparation needed. AOAC INTERNATIONAL does of Inhouse Quality Assurance Materials not set definitions for nutrients, but in the May, 1993 issue of The Referee provides validated analytical methods to (11). Recognizing that assuring an quantify defined nutrients. Therefore, adequate supply of reference materials the subcommittee recommended, and was an ongoing task, requiring the task force concurred, that the significant follow-up long after the regulatory agencies, the USDA and nutrition labeling task force would be FDA, adopt a single concise definition disbanded, the task force supported the for fat. The agencies responded by formation of the first technical division of adopting a definition of fat as the sum of AOAC INTERNATIONAL, namely the the fatty acids (regardless of source) in Technical Division on Reference the food, expressed as triglycerides (8). Materials. This division will continue the This concise definition provides a “gold efforts initiated through the task force standard” for evaluating fat analysis and will expand to reference materials methods in the future. beyond food nutrition. This division The carbohydrates subcommittee already has over 125 members and held determined that methods for total, its first annual meeting in conjunction soluble, and insoluble dietary fiber are with the AOAC INTERNATIONAL adequate. Sugar methods, in particular Annual Meeting in July, 1993. the liquid chromatography methods with Ŷ Method Validation Needs defatting steps, while adequate, should be further studied to assure validity After the final regulations for Nutrition across a broader matrix base. Complex Labeling in the US were issued by the carbohydrates as a nutrition label item USDA (12) and the USFDA (13), the had been included in the labeling task force reassessed methods proposal, but eliminated from the final adequacy and needs. The updated regulations due to the lack of a clear listings were published in the March (14) definition of the nutrient, and lack of and April (15), 1993 issues of The analytical methods to measure it. The Referee, respectively. In particular, subcommittee (and the task force) methods and/or collaborative studies recommends a concise definition for are needed for ȕ-carotene, biotin, sugar complex carbohydrates be adopted and alcohols, sugars (verification for certain has committed AOAC to validating matrices), cholesterol, copper, cyanocobalamin, defatting of samples for dietary fiber, fat (total, saturated, Therefore, the remaining three monounsaturated, and stearic acid), macronutrients, fat, protein, and folate, iodine, niacin (microbiological carbohydrate have the major impact on method), pantothenate, protein the effectiveness of an analytical (eliminating mercury use), pyridoxine, method. In a picture of a triangle with tryptophan (microbiological method), fat, protein, and moisture at the apices, vitamin A, vitamin C (where erythorbate all food samples will fit somewhere on is present), and vitamin E. Some of that triangle, assuming the sum of fat, these nutrients do have adequate protein, and carbohydrate is normalized methods, however, the methods are in to 100 per cent, and these components need of modernization and therefore are are expressed as a percentage thereof. recommended for further study. For example, a sample with 10 per cent fat, 30 per cent carbohydrate, and 10 Ŷ The Food Triangle as a Systematic Approach to Method Validation per cent protein will have normalized values of 20 per cent fat, 60 per cent A question that arose during the task carbohydrate, and 20 per cent protein. force deliberations was: How does one The triangle can be split equally in ascertain with reasonable confidence nine subtriangles, with any particular that a method is applicable to all foods nutrient lying between 0–33 per cent, without a substantial history of trouble- 33–67 per cent, and 67–100 per cent, free application to a wide variety of food respectively. By choosing 18 samples samples? Clearly, a defined systematic (two from each subtriangle), the analyst approach might be helpful to assure would be reasonably certain of covering method ruggedness across all food foods characteristic of most foods. To types while minimizing the analyst's develop further confidence in a method, efforts in assessing the method. The samples taken from a subtriangle can task force Subcommittee on Definition of be purposefully chosen to represent Foods for Analytical Purposes has particular characteristics, i.e. for the 67– proposed an approach that is currently 100 per cent carbohydrate subsection, a being considered by the Foods high fiber and a high starch sample committees and the Official Methods might be used. For the 67–100 per cent Board (16). fat section, a milk or animal fat and a The idea of requiring a collaborative vegetable fat might be chosen. The study of 40 or more samples can be system could be applied to any nutrient very discouraging, both for the associate being analyzed by using a Youden referee organizing the study and for pairing technique [closely matched potential participants. There are five sample pairs as opposed to blind macronutrient components of any given duplicates] (3) for determination of food that have a significant impact on within laboratory variability for the the performance of a method, no matter analyte of interest. If difficulty is what the analyte being measured. The experienced with getting acceptable macronutrients impact analysis of results for the method in question for various analytes by causing extraction samples from certain subtriangles, this difficulties or analyte interferences. The information could be quite helpful for five macronutrients are moisture, ash, understanding and delineating the protein, fat, and carbohydrate. Moisture cause of the ineffectiveness. A similar of nearly all samples can be adjusted if approach had earlier been suggested the level affects an assay. Water can be for reference materials for nutrition added, or the sample dried. Ash content analysis (17). The concept is extended of a sample usually has little effect on by Tanner et al. (18). assays, particularly for organic nutrients. (Eds.), John Wiley and Sons, New Ŷ Conclusion York, NY The task force has completed its (7) Nutrient Labeling Task Force (1992) objectives and reported the results of its The Referee 16, 5–10 deliberations on an ongoing basis in The (8) Anon. (1993) Federal Register 58, Referee, the official house organ of 2086–2093 AOAC. The final report has been (9) Anon. (1993) The Referee 17, 6–9 published (19). The task force (10) Anon. (1992) The Referee 16, 4–5 disbanded at the July 1993 Annual (11) Anon. (1993) The Referee 17, 6–8 Meeting of AOAC INTERNATIONAL. (12) Anon. (1993) Federal Register 58, Ŷ References 631–2063 (13) Anon. (1993) Federal Register 58, (1) Anon. (Nov 27, 1991) Federal 2065–2964 Register 56, no 229, 60301–60891 (14) Nutrient Labeling Task Force (1993) (2) Official Methods of Analysis, (1920, The Referee 17, 6–10 1925, 1930, 1935, 1940, 1945, (15) Nutrient Labeling Task Force (1993) 1950, 1955, 1960, 1965, 1970, The Referee 17, 6–8 1975, 1980, 1985, 1990) (16) Anon. (1993) The Referee 17, 1, 6– Association of Official Analytical 7 Chemists, Arlington, VA (17) Southgate, D.A.T. (1987) Fres. J. (3) Manual for the Development, Study, Anal. Chem. 326, 660–664 Review, and Approval Process for (18) Tanner, J.T., Wolf, W.R., & Horwitz, AOAC Official Methods (1993) W. (1995) in Quality and AOAC INTERNATIONAL, Arlington, Accessibility Food-Related Data, H. VA Greenfield (Ed.), AOAC (4) Nutrient Labeling Task Force (1992) INTERNATIONAL, Arlington, VA, The Referee 16, 1, 7–12 pp. 99–104 (5) Official Methods of Analysis (1990) (19) Methods of Analysis for Nutrition 15th Ed., secs 1045–1047, AOAC, Labeling (1993), D.M. Sullivan, & Arlington, VA D.E. Carpenter, (Eds.), AOAC (6) DeVries, J.W. (1985) in Methods of INTERNATIONAL, Arlington, VA, Vitamin Assay, J. Augustin, B.P. pp. 33–83 Klein, D. Becker, & P.B. Venugopal, Methods and Conventions of Nutrient Analysis

Analysis and Classification of Digestible and Undigestible Carbohydrates

Nils-Georg Asp

Department of Applied Nutrition and Food Chemistry, Lund University, Chemical Center, PO Box 124, S-221 00 Lund, Sweden

The current interest in the nutritional properties of various food carbohydrates has increased the demand for compositional data. The small-intestinal digestibility is the most important nutritional property. The digestible carbohydrates provide glucose to body tissues, whereas the undigestible carbohydrates are partially fermented and provide fermentation substrate and bulk in the colon. Mono-, di- and oligosaccharides, as well as polyols, can be determined with specific enzymatic methods, but gas-liquid chromatography (GLC) and especially liquid chromatography (LC) methods are preferable when a range of sugars is to be analyzed. Dietary fiber determination should aim to differentiate between digestible (“available”) and undigestible (“unavailable”) carbohydrates. Gravimetric and component analysis methods are complementary for different purposes. Resistant starch, i.e. undigestible starch, as well as lignin should be included in the dietary fiber. Starch is preferably analyzed with specific enzymatic methods, that should have the same cut-off as for starch removal in dietary fiber analysis.

ietary guidelines in Western countries recommend that the carbohydrate intake be increased to at least 55–60 per cent of the energy (1). In diets consumed in other D parts of the world carbohydrates may contribute more than 70 per cent of energy. Originally, the carbohydrate recommendations came as a consequence of the fat and protein recommendations. In recent years, however, the nutritional importance of the carbohydrates as such has been more and more emphasized, and new developments call for a more nutritional classification of the different food carbohydrates as a basis for more specific recommendations about intake. Labeling of foods regarding Starch generally occurs in the largest carbohydrate content is a separate, but amount in diets, followed by sucrose closely related issue. Most carbohydrate and — when milk products are content figures on food labels are still consumed — lactose. Glucose, fructose being calculated “by difference”, i.e. the and sucrose are present naturally in material remaining after moisture, ash, fruits, berries and vegetables, and also fat and protein determinations. In view may be added as refined sugars. of the quite variable nutritional effects of Polyols and fructans such as inulin are different carbohydrates, this is increasingly used as lowcalorie bulking unsatisfactory. agents, as is polydextrose. Since many A thorough characterization of the dietary guidelines recommend a limited various digestible and undigestible use of refined sugars (sucrose, fructose, carbohydrate fractions is required corn syrups, high fructose corn syrup whenever investigating the physiological etc.), the contribution of such “extrinsic” properties of a carbohydrate-containing sugars is of special interest. However, it food or diet. Compositional data on the is not possible analytically to distinguish food carbohydrates are also essential in between “extrinsic” sugars and the epidemiological research. “intrinsic” sugars present naturally. Table I shows the food carbohydrates that are quantitatively most important.

Table I. Main food carbohydrates Monosaccharides Polysaccharides Glucose Starch Fructose - amylopectin Galactose - amylose Polyols - modified food starches

Disaccharides Non-starch polysaccharides (NSP) Sucrose - cellulose Lactose - hemicelluloses Polyols - pectins - fructans Oligosaccharides - gums Į-Galactactosides - mucilages - raffinose, stachyose - algal polysaccharides other Fructans - fructo-oligosaccharides Polyols Polydextrose the other hand, are delivered to the Ŷ Nutritional Properties of Food Carbohydrates large intestine and fermented to various extents. The main products of this Small Intestinal Digestibility anaerobic fermentation are acetate, propionate and butyrate. Acetate and Carbohydrates that are digested and propionate are absorbed and absorbed in the small intestine provide metabolized in peripheral tissues and glucose, fructose and galactose to body the liver, respectively, and their possible tissues. Undigestible carbohydrates, on effects on carbohydrate and lipid metabolism are currently investigated. lowering blood cholesterol and blood Butyrate is an important source of glucose after a meal (for review, see energy for the epithelial cells of the large e.g. 10). intestine itself, and may be important in In the large intestine, the dietary fiber protecting against colonic cancer (for polysaccharides, oligosaccharides and review, see e.g. 2). Various fermentable resistant starches are fermented to carbohydrates give different proportions various extent with production of of these fermentation products (3). acetate, propionate and butyrate in As early as 1929, McCance and various proportions. Lawrence emphasized small-intestinal Rate of Carbohydrate Digestion and digestibility by introducing the term Absorption “available” carbohydrates, based on determination of starch and digestible In diabetes, patients have long been sugars (4). Correspondingly the term advised to choose carbohydrates that “unavailable carbohydrates” was used are slowly digested and absorbed, for cellulose, non-cellulose giving a limited and sustained blood polysaccharides and lignin (5). Within glucose elevation with minimum insulin the European Community legislators requirement. This has been have defined “carbohydrates” as demonstrated to improve the metabolic digestible (“metabolizable”) control in maturity onset diabetes carbohydrates and including polyols (6). (NIDDM) (11). Generally it has been Dietary fiber was first defined as the believed that starch is slowly digested remnants of plant cell-walls not digested and absorbed due to its high molecular in the small intestine (7). With this weight (“complex carbohydrate”). definition it constitutes the non-starch Sucrose and other low-molecular weight polysaccharides of the plant cell-walls, carbohydrates (“simple sugars”), on the but also undigestible protein, inorganic other hand, have been regarded as material, tannins, cutin etc. The rapidly absorbed. It is remarkable that redefinition by Trowell et al. (8) this view has been so prevalent in spite restricted the definition to of the lack of scientific evidence. On the polysaccharides and lignin, but enlarged contrary, data accumulated in the 70s it to include all undigestible and 80s showing that the height and polysaccharides. There were two shape of the blood glucose curve could reasons for this: First, purified be quite different after the intake of polysaccharides such as pectins and different foods, and that these gums were frequently used to study the differences were unrelated to the physiological effects of dietary fiber molecular size of the carbohydrates. constituents, and second, cellwall Among the low molecular weight polysaccharides could not easily be carbohydrates, fructose gives a very low differentiated analytically from glycemic response, and sucrose is undigestible polysaccharides from other intermediate between glucose and sources (9). fructose (12). Starchy foods are found Dietary fiber includes a large number across the whole range of “slow” to of polysaccharides with quite different “rapid” (for review see e.g. 13). properties, both from the chemical and A number of other carbohydrate and physiological points of view. Insoluble, food properties determining the lignified types of dietary fiber have the glycemic response have now been most prominent fecal bulking effect due identified and include gel-forming types to their resistance to fermentation, of dietary fiber, degree of gelatinization whereas soluble, gel-forming and other properties of the starch, polysaccharides are most efficient in cellular structure and gross structure (14). There is increasing evidence that However, some polysaccharides such “slow” properties of food carbohydrates as inulin and pectic substances may may also be beneficial in relation to have considerable solubility also at this blood lipid levels, satiety, physical alcohol concentration. Some sugars, performance and dental caries (12). The especially lactose, have a slow rate of cariogenic properties of foods have dissolution and limited solubility, and mainly been related to added sucrose, may need lower alcohol concentration but there is evidence that other (e.g. 50 per cent v/v) at extraction with a fermentable carbohydrates are final increase to precipitate important as well. Even starch can lower polysaccharides (17). dental plaque pH, and this property is The International Union of Pure and related to the availability of starch for Applied Chemistry (IUPAC) defines enzymatic degradation in the mouth oligosaccharides as having less than ten (15). monomeric residues. In practice, The term “complex carbohydrates” however, oligosaccharides are defined was used in 1977 by the U.S. Senate as carbohydrates soluble or extractable Committee on Nutrition and Human in aqueous ethanol. Precipitation in 78– Needs (McGovern Report) without any 80 per cent ethanol (or exact definition, but meaning in practice dialysis/ultrafiltration) is generally used digestible polymeric carbohydrate, i.e. to separate oligosaccharides and starch starch. From what has been said above, degradation products from it is obvious that starch has no polysaccharides in dietary fiber analysis nutritional advantage per se, and (9). therefore, the term “complex The Į-galactosides in leguminous carbohydrates” is questionable. In seeds and fructans in onions, artichokes Britain, it was reintroduced to mean etc. are the quantitatively most starch and non-starch polysaccharides important groups of naturally occurring (16). Although the grouping together of undigestible oligosaccharides. Inulin is a starch and non-starch polysaccharides nonstarch polysaccharide, but it is also may be relevant from the chemistry not determined as dietary fiber with any point of view, its usefulness for of the current methods in spite of a nutritional classification is questionable. degree of polymerization (DP) of 30 or Recommendations regarding complex more. Arabans in sugar beet fiber are carbohydrate intake (1) are very another example of a dietary fiber complicated to interpret in terms of polysaccharide that is extremely soluble foods even for experts. in alcohol due to its extensive branching (18). Polydextrose also falls into this Ŷ Overview of Analytical Methods category. Mono-, Di- and Oligosaccharides Physical methods, such as including Polyols polarimetry, refractive index, or density are still useful in pure systems, e.g. in Depending on the food matrix to be sugar production control. Methods analyzed, an extraction of the free based on the reduction of copper salts, sugars may be necessary. Aqueous and colorimetric methods based on ethanol is preferable due to the toxicity condensation reactions with anthrone, of methanol, that was frequently used orcinol and carbazol, can also be used earlier. A final ethanol concentration of in well known systems (17). at least 80 per cent (v/v) should be used to avoid extraction of polysaccharides. Table II. Resistant starch definition and determination 1. Difference in “NSP” glucan without and with KOH or DMSO solubilization (21). 2. Starch remaining in enzymatic, gravimetric dietary fiber residue (24). 3. Starch remaining after extensive Į-amylase hydrolysis (27). 4. Difference of total starch and starch hydrolyzed after standardized milling during pancreatin/amyloglucosidase incubation for 120 min. Separate procedures for three forms of resistant starch available (38).

The enzymatic procedures based on enclosed starch, raw Į-type starch specific, highly purified enzymes have granules, retrograded amylose and been instrumental in providing means of chemically or physically modified food specific and precise analysis of starches (23). carbohydrates in mixtures without high Originally, resistant starch was capital investments. GLC and LC measured as the starch remaining procedures, on the other hand, are associated with the dietary fiber if preferable when a number of different solubilizing agents were not used (21, carbohydrates are to be determined 24). This type of resistant starch has simultaneously. LC analysis has for long been identified as mainly retrograded been hampered by the relative amylose (25, 26). Methods capable of insensitivity of refractory index measuring also other forms of resistant detectors. However, this has been starch (27, 28) are currently evaluated overcome by systems using against in vivo measurements of starch amperometric detection (17). absorption within the European FLAIR Polysaccharides Concerted Action program (EURESTA). Chemically modified food starches Starch. Starch is the predominant (29) and dry heated starches (30) are dietary carbohydrate, and the only degraded by amylases to fragments that polysaccharide that is digestible in the are soluble in alcohol. These fragments human small intestine. Enzymatic are neither determined as starch with hydrolysis and specific glucose assay is enzymatic methods, nor as dietary fiber the method of choice today because of (30). Methods for resistant starch glucose liberation also from for instance analysis are summarized in Table II. Į-glucans at acid hydrolysis. However, Dietary Fiber. Dietary fiber is the enzymes used have to be checked analyzed according to two different for contaminating activities (19). principles (for review, see 9). In the A heat stable amylase (Termamyl) in gravimetric methods the non-fiber a combined gelatinization and hydrolysis components are removed and a residue step, has turned out to be particularly weighed. The residue can be analyzed useful (e.g. 20). for e.g. protein and ash, and corrections Resistant Starch. Resistant starch used accordingly. The crude fiber and was first defined as a starch fraction detergent fiber methods belong to this resisting amylase hydrolysis unless it category. Enzymatic gravimetric was first solubilized in KOH or DMSO methods such as those approved by the (21). It is now generally defined as the AOAC (31, 32) use alcohol precipitation sum of starch and products of starch to recover soluble fiber components and degradation not absorbed in the small can be used to measure total dietary intestine (22). It is then an undigestible fiber (TDF) or soluble and insoluble polysaccharide, that should be included components separately. Correction for in the dietary fiber. There are different protein and ash in the fiber residue is forms of resistant starch: Physically needed. Table III. Advantages and disadvantages of various method for dietary fiber analysis (9). Enzymatic Component analysis gravimetric GLC/LC Colorimetry Equipment Simple Advanced Simple Information on composition No Yes No/Yes Risk of overestimation Yesa No (Yes) Risk of underestimation No Yesb Yesb a but residue can be analyzed b if hydrolysis is complete

The component analysis methods hydrolysis of the polysaccharides. use more or less specific determination Incomplete hydrolysis or losses due to of monomeric constituents, that are then decomposition of monomers will lead to summed to yield a total fiber value. As in underestimation (for review, see 9). gravimetric methods, soluble and The current component analysis insoluble components can be methods employ acid hydrolysis and determined separately. It should be corrections for hydrolysis losses of the noted that the solubility of different components. As in amino acid polysaccharides is method dependent analysis, conditions for the hydrolysis and determined by the temperature, have to be chosen to obtain an optimal time and pH conditions used. compromise between hydrolysis yield The Southgate procedure (33) and monomer degradation. Quantitative employs colorimetric methods to hydrolysis yield is particularly difficult to determine hexoses, pentoses and obtain with acidic polysaccharides due uronic acids. The methods of Theander to the high stability of glycosyl uronic et al. (34) and Englyst et al. (35) use acid linkages towards acid hydrolysis. GLC for neutral sugar components and This fact and the more rapid a colorimetric assay for uronic acids. LC degradation of monomeric uronic acids determination is gaining in popularity. A at acidic condition are reasons why colorimetric measurement of reducing colorimetric methods are preferred for sugars has been introduced as an uronic acid determination (9). alternative to the GLC determination by Collaborative Studies of Dietary Fiber Englyst et al. (35). Analysis. A number of collaborative Advantages and disadvantages of studies of enzymatic, gravimetric dietary the two different ways of analyzing fiber determination has been carried out dietary fiber are summarized in Table III. within the AOAC. The component Enzymatic gravimetric methods are analysis method of Englyst and co- simple and robust with no requirement workers has been tested in studies of advanced equipment. There is a risk carried out by the Ministry of Agriculture, of overestimating the fiber content if Food and Fisheries in the UK (MAFF). other components remain in the residue. The studies reported prior to 1990 were However, this can be analyzed for any reviewed and compared (9). They show such contaminating components. gradually improved performance with Colorimetric methods can also be typical mean reproducibility (R95) inflated by unspecific reactions. Specific, values of 2–3 for both the gravimetric GLC or LC measurements on the other methods approved by the AOAC and for hand, require complete hydrolysis and the Englyst method. An AOAC study quantitative recovery of monomers after with the method of Theander et al. is about to be finished. The best amylose type, or lignin, would Englyst performance reported so far is a Swiss values be expected to be significantly study with the enzymatic gravimetric lower than estimates with methods AOAC method (R95=1.0–1.1) (36). An including these components. R95 value of 2.0 at a dietary fiber Delimitation problems in definition content of 10 g/100 g means that 19 out and analysis of dietary fiber have been of 20 single determinations coming from much focused on the inclusion or not of various laboratories would fall in the resistant starch and lignin. Equally range 9–11 g/100 g. important, however, is the delimitation There are few formal collaborative towards components that are not studies covering more than one method. precipitated in the 70–80 per cent (v/v) Usually, studies have included just one ethanol used in all the methods to or a few laboratories running a different separate water-soluble fiber methods, which makes strict components. Inulin is an undigestible intermethod comparison difficult. In a polysaccharide that is not precipitated recent study coordinated by the and therefore not recovered in any of European Community Bureau of the methods. Polydextrose is another Reference (BCR) dietary fiber values undigestible oligo-polysaccharide also with the AOAC method could be not determined as dietary fiber. As certified for three different materials. discussed above, these components Indicative values only could be given for should be grouped together with the the Englyst GLC and colorimetric dietary fiber rather than with digestible methods, but means with these methods carbohydrates. The same is true for the were similar to those obtained with the undigestible oligosaccharides. Specific AOAC method (37). enzymatic or HPLC methods then have For most foods, estimates of total to be employed for these components. dietary fiber with the enzymatic Difference methods are unspecific gravimetric method of the AOAC or and accumulate analytical errors from according to Theander et al. (both the fat, protein, ash and moisture including the retrograded amylose type determinations. When keeping these of resistant starch and lignin) would not limitations in mind, however, these be significantly different from estimates methods are capable of giving a of nonstarch polysaccharides with the reasonable estimate of “total available Englyst methods. This means that the carbohydrates” in many foods if the confidence intervals for the different dietary fiber is measured, e.g. with the methods overlap (9). It should be noted enzymatic gravimetric AOAC method. also that two collaborative studies have Difference calculations are also useful in shown consistently higher values with the laboratory to check the the colorimetric Englyst method than standardization of methods for with the original GLC variety (9). Only in proximate analysis. foods with particularly high levels of resistant starch of the retrograded Fibre— A Component of Food, T.F. Ŷ References Schweizer & C.A. Edwards (Eds.), (1) World Health Organization (1990) Springer-Verlag, London, pp. 137– Diet, Nutrition, and the Prevention 150 of Chronic Disease. Technical (3) Edwards, C.A., & Rowland, I. Report Series, 797, WHO, (1992) in Dietary Fibre—A Copenhagen Component of Food, T.F. (2) Rémésy, C., Demigné C., & Schweizer & C.A. Edwards (Eds.), Morand, C. (1992) in Dietary Springer-Verlag, London, pp. 119– (18) Asp, N.-G. (1990) in New 136 Developments in Dietary Fiber, I. (4) McCance, R.A., & Lawrence, R.D. Furda & C.J. Brine (Eds.), Plenum (1929) Medical Research Council Press, New York, pp. 227–236 Special Report Series No. 135, (19) Åman, P., & Graham, H. (1987) J. London, HMSO Agric. Food Chem. 35, 704–709 (5) McCance, R.A., & Widdowson, E.M. (20) Holm, J., Björck, I., Drews, A., & (1940) Medical Research Council Asp, N.-G. (1986) Starch/Stärke, Special Report Series No. 235, 38, 224–226 London, HMSO (21) Englyst, H.N., Wiggins, H.S., & (6) Official Journal of the European Cummings, J.H. (1982) Analyst, Community (1990) Directive NOL 107, 307–318 276/40 (22) Asp, N.-G. (1992) Eur. J. Clin. Nutr. (7) Trowell, H.C. (1972) Am. J. Clin. 46 (Suppl), S1 Nutr. 25, 926–932 (23) Asp, N.-G., & Björck, I. (1992) (8) Trowell, H.C., Southgate, D.A.T., Trends Food Sci. Technol. 3, 111– Wolever, T.M.S., Leeds, A.R., 114 Gassull, M.A., & Jenkins, D.J.A. (24) Johansson, C.-G., Siljeström, M., & (1976) Lancet, i, 967 Asp, N.-G. (1984) Lebensm. Unters (9) Asp, N.-G., Schweizer, T.F., Forsch. 179, 24–28 Southgate, D.A.T., & Theander, O. (25) Siljeström, M., Eliasson, A.-C., & (1992) in Dietary Fibre- A Björck, I. (1989) Starch/Stärke, 41, Component of Food, T.F. 147–151 Schweizer & C.A. Edwards (Eds.), (26) Russel, P.L., Berry, C.S., & Springer-Verlag, London, pp. 57– Greenwell, P. (1989) J. Cereal. Sci. 101 9, 1–15 (10) Asp, N.-G., Björck, I., & Nyman, M. (27) Berry, C.S. (1986) J. Cereal Sci. 4, (1993) Carbohydrate Polymers 21, 301–314 183–187 (28) Englyst, H.N., & Cummings, J.H. (11) Brand Miller, J.C. (1994) Am. J. (1990) in New Development in Clin. Nutr. 59 (Suppl), 747S–752S Dietary Fiber, Furda, I., Brine, J. (12) Truswell, A.S. (1992) Eur. J. Clin. (Eds.), Plenum Press, New York, Nutr. 46 (Suppl 2), S91–S101 NY, pp. 205–225 (13) Würsch, P. (1989) World Rev. Nutr. (29) Björck, I., Gunnarsson, A., & Diet. 60, 199–256 Östergård, K. (1989) Starch/Stärke (14) Björck, I., Granfeldt, Y., Liljeberg, 41, 128–134 H., Tovar, J., & Asp, N.-G. (1994) (30) Siljeström, M., Björck, I., & Am. J. Clin. Nutr. 59 (Suppl), 699S– Westerlund, E. (1989) 705S Starch/Stärke 41, 95–100 (15) Lingström, P., Holm, J., Birkhed, D., (31) Prosky, L., Asp, N.-G., Schweizer, & Björck, I. (1989) Scand. J. Dent. T.F., DeVries, J.W., & Furda, I. Res. 97, 392–400 (1988) J. Assoc. Off. Anal. Chem. (16) The British Nutrition Foundation. 71, 1017–1023 (1990) Complex Carbohydrates in (32) Lee, S., Prosky, L., & DeVries, J. Foods, Chapman and Hall, London, (1992) J. Assoc. Off. Anal. Chem. pp. 1–164 75, 395–416 (17) Greenfield, H., & Southgate, D.A.T. (33) Southgate, D.A.T. (1969) J. Sci. (1992) Food Composition Data Food Agric. 20, 331–335 Production, Management and Use, (34) Theander, O., Åman, P., Elsevier Applied Science, London, Westerlund, E., & Graham, H. pp. 94–104 (1990) in New Developments in Dietary Fiber. I. Furda & J. Brine Lebensmitteluntersuchung und (Eds.), Plenum Press, New York, Hygiene 79, 57–68 NY, pp. 273–281 (37) Hollman, P.C.H., Boenke, A., & (35) Englyst, H.N., Cummings, J.H., & Wagstaffe, P.J. (1993) Fres J. Anal. Wood, R. (1987) J. Assoc. Publ. Chem. 345, 174–179 Analysts 25, 73–110 (38) Englyst, H.N., Kingman, S.M., & (36) Schweizer, T.F., Walter, E., & Cummings, J.H. (1992) Eur. J. Clin. Venetz, P. (1988) Mitteilungen aus Nutr. 46, S33–S50 dem Gebeit der

Methods and Conventions of Nutrient Analysis

Recent Developments in the Determination of Water-Soluble Vitamins in Food—Impact on the Use of Food Composition Tables for the Calculation of Vitamin Intakes

Paul M. Finglas

Nutrition, Diet & Health Department, Institute of Food Research, Norwich Research Park, Colney, Norwich, NR4 7UA, UK

There is a need for improvements in the determination of vitamins in food, in particular, the establishment of properly validated and robust techniques that are applicable to a wide range of food matrices. This paper addresses three main topics: first, recent developments in methods including LC techniques and biospecific methods utilizing antibodies and naturally occurring vitamin binding proteins which are both based on the microtitration plate format; second, results from a European Union project under the Measurement and Testing Programme concerned with the improvement in vitamin analysis in food by intercomparisons of methods, optimization of extraction conditions and the preparation of food reference materials (RMs); and third, the impact of the improvements in methods on the quality of vitamin data currently presented in UK food tables by comparing calculated vitamin intakes obtained using both the 4th and 5th editions of McCance & Widdowson's The Composition of Foods and direct analysis of duplicate diets.

t is important that the method of analysis chosen for any vitamin should be that which most closely reflects the vitamin activity of the food in question since the primary Iobjective for use of the data is for nutritional purposes. The term “vitamin” reflects a certain physiological activity which is related to the chemical substances or “vitamers” responsible for this activity (1). Ideally methods would be chosen that could determine each vitamer separately and then by calculating the sum of the individual activities, a total activity of the food could be obtained (2). However, in practice this is rarely possible as procedures are not specific for the compounds of interest.

estimates of the biological activity of the Ŷ Methodology various forms, can be used to provide Liquid Chromatography better estimates of the vitamin activities of foods than are currently available in The major advantage of this technique food tables. is that individual forms of the vitamin can be measured, and together with Although LC techniques have been fluorescence but thiamin requires widely used for the determination fat- conversion to thiochrome with alkaline soluble vitamins in food, their application potassium ferricyanide solution. The to water-soluble vitamin analysis has latter can be performed either manually mainly been limited to vitamins B1, B2, prior to injection onto the analytical B6 and C. This has largely been due the column, or by post-column availability of sufficiently sensitive and derivatization. Reverse phase LC specific detection systems that are analysis of thiochrome invariably capable of quantifying several vitamers involves the injection of high salt from a complex mixture of compounds. concentrations on to the analytical The use of LC with UV detection has column necessitating frequent washing been used for the determination of these and much reduced column life (4). This vitamins in food but this form of can be overcome to some extent by the detection is generally not sufficiently use of guard columns but this may be sensitive nor specific due to the low uneconomic. Alternatively, thiochrome levels found, especially in unfortified can be selectively extracted into an foods (3). LC with fluorescence organic solvent, normally isobutanol, detection has been preferred using allowing fluorometric determination after either the natural fluorescence of the a normal phase LC separation of vitamin (e.g. riboflavin), or with thiochrome from any remaining derivatization to form a suitable fluorescing interferences (5). This fluorescent complex (e.g. thiamin and approach can give increased sensitivity vitamin C). This form of detection gives and rapid automated analysis without better sensitivity and specificity the need for post-column derivatization. compared to UV detection (3). Thiamin methods based on the Examples of LC procedures available thiochrome reaction after acid hydrolysis for selected water-soluble vitamins are and treatment with enzyme to release given in Table I. the phosphorylated forms give total Thiamin and Riboflavin. These thiamin concentrations. The separation vitamins are usually extracted from of thiamin and its phosphorylated forms foods using dilute mineral acids and [thiamin monophosphate (TMP), thiamin autoclaving at 121°C, followed by di- or pyro-phosphate (TPP) and thiamin enzymatic hydrolysis to release the triphosphate (TTP)] has also been bound forms of the vitamins. Riboflavin reported (6). can be measured directly with Table I. Selected LC procedures available for some water-soluble vitamins Vitamin Principle Column Mobile phase Detectiona Reference (nm) Thiamin 1 Post-column Silica (radpak) 0.05M F (265/418) 37 oxidation to Phosphate thiochrome buffer: EtOH 2 Pre-column oxidationµ-Bondapak C-MeOH:H2O F (365/435) 38 to thiochrome 18 3 Free thiamin + Micropak Ax5 Ammonium UV (245) 6 phosphorylated phosphate forms Riboflavin Native fluorescence Apex ODS 2 MeOH:H2O F (450/510) 37 (B2, FMN) Vitamin B6 1 Acid digestion with Spherisorb 0.04M F(290/395) 2 autoclaving (PM, PL, ODS 2 H2SO4:MeOH PN) 2 HC104 extraction + Lichrosphere 0.03M F (340/400) 9 post-column reaction RP 18 Phosphate with bisulphate (PM, buffer: PL, PN, PMP, PLP) McOH:ion pair 3 TCA extraction + Lichrosphere 0.1M F (325/385) 7 post-column reaction RP18 KH2PO4:EtOH with KH2PO4 acetonitrile: ion pair 4 Acid phosphatase Octysilyl 0.05 KH2PO4: F (290/395) 10 hydrolysis. PMĺPL acetonitrilc:ion using glyoxylic pair acid/Fe2+; PLĺPN using NaBH4/NaOH Vitamin C 1 Acid extraction + Hypersil ODS 0.8 M F (367/418) 39 oxidation with Phosphate ascorbate oxidase) buffer:MeOH (AA + DHAA) 2 Acid extraction (AA Partisil P5 H2O:MeOH: UV (248) Leatherhead only) acetic acid Food RA 3 Acid extraction, Apex ODS 2 Acetate EC 11 homocysteine buffer:NaOH reduction (Total AA, AA) Folates Hog kidney µ-Bondapak Phosphate F(9295/365) 14 deconjugation, anion phenyl buffer: exchange cleanup, acetonitrile post-column reaction with Ca(OCl)2 a Fluorescence (excitation and emission wavelengths); EC, electrochemical; UV, ultraviolet

Results from the manual thiochrome found to be the extraction/enzyme procedure agree well with the LC hydrolysis step (7). fluorometric procedures but are Similarly, for riboflavin, LC methods generally lower when compared to the generally give lower results compared to microbiological results (7). The main the microbiological assay (MA). Most cause of variability between laboratories test organisms used in the MA give an for the determination of thiamin was equivalent growth response to riboflavin and flavin mononucleotide (FMN), which are both vitamin active. In LC tact. Typical acids used include procedures, riboflavin is separated from sulfosalicylic acid, SSA (8), FMN and the latter is not measured. In trichloroacetic acid, TCA (7) and the same study as above several perchloric acid, PER (9). The SSA laboratories reported incomplete procedure uses anion exchange sample conversion of FMN to riboflavin and thus clean-up to remove SSA and interfering lower LC results compared to the MA. compounds prior to separation and The enzyme conversion of FMN to quantification on a similar column with riboflavin needs to be improved if LC is fluorometric detection. All five B6 forms used for quantification (7). and the internal standard, 3- Vitamin B6. LC provides a technique hydroxypyridine, are measured by that is capable of measuring in a single means of detector wavelength chromatographic run all vitamin B6 forms switching. The inclusion of an internal [pyridoxamine (PM), pyridoxal (PL), standard before sample extraction is pyridoxine (PN), pyridoxamine useful as corrections for any dilution phosphate (PMP) and pyridoxal errors and instrument variation can be phosphate (PLP)] and is now more made. This method has been compared widely used than the microbiological to a microbiological assay and assay in food composition work. All satisfactory agreement obtained for these five forms have equal biological meat samples (8). Differences were activity (1). found, however, between the two A major limitation of the MA is the methods for fruits and vegetables. This lower growth response to PM compared was attributed to incomplete extraction to PL and PN obtained using some of vitamers for the LC procedure (8). organisms. This can obviously result in The other two systems with TCA and a substantial underestimation in the total PER use reverse-phase LC with ion-pair vitamin B6 activity of foods that are high regents and a post-column reaction with in the PM form (8). Two approaches potassium dihydrogen phosphate (7) have been used in the development of and sodium bisulfite (9), respectively, to LC methods for vitamin B6 depending on enhance the flurometric response of the whether the separation and various forms. The internal standard quantification of the phosphorylated used was 4-deoxypyridine. forms are required in addition to PM, PL Chromatograms for these procedures and PN. The normal acid digestion with tend to give fewer interfering peaks and autoclaving and enzymatic digestion is better resolution compared to those LC used for the extraction of total amounts methods measuring total amounts of of PM, PL and PN and these forms can PM, PL and PN forms. For the be quantified using a reverse phase LC determination of vitamin B6 in fortified system with fluorometric detection (2). foods where PN is predominantly This type of extraction gives an array of present, LC systems can be greatly potential interfering compounds that can simplified. make accurate quantification difficult. In A recent LC procedure using ionpair addition, the conversion of PMP to PM reagent and fluorometric detection has can be incomplete unless autoclaving been reported (10). After for 2 hours is performed. However, such dephosphorylation using acid conditions may lead to thermal phosphatase, PM is transformed to PL 2+ degradation of vitamin B6 vitamers (9). with glyoxylic acid/ Fe as catalyst, The second type of extraction system followed by conversion of PL to PN developed for vitamin B6 employs less using alkaline sodium borohydride. The vigorous conditions in order to keep the conversion of PM to PL was reported to phosphorylated forms (PMP and PLP) in be 95 per cent but it was found the conversion of PL to PN was always less Ag/AgCl electrode (11). Concentrations than 100 per cent and variable from of DHAA have been found to be fairly food to food. The procedure was only low in individual foods and mixed diets tested on four foods [yeast, wheat germ compared to AA (12). Good agreement and two types of breakfast cereals (with between the vitamin C results obtained bran and muesli)] and no comparison using various types of LC methods and with the MA was made. However, the the manual fluorometric procedure has method has the advantage of much been found (7). simplified chromatograms with only the Folates. Various LC methods have separation and quantification of PN been developed for the determination of required. folates but most of these procedures Vitamin C. Vitamin C activity is have only been applied to standard exhibited by L-ascorbic acid (AA) and mixtures and fortified foods such as Ldehydroascorbic acid (DHAA) but not infant foods and breakfast cereals by D-isoascorbic acid (erythorbic acid, where pteroylmonoglutamic acid (PGA) IAA) (1). The latter is also used as an is the predominant form. When these antioxidant in foods and it is therefore methods have been applied to natural important to have methods that can levels found in most foods, extensive measure both AA and DHAA either sample clean-up and purification is separately or together, and required (13). differentiating them from IAA (1). Most of the LC methods developed A variety of LC methods have been have concentrated on the analysis of reported for the determination of vitamin folate monoglutamates which requires C in food and biological materials have conjugase treatment of samples prior to been reported either as a total value analysis. The selection of a (AA+DHAA), or with separate deconjugase enzyme and of reaction determinations of each form. The latter conditions that provide complete can increase the complexity and length hydrolysis of folate polyglutamates to of the procedure and introduce the monoglutamate level is essential for additional errors and thus LC methods accurate LC quantification (14). Three which give total vitamin C values are types of deconjugase enzymes are preferred (1). Three types of detection available: hog kidney (HK), human systems have been used. Fluorescence plasma (HP) and chicken pancreas detection requires the oxidation of AA to (CP). The first two enzymes produce DHAA with ascorbate oxidase or mainly monoglutamate products charcoal followed by a reaction with o- whereas CP gives essentially phenylenediamine to form a fluorescent diglutamates. Although CP may be less quinoxaline derivative (1). These susceptible to inhibition by certain food methods measure total vitamin C, i.e. components compared to HP and HK ascorbic and dehydroascorbic acids. enzymes (15), it is not advisable to use Those LC methods that employ UV this enzyme in chromatographic studies detection are less specific and can only as identification and quantification of measure AA since DHAA is non-UV diglutamate products is difficult due the absorbing (2). In addition, they may not lack of availability of commercial separate AA from IAA (2). standards for these forms. The most recent form of detection Although HK enzyme gives used for vitamin C is electrochemical essentially monoglutamate products, which is extremely sensitive. DHAA is deconjugase activity can be low reduced to AA by homocysteine and the compared to CP necessitating extensive total AA measured enzyme clean-up and purification before electrochemicallychemically using an use (16). There has been renewed interest in (Aspergillus oryzae; Sigma) and the use of HP as the deconjugase Pronase (Calbiochem-Behring Corp.) enzyme because it is readily available, it has been reported to give higher folate can be used without any purification, it values in some starchy foods compared contains low levels of folate and it gives to CP alone (18). Similar results have a monoglutamate end-point (17). also been found in cereal products using However, a recent collaborative study CP, a heat resistant Į-amylase has found it may not be as effective as (Termamyl; Novo) and CP or HK in three foods (wheat flour, amyloglucosidase (Sigma) (19). It is milk flour and yeast powder) (Lumley & clear that whatever conjugase enzyme Finglas, unpublished data). A triple is used for folate deconjugation, it enzyme combination consisting of CP should be fully optimized for each food (acetone preparation; Difco), Į-amylase analyzed.

Table II. Folate content (nmol/g) and distribution in selected foods using LC and microbiological assay (MA)a LC MA

Food THF 5-CH3-THF 5-CHO-THF Total Whole cow's milk ndb 0.07 nd 0.07 0.03 Fresh cabbage 0.07 0.11 nd 0.18 0.29 Orange juice nd 0.37 nd 0.37 1.90 Wholewheat flour nd nd nd 0 1.59 a Adapted from (14) b nd, not detected There has also been recent interest fluorometer during the chromatographic in being able to use LC in quantifying run to give greater sensitivity. For PGA, not only any type of folate in respect to THF and dihydro-folic acid, a oxidative its carbon substitution but polyglutamate post-column derivatiation system using forms as well (20). This would obviously calcium hypochlorite converts these eliminate the need for the lengthy compounds to highly fluorescent pterin deconjugation step and facilitate the use compounds. The agreement between of simplified extraction systems. this LC procedure and the MA for six The most promising LC procedures foods was found to be poor (Table II) are those based on methods and further work is needed to explain incorporating anion-exchange clean-up the reasons for this. followed by hog kidney deconjugation Microbiological Procedures and quantification using the native fluorescence of the reduced folates in Despite many of the drawbacks an acidic mobile phase (14). Extraction concerning the use of microbiological conditions were chosen so that 10- assays for the analysis of water-soluble formyl-tetrahydrofolic acid (10-CHO- vitamins, they are still widely used for THF) was quantitatively converted to 5- folate, vitamin B12, pantothenic acid, formyl-tetrahydrofolic acid (5-CHO- niacin and biotin. Over the years MAs THF). While tetrahydrofolic acid (THF) have received a great deal of attention and 5-methyl-tetrahydrofolic acid (5- especially for folates (17). Although the CH3THF) can be readily quantified using MA has been modified and improved, their natural fluorescence, the relatively the basic concept of the assay has not weak fluorescence of 5-CHOTHF is a changed. limiting factor and requires careful In our laboratory, the response of control of the slit widths of the folate monoglutamates to L. rhamnosus (formerly L. casei) has been extensively studied (21–23) and a pH of 6–6.2 found well is measured using a microplate to give equivalent response of the major reader at 600–670 nm, the entire plate forms over the 0–1 ng calibration range. taking 2–3 minutes to read completely. The most recent development in the Seven levels of standard and 16 MA for folates has been the introduction samples (all in quadruple) are included of the 96-well microtitration plate based on each plate, with ten plates in an assay (24, 25). In our experience this assay run. Considerable time is saved has improved assay reliability and using the microtitration plate assay over increased sample through-put conventional procedures using test dramatically. Bacterial growth in each tubes.

Table III. Biospecific methods currently developed for the determination of watersoluble vitamins in fooda Vitamin Type Immunochemicals Foods analysed Reference Biotin EBPA 1.Avidin-HRPb, Biotin-KLH 1.Liver 1.27 2.Avidin/Streptavidinb Biotin- 2.Breakfast cereal 2.39 HRP Biotin ELISA 1.Anti-biotin antiserac Biotin- 1.Liver 1.- KLH 2.Anti-biotin antiserab Biotin- HRP Pantothenic acid ELISA Anti-pantothenic acidc Milk, eggs, bread, potatoes, 26 (PA) antisera, PA-KLH liver, lettuce Bread, milk, rice, vegetables, 30 chicken, beef Pyridoxamine ELISA Anti-pyridoxamine antiserac, Breakfast cereals 29 Cyanocobalamin EPBA PL-KLH R-protein-HRP, B12- (B12) KLH Folates, total EPBA 1. FBP-HRP, PGA-KLH 1.Vegetables, animal feeds 1.27 2. FBP-HRP, BSA-KLH 2.Breakfast cereals, fruit 2.39 juices PGA ELISA 3.Anti-folate antiserab PGA- 3.Standard mixtures 3.- KLH 5-CHO-THF 4.Anti-folic antiserab PGA- 4.Multivitamin preparations 4.39 5-CH3-THF HRP a Adapted from (44) b Commercially available c In-house antisera Biospecific Procedures sample through-put. ELISAs have the greater potential for the measurement of During the last five years biospecific specific vitamers, whereas EPBAs is methods of analysis using either more useful for broad specificity assays antibodies [enzyme-linked where the analysis of a group of immunosorbent assay (ELISA) format], vitamers is required (e.g. folates). or naturally occurring vitamin binding Most of the ELISAs developed at proteins [enzyme protein-binding assay Norwich have used in-house antisera (EPBA) format], have been shown to be which can take several months to capable of producing sensitive and produce. Antisera have been raised specific assays for a number of B-group against the major, naturally occurring vitamins (Table III). Both types of assay forms of each vitamin. There is, are based on the microtitration plate however, a range of commercial format which is ideally suited to high antisera (polyclonal and monoclonal) currently available to several vitamins vitamin B12 by EPBA and MA (27–29) including PGA, 5-MTHF, 5-CHO-THF, and PM by ELISA and LC in a range of vitamin B12 and biotin. This should foods (30). In general, the agreement facilitate the development of suitable between the results obtained using the ELISAs for food use. biospecific methods and the more There are limited comparative data conventional techniques has been good available comparing results from for those foods analyzed. These biospecific methods to MA and HPLC techniques also offer the potential of procedures. We have compared the developing simplified extraction systems determination of pantothenic acid by as it should be possible to measure ELISA and MA (26); biotin, folate and bound vitamin forms. Table IV. Food reference materials (RMs) currently under stability testing for fatand water-soluble vitaminsa RM Fat-soluble Water-soluble Margarineb A,E,D,ȕ-carotene - Wholemcal flour - B1, B2, B6, folate, niacin, biotin c Milk powder A, E, D, ȕ-carotene C, folate, B12, B1, B2, B6 d Brussels sprouts -C,B1, B2, folate, B12, niacin d Pig liver A, E, D, ȕ-carotene Folate, B12, C, B1, B6, biotin d Mixed vegetables Carotenoids C, B1, B2, folate, B12 a Adapted from (45) b Canned product c Vtamin enriched d Lyophilised dry foods but taking care to control the Ŷ Community Bureau of Reference (BCR) lyophilization conditions (time, temperature etc.) in order to minimize In 1988, BCR undertook a research vitamin losses. The final materials have project to improve the quality of vitamin been packaged into food grade, analysis carried out by its Member aluminum sachets under an inert States for nutritional labeling purposes atmosphere. The homogeneity and and the production of food composition stability of a wide range of vitamins are data. The project involves about 50 being vigorously tested for each RM. European laboratories specialized in vitamin determinations and includes Method Intercomparison Studies method intercomparison studies to The first intercomparison of methods for identify and eliminate sources of error, vitamins was organized to study the and the preparation of suitably stable state of the art in a group of European and homogeneous food reference laboratories experienced in vitamin materials (RMs). determinations. A summary of the main Reference Materials (RMs) results is given in Table V. The agreement between the participants for The use of certified RMs is an essential vitamins B1 and C, and niacin was good part of quality assurance and provides a and indicative values were given for means by which results can be traced three candidate RMs. back to a certified value. RMs are For vitamin B6, two problem areas especially useful in the development of were identified. Firstly, the identification new techniques. The range of RMs of the various vitamers proved difficult currently under stability testing are given and consequently there were large in Table IV. The approach used in the variations in both individual vitamers preparation of RMs has been to produce and total B6 values obtained by the participants (Table VI). Secondly, it was included in the total B6 activity. The found that glucoside derivatives of microbiological results for B6 were higher pyridoxine were not hydrolyzed by due to inclusion of the glucoside forms. phosphatase and takadiastase enzymes There was also large variation in the used and required an additional ȕ- results for vitamin B2 (Table V) and it glucosidase enzyme to release these was concluded that this was due to the forms for LC analysis. However, these hydrolysis/de-phosphorylation step of forms are unlikely to be absorbed by the procedure. man and should therefore not be

Table V. Summary of the results of the First BCR Intercomparison of Vitamins in the candidate food reference materials Vitamin Methods usedb Number of Within-and between- Comments laboratories laboratory variation (%CV)

%Cvw %CVB Thiamin LC, fluorometric, MA 10 3–5 11–18 Indicative values given for three RMs Riboflavin LC,MA 12 4–7 28–74 Problems in extraction/dephosphoryl ation step Vitamin B6 LC, MA 6 4–7 18–51 Problems in peak identification of extraction/hydrolysis Niacin MA 7 3–5 9–15 Indicative values given for three RMs Vitamin C LC, fluorometric 10 5 15 Indicative value given for one RM a adapted from (7) b LC, liquid chromotography; MA, microbiological assay

Table VI. Results for B6 content and individual vitamers in three food reference materials (RM)a. Results for individual vitamers are means (with ranges in parentheses) from five laboratories using LC procedures.

Individual vitamers Total B6 (% of total B6) (mg/100 g dry mass) RM PMb PLb PNb Mean %CV Milk powder 25 70 6 0.33 18 (21–31) (60–78) (0–16) Pork muscle 22 67 11 1.32 35 (0–47) (53–76) (0–31) Haricot vert beans 50 31 19 0.17 22 (26–100) (0–54) (0–36) a Adapted from (7) b PM, pyridoxamine; PL, pyridoxal; PN, pyridoxine. Figure 1. BCR-intercomparison on the determination of folates in food Determination of folate in a brussels sprouts RM using microbiological assay (MA), enzyme protein-binding assay (EPBA), radioassay kit (RIA) and liquid chromatography (LC). (¨), chicken pancreas, (×) human plasma and (+) hog kidney deconjugase enzymes. Results are means +/- 1SD (Reprinted with permission from Food Chemistry, Copyright 1992, Elsevier Applied Science Publishers Ltd).

Few laboratories were able to and poor calibration were found in the perform analyses for folate and vitamin LC procedures used. In general, EPBA B12 and thus a second intercomparison and RIA results were higher than MA was organized for folate analysis in a and LC values but much more variable. lyophilized brussels sprouts material It was concluded that the response of using MA, LC, EPBA and radio-protein the individual folate forms to the binding binding kit (RIA) (13). Three types of protein used in the assay is crucial and deconjugation were investigated: HP, careful control of assay pH and type of CP and HK. Good agreement was calibrant is required if an equivalent obtained with laboratories using MAs response to the main folate forms is to (Figure 1). CP deconjugation gave be obtained. Further work is needed about 20 per cent higher folate levels in before LC and other techniques can be this foodstuff compared to HP. The use used for routine folate analysis in food. of autoclaving followed by CP or HP The extraction conditions for the LC deconjugation gave lower (10–20 per determination of vitamins B1, B2 and B6 cent) levels determined by MA when in three food materials (pork muscle, compared to refluxing and milk powder and wholemeal flour) have deconjugation with the same enzymes. also been investigated. The aim of this Problems in the identification of peaks study was to optimize conditions for enzyme hydrolysis (pH, time, Many nutritional studies have relied temperature and sample-enzyme ratio). upon the 4th edition of McCance and Four commercially available enzymes Widdowson's The Composition of Foods which are most commonly used were (MW4) for the calculation of nutrient selected: takadiastase (Pfaltz & Bauer, intakes (31). This has been superseded Serva), phosphatase (Sigma) and a by a revised 5th edition (MW5) (32) mixture of takadiastase (Fluka) and containing a much wider range of foods phosphatase (Sigma). Optimum levels and new analytical data using improved of B6 in pork and milk powder were methods of analysis. found using either takadiastase (Pfaltz & One way of assessing the impact of Bauer) or phosphatase (Sigma) with a improved vitamin methods is to compare pH of 4.8–5.5, an overnight incubation vitamin intake values computed from and a temperature of 37–45C. The food composition tables with values minimum amounts of enzyme needed obtained by direct analysis of diets were 500 mg takadiastase or 50 mg consumed. This is also important as phosphatase per gram of sample. For B1 many nutritional studies in the literature in pork and flour, optimum levels were have relied on MW4 for the calculation only found using takadiastase (Pfaltz & of vitamin intakes and it is therefore Bauer) with a pH of 4 and incubation essential to assess the likely impact on time of four hours. The minimum the interpretation of such studies if MW4 amount of enzyme needed was 100 is now replaced with revised data given mg/g pork and 20 mg/g flour. The in MW5. conversion of thiamin monophosphate In a study in Norwich, nutrient intake to free thiamin was only about 30 per data, collected from 54 adolescents cent using the using a 7-day weighed inventory takadiastase/phosphatase mixture and (recorded every 6th day for 7 weeks), these enzymes should be avoided for calculated using MW4 with the same LC work. Similar results were obtained data calculated using MW5. In addition, for B2 in pork and flour except an intake values obtained using MW4 and overnight incubation and higher enzyme MW5 are compared with values levels (100 mg takadiastase/g pork, 20 obtained by direct analysis of duplicate mg takadiastase/g flour) were required. diets collected on the same day of The conversion of flavin mononucleotide dietary recording (33, 34). In particular, (FMN) to free vitamin was not complete MW5 versus analyzed values are using both enzyme preparations. The examined for those nutrients where the takadiastase/phosphatase mixture gave agreement was found to be poor when about 60 per cent conversion whereas calculated intake values were computed takadiastase (Pfaltz & Bauer) gave using MW4, for example folate and nearly 80 per cent conversion. This is vitamin B6 (33). A summary of methods not a problem in the MA as FMN gives used to analyze duplicate diets for an equivalent growth response selected water-soluble vitamins are compared to riboflavin. given in Table VII. Typical methods used for food composition data for these Ŷ Impact of Improvements in the Determination of Water-Soluble vitamins in MW4 and MW5 are also Vitamins on the Quality of Food included. Composition Data in UK Food Tables Table VII. Summary of methods used to analyse duplicate diets of schoolchildren in Norwich study for energy and selected water-soluble vitamins and typical methods used for UK food composition tables (MW4 & MW5)a Nutrient Methodsb used for Methodse used for UK food composition folates duplicate diet analysis (reference) (reference) MW4 MW5 Energy Bomb calorimetry Calculation from protein, Calculation from protein, corrected to metabolizable fat, carbohydrate and fat, carbohydrate and energy (40) alcohol alcohol Thiamin (B1) LC/F (41, 42) F (43), MA (35) MA (34), F (43), LC/F (37) Riboflavin LC/F (41, 42) MA (34), F (43) MA (34), LC/F (35, 37) (B2) c Vitamin B6 LC/F (2, 42) MA (34) MA (34), LC/F (2) Vitamin Cd LC/EC (11,42) T (AA only; 43), T (AA only; 43) F (Total; 43) F(total; 43) Folate (total) MA (21,42) MA (34) MA (34), MA (21) a (31); (32) b LC/F, liquid chromatography with fluorometric detection; LC/EC, liquid chromatography with electrochemical detection and MA, microbiological assay using Lactobacillus rhamnosus at pH 6.2 c Total vitamin B6 = pyridoxamine (PM) + pyridoxal (PL) + pyridoxine (PN) d Total vitamin C = ascorbic acid (AA) + dehydroascorbic acid (DHAA) e F, fluorometric & T, titrimetric. For MW4, MA (riboflavin): Streptococcus zymogenes; MA (thiamin): Lactobacillus viridescens or L. fermienti; MA (B6): Saccharomyeces carlsbergensis and MA (folate): Lactobacillus casei. For MW5, MA (thiamin): L. viridescens; MA (B6): Kloeckera apiculata; MA (folate): Lactobacillus rhamnosus & MA (B2): Streptococcus zymogenes

The possible extent of under- conditions, i.e. initial pH of 6.8 for recording of total food intake in this organism growth and 0–1 ng calibration study was assessed using mean daily range, the growth response of analyzed energy intake values and Lactobacillus rhamnosus (casei) to 5- estimates of daily energy expenditure MTHF was poor in comparison to PGA, for similar subjects. It was concluded the normal folate used to calibrate the that there was no gross under- assay. If the pH is lowered to 6.2, the estimation of habitual energy intake and response of 5-MTHF and PGA is the the calculated nutrient intake values same (21). Although PGA is used for represented a good reflection of their food fortification because of its greater habitual intake (12). The results for stability and lower cost, it does not occur energy and water-soluble vitamins are naturally. The implication of this is that given in Table VIII for girls only. Similar in foods that contain appreciable trends were found for boys (34). amounts of natural folates, the folate The major differences in vitamin values given in MW4 are likely to be intake values calculated from MW4 and grossly under-estimated. This has MW5, and compared to analyzed intake subsequently led to an apparent values, were found for folate and vitamin underestimate in the true folate intake in B6. the UK (22, 23). Analysis of 128 Folates vegetables for folate using an improved MA found that the revised folate values In the UK, the microbiological assay were about two-fold higher compared to conditions of Bell (35) were used to MW4 data (36). obtain values for the folate content of foods given in MW4. Under these Table VIII. Average daily intakes of energy and water-soluble vitamins calculated from MW4, MW5 and direct analysis of duplicate diets Values are means with standard errors for 35 girls (13–14 year olds) with ranges in parentheses.a Nutrient Calculated Analyzedb Correlation Calculated Analyzedb Correlation MW4d coefficient, rc MW5d coefficient, rc Energy 7.23±0.26 7.36±0.24 0.94 7.18±0.24 7.36±0.24 0.95 (MJ) (4.3–11.1) (4.5–11.00) (4.2–10.9) (4.5–11.0) Thiamin 1.1±0.05** 1.5±0.1 0.43 1.3±0.07* 1.5±0.1 0.19 (mg) (0.6–1.8) (0.6–2.7) (0.8–2.6) (0.6–2.7) Riboflavin 1.3±0.08* 1.2±0.07 0.77 1.3±0.07 1.2±0.07 0.78 (mg) (0.4–2.6) (0.4–2.3) (0.4–2.4) (0.4–2.3) Vitamin 1.1±0.04* 0.7±0.04 NSe 1.7±0.08 0.7±0.4 0.35 B6 (mg) (0.7–1.6) (0.4–1.4) (0.9–3.2) (0.4–1.4) Vitamin C 75±7 84±5 0.66 71±6 84±5 0.56 (mg) (19–182) (43–153) (26–144) (43–153) Folate 145±8** 252±17 0.42 212±9* 252±17 0.66 (mg) (61–278) (96–510) (103–340) (96–510) a Adapted from (42) b For details of methods see Table VII c Correlation coefficient (r) between calculated MW5 & MW4 and analyzed data for girls [statistical significance of r=P<0.05] d Calculated MW4/MW5 intake was significantly different from analyzed data: * P<0.05, ** P<0.001 e NS, not significant

In this study, calculated folate intake may be underestimated (34). This values using MW5 are considerably possible underestimation will vary from higher (50 per cent) than MW4 intake food to food depending on the relative values and much nearer the values amounts of each vitamer present. Much obtained by duplicate diet analysis. The of the newer data for vitamin B6 revised folate content of several appearing in MW5 has been obtained breakfast cereals, especially cornflakes using the MA with Kloeckera apiculata and muesli, using the improved MA which should give better estimates for were much higher and made a the total vitamin activity than other significant contribution to the total daily organisms used. This is likely to have intake for this vitamin. Although a weak contributed to the large increase in correlation was found for between MW4 calculated intake values found when calculated and analyzed intake values using MW5 compared to MW4. The for girls, no association was found for vitamin B6 MW5 values of muesli and boys (33). However, the correlation cornflakes, which are two of the major between MW5 and analyzed intake contributors to total daily intake of this values is improved, and significant for vitamin, are considerably higher than both sexes. values appearing in MW4 for these foods. Vitamin B6 Some of the vitamin B6 values Some doubt has been expressed over appearing in MW5, for example raw and the vitamin B6 data appearing in MW4, cooked vegetables, have been obtained which has largely been obtained by MA by an LC procedure which permits the using a variety of test organisms, some separation and quantification of PM, PL of which may not respond equally to all and PN forms and the sum of which forms and thus total vitamin B6 activity gives the total vitamin B6 activity (36). Although the analyzed intake values are (2) Brubacher, G., Müller-Mulot, W., & about 50 per cent lower than the revised Southgate, D.A.T. (1985) Methods MW5 calculated intake values for total for the Determination of Vitamins in vitamin B6, there is now a significant Food, Elsevier Applied Science, correlation between the data for both London, pp. 129–140 sexes, which was not found for the MW4 (3) Finglas, P.M., & Faulks, R.M. calculated intake versus analyzed (1987) J. Micronutr. Anal. 3, 251– values. Clearly further improvements in 283 LC methodology is needed for this (4) Ang, C.Y.W., & Moseley, F.A. vitamin, especially extraction and peak (1980) J. Agric. Food Chem. 28, identification, before more reliable food 483–486 composition data can be obtained. (5) Bailey, A.L., & Finglas, P.M. (1990) J. Micronutr. Anal. 7, 147– 157 Ŷ Conclusions (6) Hilker, D.M., & Clifford, A. J. (1982) There has been a steady improvement J. Chromatogr. 231, 433– 438 in methods for the determination of (7) Hollman, P. C. H., Slangen, J.H., watersoluble vitamins in food over Wagstaffe, P.J., Faure, U., recent years notably with the Southgate, D.A.T., & Finglas, P.M. development of LC and biospecifc (1993) Analyst. 118, 481–488 procedures. These techniques allow the (8) Polansky, M.M., Reynolds, R.D., & quantification of individual vitamers and Vanderslice, J.T. (1985) in Methods thus give better estimates of the vitamin of Vitamin Assay, J.A. Augustin, activities of foods than is currently B.P. Klein, D. Becker & P.B. available in food composition tables. Venugopal (Eds.), John Wiley & Further work is needed on the Sons, New York, pp. 427–428 optimization of (9) Bitsch, R., & Moller, J. (1989) J. extraction/dephosphorylation conditions Chromatogr. 463, 207–211 and comparative data between (10) Reitzer-Bergaentzle, M., Marchioni, methods. The availability of a range of E., & Hasselmann, C. (1993) Food certified RMs will greatly assist in Chem. 48, 321–324 method validation and improving the (11) Behrens, W.A., & Madere, R. quality of data generated for food (1987) Anal. Biochem. 165, 102– composition tables. Care should also be 107 exercised when interpreting dietary (12) Finglas, P.M., Bailey, A.L., Walker, intake data from nutritional studies A., Loughridge, J.M., Wright, A.J.A., particularly the source, limitations and & Southon, S. (1993) Br. J. Nutr. reliability of the analytical techniques 69, 563–576 used to acquire them. (13) Finglas, P.M., Faure, U., & Ŷ Acknowledgments Southgate, D.A.T. (1993) Food Chem. 46, 199–213 Various parts of this work were (14) Gregory, J.F., Sartain, D.B., & Day, supported by the Office of Science and B.P.F. (1984) J. Nutr. 114, 341–353 Technology and the European (15) Eigen, E., & Shockman, G.D. Communities' Community Bureau of (1963) Fed. Proc. 42, 2105 Reference. (16) Keagy, P.M. (1985) in Methods of Ŷ References Vitamin Assay, J. Augustin, B.P. (1) Greenfield, H., & Southgate, D.A.T. Klein, D.A. Becker & P.B. (1992) in Food Composition Data, Venugopal (Eds.), John Wiley & Elsevier Science Publishers, Sons, New York, pp. 450–452 London (17) Tamura, T. (1990) in Folic Acid Metabolism in Health & Disease, M.P. Picciano, E.L.R. Stokstad & Composition of Foods, 5th Ed., J.F. Gregory (Eds.), Wiley-Liss, Royal Society of Chemistry, New York, pp. 121–137 Cambridge (18) De Souza, S., & Eitenmiller, R. (33) Southon, S., Wright, A.J.A., Finglas, (1990) J. Micronutr. Anal. 7, 37–57 P.M., Bailey, A.L., & Belsten, J. (19) Goli, D.M., & Vanderslice, J.T. (1992) Proc. Nutr. Soc. 51, 315– (1992) Food Chem. 43, 57–64 324 (20) Gregory, J.F. (1985) in Methods of (34) Bell, J.G. (1974) Lab. Pract. 23, Vitamin Assay, J. Augustin, B.P. 235–242, 252 Klein, D.A. Becker & P.B. (35) Kwiatkowska, C.A., Finglas, P.M., & Venugopal (Eds.), John Wiley & Faulks, R.M. (1989) J. Hum. Nutr. Sons, New York, pp. 473–496 Diet 2, 159–172 (21) Phillips, D.R., & Wright, A.J.A (36) Schrijver, J., Speek, A.J., Klosse, (1982) Br. J. Nutr. 47, 183–189 J.A., van Rijn, H.J.M., & Schreurs, (22) Wright, A.J., & Phillips, D.R. (1985) W.H.P. (1982) Ann. Clin. Biochem. Br. J. Nutr. 53, 569–573 19, 52–56 (23) Finglas, P.M., Wright, A.J.A., (37) Finglas, P.M., & Faulks, R.M. Faulks, R.M., & Southgate, D.A.T. (1984) Food Chem. 18, 37–44 (1990) in Recent Knowledge on Iron (38) Speek, A.J., Schrijver, J., & and Folate Deficiencies in the Schreurs, W.H.P. (1984) J. Agric. World, Vol. 197, S. Hercberg, P. Food Chem. 32, 352–355 Calan & H. Dupin, (Eds.), Inserm, (39) Rubach, K., & Reichert, N. (1991) Paris, pp. 385–392 Deutsche Lebensmittel-Rundschau (24) Newman, E.M., & Tsai, J.F. (1986) 88, 341–347 Anal. Biochem. 154, 509–515 (40) Miller, D.S., & Payne, P.R. (1959) (25) Horne, D.W., & Patterson, D. Br. J. Nutr. 13, 501–508 (1988), Clin. Chem. 34, 2357–2359 (41) Bailey, A.L., & Finglas, P.M. (1990) (26) Finglas, P.M., Faulks, R.M., Morris, J. Micronutr. Anal. 7, 147– 157 H.C., Scott, K.J., & Morgan, M.R.A. (42) Southon, S., Wright, A.J.A., Finglas, (1988) J. Micronutr. Anal. 4, 47–59 P.M., Bailey, A.L., & Loughridge, (27) Finglas, P.M., Faulks, R.M., & J.M. (1994) Br. J. Nutr. (in press) Morgan, M.R.A. (1986) J. Micronutr. (43) Official Methods of Analysis (1975) Anal. 2, 247–257 12th Ed., AOAC, Arlington, VA (28) Finglas, P.M., Kwiatkowska, C., (44) Finglas, P.M., Alcock, S.A., & Faulks, R.M., & Morgan, M.R.A. Morgan, M.R.A. (1992) in Food (1988) J. Micronutr. Anal. 4, 309– Safety and Quality Assurance: 322 Applications of Immunoassay (29) Alcock, S.A., Finglas, P.M., & Systems, M.R.A. Morgan, C.J. Morgan, M.R.A. (1992) Food Chem. Smith, & P.A. Williams (Eds.), 45, 199–203 Elsevier Applied Science, London, (30) Alcock, S.A., Finglas, P.M., & 401–409 Morgan, M.R.A. (1990) Food Agric. (45) Finglas, P.M., Faure, V., & Immunol. 2, 197–204 Wagstaffe, P.J. (1993) Fres. J. (31) Paul, A.A., & Southgate, D.A.T. Anal. Chem. 345, 180–184 (1978) in McCance & Widdowson's The Composition of Foods, 4th Ed., HMSO, London (32) Holland, B., Welch, A.A., Unwin, I.D., Buss, D.H., Paul, A.A., & Southgate, D.A.T. (1991) in McCance & Widdowson's The Methods and Conventions of Nutrient Analysis

Update on the Analysis of Total Lipids, Fatty Acids and Sterols in Foods

Andrew J. Sinclair

Department of Medical Laboratory Science, Royal Melbourne Institute of Technology, GPO Box 2476 V, Melbourne, Vic 3001, Australia

In the last 20 years, there has been an increasing awareness of the nutritional importance of lipids in foods. This has led to a requirement to improve the quality and quantity of data on food lipids in food databases. This paper discusses the extraction of lipids from different food matrices, the use of manual versus automated procedures and problems which occur during the extraction process. Methodological approaches are discussed for the analysis of fatty acid composition and concentration, and the analysis of the sterols by gas chromatography and high pressure liquid chromatography. This paper raises the future requirement for a wider range of food lipid data including quantitative information on tocopherols and tocotrienols, molecular species of triacylglycerols, distribution of fatty acids on the 2-position of the triacylglycerols, cholesterol oxides and other lipid oxidation products.

ipids are an important group of substances found in food where they make major contributions to taste, flavor and the energy content of the food. Lipids are a Lheterogeneous class of compounds which makes it difficult to provide a precise definition, however they are classified as those substances insoluble in water and soluble in a range of organic solvents. The information required on lipids for a food database include the total lipid content for the calculation of energy content as well as comparison between foods, the fatty acid types which include the saturates, cis- and trans-monounsaturates and Ȧ-6 and Ȧ-3 polyunsaturates, and the sterol content and composition, including the proportion of cholesterol in the food. While greater than 95 per cent of most food lipids in westernized countries consist of triacylglycerols (TAG), there are other lipids which need to be considered because of their presence in high concentration in certain foods. These include wax esters (found in high concentration in certain species of fish (1) and phospholipids (found in high concentration in eggs). Fatty acids, although rarely present in foods as such, are major components of all food lipids apart from sterols. Interested readers are referred to specialized books on lipids and lipid analysis (2–5). with glass stoppers or Teflon-lined caps Ŷ Extraction of Samples is mandatory for the extraction and The choice of technique to be adopted subsequent processing steps and use of for the analysis of total lipids can redistilled solvents has been depend on a number of factors including recommended, although highly purified the type of food being analyzed and solvents are available from suppliers. therefore the lipid classes present, the Supercritical CO2 can also be used to number of samples to be analyzed and extract lipids from foods followed by the laboratory facilities. The general weighing of the residue (5, 14). This approach to the extraction of lipids from technique has a number of advantages biological materials is to denature including extraction of the samples at lipoproteins and enzymes in alcohol and relatively low temperatures, and use of a to extract the lipids into an organic non-toxic and inert gas which may solvent. Simple soxhlet extraction in satisfy regulatory authorities who are ether has been used in the case of concerned with the use and disposal of foods with very high concentrations of hazardous solvents in laboratories. TAG, such as meat fat and milk fat however this can lead to poor extraction Ŷ Removal of Non-Lipid of the more polar lipids (6). There have Contaminants been many different procedures Most polar solvents also extract non- published for the extraction of lipids (2, lipids such as sugars, urea, amino acids 5) including refluxing liver tissue in and salts. It has been claimed that pre- ethanol (7), use of n-butanol saturated extraction with 0.25 per cent acetic acid with water for cereals (8), the use of will remove contaminants and destroy isopropanol as a preliminary extractant lipolytic enzymes (15), however this for plant tissue to inhibit phospholipase procedure has not been widely adopted. D activity (2), and a 5-stage extraction Most contaminants can be removed by process using chloroform and methanol washing the organic solvent extract with as well as acidic and basic solvents water or dilute salt solution, with the systems (9). A dry column method for proportions of solvents to aqueous the extraction of meat ground up with phase being important to prevent losses anhydrous sodium sulfate using of polar lipids (2, 12, 13). These columns containing celite followed by contaminants can also be removed by elution of the lipids from the column with passing the solvent through Sephadex a variety of solvents has been described G25, a procedure which is strongly (10). Extraction of lipids from lyophilized recommended by Nelson (16). He samples is known to result in difficulty of showed that the extraction of 20 mL of complete lipid extraction, as illustrated plasma yielded 83.2 mg of organic recently by poor recovery of TAG from solvent soluble material before passing oysters (11). The most common process through Sephadex where the yield was for animal and fish tissues, however, is only 50.1 mg of lipid. Since the total lipid to blend samples in ten volumes of content of foods is usually estimated methanol followed by the addition of 20 gravimetrically, this example highlights volumes of chloroform, with additional the importance of removal of non-lipid re-extraction of the sample (2, 5). This material. This is particularly important approach was developed by Folch et al. when the individual fatty acid content of (12) and Bligh and Dyer (13). The foods is based on total lipid value rather addition of antioxidants to the solvents is than using an internal standard to recommended, at a level of 50–100 quantitate the fatty acids (17). Following mg/L in the case of butylated this clean-up step, the solvent is hydroxytoluene. Use of glass containers removed in a rotary evaporator with the water bath set at or near room analysis of lean meats, such as lean temperature to protect the lipids which pork, chicken breast and certain cuts of are then transferred in chloroform to a beef (18). storage tube/flask. The total lipid content Supercritical CO2 can be used to is estimated by weighing an aliquot from extract lipids from foods in a continuous which the solvent has been evaporated process. The extraction apparatus can using a stream of nitrogen gas. The be connected to a gas chromatograph remaining lipids should then be stored at (GC) or super-critical fluid -20°C, prior to future analysis of lipid chromatography unit for the analysis of classes, fatty acids, and sterols. The the component lipids or for pesticide procedures described above are very analysis (5, 14). time consuming, labor intensive and Foods or feeds can also be analyzed require the use of large volumes of using infrared analysis, which is a solvents. nondestructive technique, using a single sample at a time and which provides Ŷ Automated Lipid Extraction Procedures data on total lipids, water and protein. The advantages of these automated There are a number of different techniques are speed of analysis and automated techniques which can be the use of less solvent or no solvent at used to analyze large numbers of food all. Disadvantages include the small samples which would be ideally suited sample size, the necessity to for the analysis of similar materials for standardize the infrared analyzer and quality control purposes. These include the CEM analyzer for each food type the soxhlet process either with or and the fact that it would be necessary without acid digestion using solvents to use a separate conventional such as diethyl either or chloroform: extraction technique for the analysis of methanol (6). This technique can allow fatty acids and sterols. samples to be analyzed in a batch process, and following extraction and Ŷ Sampling evaporation of the solvent, the weight of A major problem exists in relation to the total fatty acids (after acid treatment) determining a representative sample of or total lipids is determined food to be analyzed when the foods are gravimetrically. not homogeneous, as is the case with Another technique is to use a rapid fresh meat and fish. This problem is automated procedure to remove water, exacerbated if the technique adopted extraction of the dry residue with requires the use of only a small dichloromethane and re-weighing the analytical portion. For example, we have defatted residue which gives a figure for experienced difficulty in taking total lipids and percent water. This representative samples for the analysis technique, using an instrument known of meat using the CEM analyzer which as the CEM automated analyzer (CEM uses only 3–5 g meat/analysis. The fat Corp., Indiana Trail, NC), allows the content of these retail meat samples can analysis of one laboratory sample at a range from 4 to 50 per cent lipid (19) time and is rapid (about 5 min/sample), and the problem occurs at the stage however this process is not suited for where the lean meat and visible fat are further processing of the extracts due to mixed and blended together, and is loss of the lipid in the process. We have most evident in meat with a high lipid shown that this procedure does not content. It may be necessary to use a extract all the phospholipids from meat larger food sample to analyze foods and this can influence the total lipid which are not homogeneous. value significantly in the case of the Ŷ Analysis of Lipid Classes essential to purify the esters on a small Prior to the analysis of unfamiliar tissues column of fluorisil which removes sterols for fatty acid composition and sterols, it and oxidized material (2). The most is considered essential to examine the common internal standard is lipid extract by thin layer heptadecanoic acid (C17:0), although chromatography using a solvent system other odd-chain fatty acids have been to separate the non-polar lipids from the used including C13:0, C15:0, C19:0, phospholipids which remain at the C21:0 and C23:0. The choice depends origin. The lipids can be visualized using on the fatty acids found in the food and cupric acetate-orthophosphoric spray the separation of the internal standard (20), followed by heating in an oven at from the sample fatty acids. It is 100°C which leads to sterols showing a preferable to add the internal standard characteristic purplish color initially and as the TAG if this is the main lipid class, finally all the lipids char. This procedure rather than as a free fatty acid. There can assist in determining the likely has been some discussion on the identity of the lipid classes (e.g. choice of internal standards when the presence of wax esters in fish flesh) as food is rich in long-chain PUFA with well as the approximate proportions of Ackman (24) arguing for the use of a each class which is useful to estimate standard with a similar retention time amounts of internal standards to be (e.g. C23:0) as the fatty acids of used in the fatty acid and sterol interest. determinations. More sophisticated The separation of the FAME is best options include use of liquid achieved using polar capillary columns, chromatography (LC) (3, 5) or the with a wide range of columns of different Chromarod-Iatroscan technique to lengths, internal diameters and phase separate and quantitate lipid classes (5). thicknesses being commercially available. Columns designed for optimal Ŷ Determination of Fatty Acid separation of FAME are available (e.g. Composition and Content of Foods BPX70a, CPSil 88a, SP-2560a and For a food database, information is Omegawaxa) and the choice depends on required on chain length of saturates, the specific separations required. The cis/trans isomers of the separation of cis- and trans- monounsaturates, and Ȧ-6 and Ȧ-3 monounsaturated FAME as found in polyunsaturated fatty acids (PUFA) (21). margarines can be achieved There is also a developing interest in the successfully on 50–100m columns of positional distribution of fatty acids on BPX70 (25) (Figure 1) and SP2560a TAG molecule (22) and the molecular (26). Standards for determination of species of TAG present in the sample retention times are available (23). The general procedure for the commercially with a wide range of analysis of the fatty acids is to hydrolyze mixtures of known composition being the fatty acids, form derivatives and then available from Nu Chek Prep. Inc. analyze these by GC or LC with (Elysian, MN). Craske and Bannon (27), quantitation by use of internal amplifying the work of Ackman and standards. At the present time, GC is Sipos (28) demonstrated that, for a the method of choice for analysis of fatty number of common saturated and acids as methyl esters. There are many unsaturated fatty acids, the flame different methods for forming the fatty ionization detector theoretical response acid methyl esters (FAME), including the factors can be calculated from the use of BF3 in methanol, 5 per cent content of carbon bonded to hydrogen in anhydrous HCl in methanol, and 1–2 per the molecule. However, primary cent H2SO4 in methanol (2). It is usually standards are still essential to determine that both the gas chromatographic standard AOAC method (32), and parameters and the analyst's application of this method to the manipulative skills are efficiently analysis of eggs, meats and milk optimized. Assistance in the products. A disadvantage could be the confirmation of the identity of unknown very small analytical portion weight FAME can be gained by using non-polar which is likely to be problematic with capillary columns (e.g.. methyl silicone) heterogeneous foods (e.g. in the case of since the unsaturated FAME elute eggs, the sample weight amounts to before the saturated FAME using these only 200 mg). The compounds present phases (4). The use of GC mass in the non-saponifiable fraction of foods, spectrometry (MS) to confirm the apart from cholesterol, include plant identity of FAME is becoming more sterols, tocopherols and tocotrienols, common with the wider access to cholesterol oxides and other modestly priced instruments. The use of hydrocarbons such as squalene (30). picolinyl esters (13-OH methyl pyridine) A common approach to the of fatty acids for the GC-MS has been separation of sterols in foods is the use recommended for FAME with unusual of capillary GC with non-polar columns structures (29). and using a GC with a flame ionization detector. Internal standards of 5 - Ŷ Determination of Sterol Į Composition and Content in Foods cholestane or 5 Į-cholestanol have been used, however it is more difficult to The main sterol of interest from a separate cholesterol and 5 Į- nutritional viewpoint in foods is cholestanol, compared with 5 Į- cholesterol and it is now widely cholestane and cholesterol. Separation acknowledged that the earlier of the sterols as trimethyl silyl (TMS) spectrophotometric methods employed ether derivatives has been the preferred for the estimation of the cholesterol method since it is regarded that this content of foods led to an overestimate improves peak shape, decreases of the true amount present (30). The retention times and improves sensitivity, general approach to the analysis of food however problems associated with the sterols is to saponify the lipids, isolate use of TMS derivatives include the non-saponifiable fraction and hydrolysis and the fact that the reagents analyze the sterols by GC or LC using are toxic, flammable and corrosive (30). appropriate internal standards. With the development of inert fused Cholesterol can also be determined silica capillary GC columns, many enzymatically (31) using kit methods workers no longer derivatize sterol which are widely available. samples prior to GC analysis (30). LC In a review of the analysis of sterols has been used to separate food sterols, by GC or LC (30), details of the direct however, while cholesterol does not saponification methods for the analysis have a strong absorption in the UV of food sterols are discussed. The region, absorption at 205 nm can be advantages include reduced solvent used (30). volumes and sample preparation times, excellent recoveries compared with the Fig 1(a)

Fig 1 (b)

Figure 1. Gas chromatograms on 50 metre × 0.22mm BPX70 column of (a) margarine and (b) butter showing separation of the main trans 18:1 and cis 18:1 positional isomers (adapted from Mansour & Sinclair, 25) (6) Sahasrabudhe, M.R., & Smallbone, Ŷ The future B.W. (1983) J. Am. Oil Chem. Soc. Cardiovascular disease is still the 60, 801–805 leading cause of death in most (7) Lucas, C.C., & Ridout, J.H. (1970) industrialized societies. Recent research Prog. Chem. Fats 10, 1–150 into dietary aspects of cardiovascular (8) Morrison, W.R., Tan, S.L., & Hargin disease have indicated that it will no K.D. (1980) J. Sci. Food Agric. 31, longer be sufficient for food databases 329–340 to have information on the total fat, (9) Rouser, G., Kritchevsky, G., & cholesterol and major fatty acids types Yamamoto, A. (1967) in Lipid of different foods. Because of the Chromatographic Analysis, Vol 1, increasing interest in oxidation of low G.V. Marinetti (Ed.), Edward Arnold density lipoproteins (33), oxidation of Ltd., London, pp.99–162 cholesterol in food (34, 35), the role of (10) Marmer, W.N., & Maxwell, R.J. trans fatty acids in lipoprotein (1981) Lipids 16, 365–371 metabolism (36) and the beneficial (11) Dunstan, G.A., Volkman, J.K., & effects of the Ȧ-3 PUFA on various Barrett, S.M. (1993) Lipids 28, 937– aspects of cell metabolism (37), in the 944 future food databases will require (12) Folch, J., Lees, M., & Sloane- detailed information on the tocopherol Stanley, G.H.S. (1957) J. Biol. and tocotrienol isomers, cis and trans Chem. 226, 497–509 fatty acid isomers, cholesterol oxides (13) Bligh, E.G., & Dyer, W. (1959) Can. and other oxidation products of food J. Biochem. Physiol. 37, 911– 917 lipids, and molecular species and (14) King, J.W. (1993) INFORM (J. Am. positional distribution of fatty acids in Oil Chem. Soc.) 4 1089–1098 TAG (21–23). Another issue which is (15) Phillips, F.C., & Privett, O.S. (1979) likely to emerge as a challenge for Lipids 14, 949–952 analysts in the near future will be the (16) Nelson, G.J. (1993) in Analysis of development of standard methods for Fats, Oils & Derivatives, E.G. the wide range of different fat Perkins (Ed.), AOCS Press, substitutes which have been developed Champaign, IL, pp. 20–59 in recent years (38). (17) Weirauch, J.L., Posati, L.P., Ŷ References Anderson, B.A., & Exler, J. (1977) J. Am. Oil Chem. Soc. 54, 36–40 (1) Vlieg, P., Body, D.R., & Burlingame, (18) Mann, N.J., Sinclair, A.J., Watson, B. (1991) J.N.Z. Diet. Assoc. 45, M.J., & O'Dea, K. (1991) Food Aust. 29–30 43, 67–69 (2) Christie, W.W. (1982) Lipid (19) Watson, M.J., Mann, N.J., Sinclair, Analysis, 2nd Ed., Pergamon Press, A.J., & O'Dea, K. (1992) Food Aust. New York, NY 44, 511–514 (3) Christie, W.W. (1987) High (20) Fewster, M.E., Burns, B.J., & Mead, Performance Liquid J.F. (1969) J. Chromat. 43, 120– Chromatography and Lipids, 126 Pergamon Press, New York, NY (21) Sinclair, A.J. (1993) Food Aust. 45, (4) Christie, W.W. (1989) Gas 226–231 Chromatography and Lipids, The (22) Kritchevsky, D. (1988) Nutr. Rev. Oily Press, Ayr, Scotland 46, 177–181 (5) Analysis of Fats, Oils & Derivatives (23) Currie, G.J., & Kallio, H. (1993) (1993) E.G. Perkins (Ed.), AOCS Lipids 28, 217–222 Press, Champaign, IL (24) Ackman, R.G. (1991) Lipids 27, 858–862 (25) Mansour, M.P., & Sinclair, A.J. (32) Official Methods of Analysis (1980) (1993) Asia Pacific J. Clin. Nutr. 3, 13th Ed., AOAC, Washington, DC, 155–163 sec. 43.235 (26) Firestone, F., & Sheppard, A. (33) Steinberg, D., Parthasarathy, S., (1992) in Advances in Lipid Carew, T.E., Khoo, J.C., & Witztum, Methodology, Vol. 1, W.W. Christie J.L. (1989) New Eng. J. Med. 320, (Ed.), The Oily Press, Ayr, 915–924 Scotland, pp. 273–322 (34) Chisolm, G.M. (1991) Current (27) Craske, J.D., & Bannon, C.D. Opinion in Lipidology. 2, 311–316 (1987) J. Am. Oil Chem. Soc. 64, (35) Sarantinos, J., O'Dea, K., & Sinclair, 1413–1417 A.J. (1993) Food Aust. 45, 485–490 (28) Ackman, R.G., & Sipos, J.C. (1964) (36) Mensink, R.P., Zock, P.L., Katan, J. Am. Oil Chem. Soc. 41, 377–383 M.B. & Hornstra, G. (1992) J. Lipid (29) Harvey, D.J. (1992) in Advances in Res. 33, 1493–1501 Lipid Methodology, Vol. 1, W.W. (37) Sinclair, A.J. (1993) Asean Food J. Christie (Ed.), The Oily Press, Ayr, 8, 3–13 Scotland, pp. 19–80 (38) (38) Haumann, B.F. (1993) (30) Fenton, M. (1992) J. Chromat. 624, INFORM (J. Am. Oil Chem. Soc.) 4, 369–388 1227–1235 (31) Shen, C.J., Chen I.S., & Sheppard, A.J. (1982) J. Assoc. Off. Anal. Chem. 65, 1222–1224 Methods and Conventions of Nutrient Analysis

Conventions for the Expression of Analytical Data

David A. T. Southgate

formerly of AFRC Institute of Food Research, Norwich Laboratory, Colney, Norwich, NR4 7LY, UK

The modes of expression and the conventions used in citing analytical data in nurtritional databases are critical for the accurate use of the database. This is especially true for the nutritionist who wishes to use data from a number of different databases or who wishes to merge data from databases in different countries for a specific study. While international agreement on the modes of expressions and conventions is the preferred approach, at the very least it is essential that the documentation of all databases describes in detail the modes of expression and conventions explicitly so that users will know where data are compatible. The major conventions in use are discussed and proposals made for establishing some common positions in the preparation of nutritional databases.

he nomenclature, conventions and modes of expression used to describe the nutrient values have a profound influence on the accurate use of nutritional Tdatabases. They are especially important when one is working using a series of different databases. Many nutritional epidemiological studies are being made internationally. For such studies it is often necessary to construct a special database from a number of different sources, and it becomes critically important to ensure that compatible data are being combined. This involves consideration of, first, the compatibility of the analytical methods used to generate the data (1) and, second, that the modes of expression, and for many nutrient values, the conventions used in deriving the values, are also compatible. Where the analytical methods for a factors applied to the amounts of nutrient are specific and the range of protein, fat, carbohydrates and alcohol methods in use is known to produce in the food. comparable values then expression on The energy conversion factors in use an appropriate weight basis is straight- in most databases are based on the forward (2). studies made by Atwater and his There are three major categories of colleagues in the early years of this nutrients where the conventions and century and it is a tribute to their modes of expression are a major cause experimental skills that the system, of incompatibility. First, where the despite the assumptions which had to nutrient values are calculated from be made at that time, remains a some other analytical measurement. practical approximation (3). The Atwater Second, where the nutrient is a complex factors adjust the protein, fat and mixture and some analytical carbohydrate heats of combustion to compromises have been adopted in the allow for the fecal losses of these measurement of the mixture. The third constituents by multiplying the average category includes those vitamins for heats of combustion for mixed proteins, which there are a number of active fats and carbohydrates, by the forms (vitamers) which differ in respective apparent digestibilities of the biological activity. three components. The energy loss in urine was allowed for by correcting the Ŷ Nutrient Values Calculated from Other Analytical Measurements protein value for the energy per unit nitrogen lost in urine. These cover two of the most important These calculations give the familiar, conventions used in the expression of 4, 9, 4 factors (kcal per g of protein, fat, nutrient values in databases; the and carbohydrate, respectively), which expression of values for energy and are still widely used. protein. There are two other conventions Energy Values of Foods used in nutritional databases. Specific Energy Conversion Factors. In nutritional databases (and in general This was originally developed by nutritional usage) the energy values of Atwater but not used by him. It was later foods are, strictly, the “metabolizable advocated by Merrill and Watt (4). This energy values” in other words, the convention is based on the premise that energy that is available for use by the it would be better to use heats of body. In formal terms metabolizable combustion values specific for the energy of the dietary intake is equal to different foods (or food groups) and to the gross (the heat of combustion) use specific apparent digestibility values energy intake minus the energy lost in for the constituents of each food (or feces and urine (other losses should group), rather than to use average also be included, for example, energy values derived for mixed diets. This lost in other secretions or gases but in system depends on having experimental human nutrition these are customarily data for the heats of combustion for all discounted). Metabolizable energy components of foods and extensive data values thus are, in strict terms, an from human metabolic studies, and at attribute of the dietary intake and the the present time relies heavily on early calculation of values for foods is a data, some from Atwater himself. It is pragmatic approximation. important that databases using this Metabolizable values for foods are system document, precisely, which calculated using energy conversion factors have been used for specific foods as the published account can be suggested (7) that a radically different interpreted in different ways system based on measured heats of Modified Atwater System Used in the combustion of foods would be UK Nutritional Database. In the United intellectually more satisfying but in Kingdom a different system was practical terms the differences produced developed by McCance and Widdowson by the three systems are much smaller (5). The need for this arose because than the errors inherent in ignoring these authors measured carbohydrates individual differences in digestibility and directly (as opposed to Atwater's use of the errors in measuring food intakes. the “by difference” method) and Provided that it is recognized that these furthermore, they divided the food conventions are approximations and carbohydrates into two categories; energy values are not cited to four “available” the digestible sugars and significant figures the practical starches, which were glucogenic in man incompatibilities are not significant (2). and “unavailable”, those carbohydrates Protein not digested in the small intestine and not providing the body with absorbable Virtually all nutritional databases give carbohydrates. For the available values for protein that are derived by carbohydrates they assigned the energy calculation from measured total nitrogen conversion factor 3.75 kcal per g (this is values. These have been customarily the heat of combustion of measured by the Kjeldahl method. monosaccharides) because available These calculated values should be carbohydrates were expressed in this called “crude protein” and recognized as way. Unavailable carbohydrates were a approximations and not as estimates assigned a zero energy value, not of protein in a biochemical sense. The because they did not provide energy but factors used for calculating protein are because these components were known based on the percentage of nitrogen in a to reduce the apparent digestibility of typical proteins. Thus 6.25 is proteins and fats and therefore any appropriate for proteins with 16 g N per energy from the short chain fatty acids 100 g protein. produced by fermentation in the large The FAO expert report on protein bowel was discounted. requirements in 1973 (8) recommended Detailed evaluation of this approach that different factors should be used for in experimental studies on a large calculating “crude” protein values for number of subjects showed that it gave different food groups to take account of a good prediction of metabolizable the N-content of different food proteins energy intakes (6). The UK system and this approach has been followed in gives lower energy values for plant the UK nutritional database (5, 13). foods that are rich in plant cell wall When the protein and energy material (and incidentally organic acids). requirements were reviewed again in Comparisons of the Three 1985 (9), it was clear that the use of Conventions. In practical terms the three these factors produced anomalies when conventions give similar values. The making recommendations for protein Atwater system tends to give over- intakes. This arises because all the estimates of metabolizable energy and experimental studies on protein the UK system under-estimates energy requirements have been based on the intakes at very high unavailable measurement of nitrogen metabolism carbohydrate intakes because of the and it has been argued that it would be increasing importance of the energy better to avoid confusion by using only from the fermentation products of the 6.25 as a conversion factor. unavailable carbohydrates. It has been In nutritional terms it would be should be given to values obtained preferable to move to more using methods that give complete biochemically coherent measures where extraction such as those used to protein values were based on estimates prepare extracts for fatty acid analysis of amino acid nitrogen content using (2) with all values documented by amino acid composition data. The method. pratice of correcting for non-protein, Carbohydrates non-amino acid nitrogen is desirable for some foods especially cartilaginous The main difficulties arise from the fishes and some fungi in order to avoid incompatibility between carbohydrates over-estimating the protein in these values obtained “by difference” and by foods. direct analysis. Direct analysis provides As a general principle it is desirable values for the different classes of that all databases include total nitrogen sugars, mono-, di- and oligo- values so that when merging data a saccharides starch and non-starch consistent approach to calculating crude polysaccharides (the major components protein values can be adopted. It is of dietary fiber) (10). The sum of the important to ensure that users do not individual components approximates to equate (N × a factor) with a biochemical the total carbohydrate “by difference” if concept of protein as functional the food does not contain substantial polymers of amino acids and to amounts of non-carbohydrate recognize the conventional components. It is important to recognize approximation. that “by difference” values also include errors in the measurement of water Ŷ Complex Nutrients Involving content, ash, protein and total lipid. Analytical Compromises In the UK database the available The nutrients which fall into this carbohydrates are expressed as category include fat, the carbohydrates monosaccharides, which means that the and the folates. The essential difficulties summation of the values will exceed the arise because these nutrients are very total carbohydrate “by difference” by a complex mixtures the determination of significant amount especially in starch- which requires time-consuming rich foods because of the addition of separations which are not widely used in water on hydrolysis of disaccharides routine food analysis. This means that and polysaccharides. most of the data available to database The ideal from the analytical point of compilers is based on simple methods view is to move away from the and the values obtained tend to be technically obsolete “by difference” method-dependent. method; but at the present time it is Fat essential to identify the method used when giving carbohydrate values. This The values for fat in most databases are is especially true for dietary fiber where for “total lipid-solvent soluble material”, the methods in use measure which includes the triacylglycerol conceptually different fractions. Thus compounds, sterols and depending on dietary fiber values should be cited the method used, phospholipds. Several (measured as total dietary fiber, AOAC, of the methods in common use do not or as non-starch polysaccharides etc.). extract the lipids efficiently from some food matrices and the values cited in Folates databases may be incompatible A large number of folate vitamers exist because of this method-dependence. In in foods. The biological availability of the compilation of databases preference them is an active topic of research and there are chromatographic methods for efficiency of conversion to retinol. These the separation of the different forms, conversion factors are based on studies although these are technically exacting with mixed diets and there is growing and not suitable for routine use at the evidence that the efficiency of present time. Most values in nutritional conversion is profoundly affected by the databases are based on microbiological amount of fat in the diet and on the determinations using either structure of the food matrix, so that Streptococcus faecalis or Lactobacillus under some conditions the carotenoids caseii (rhamnosus). Deconjugation of may not contribute to vitamin A status. the polyglutamyl chain has to be used to The convention is to divide ȕ-carotene measure total folates. At one time it was values by 6 and other carotenoids by 12 assumed that a value measured before to convert them to retinol equivalents. deconjugation represented free Some authors argue that the conversion monoglutamyl folates but is now clear efficiency of the ȕ-carotene in dairy that this value includes contributions foods is under-estimated by the use of 6 from the polyglutamyl forms and has no and suggest 3 but the evidence for this value for the nutritional classification of is not totally convincing (12). the folates. S. faecalis does not respond At the present time there is growing to some folates and the values obtained evidence that the carotenoids are are significantly lower than when using important nutrients in addition to their L. rhamnosus. It is thus critically role as provitamin A compounds and important to identify the organism used this suggests that databases should in the assay when citing folate values firstly, give values for retinol and the (11). different carotenoids separately. If a total vitamin A activity is deemed useful Ŷ Assigning Biological Activity to Different Vitamers in addition, the convention used must be described. Such an approach permits This concerns those vitamins where users to assign their own convention there are a range of vitamers in foods and to use different conventions as which have differing biological activities understanding of the factors determining and where it is customary in nutritional the efficiency of conversion improves. databases to give some kind of aggregated value for the vitamin activity Vitamin E Activity of the food. The major vitamin which The biological activity of the tocopherols falls into this category is vitamin A, and tocotrienols vitamers is related in where it has become accepted to give a this convention to the activity of Į- vitamin A value expressed as tocopherol. The biological evidence for micrograms retinol equivalents, based calculating the biological activity factors on summation of the amounts of retinol is limited but the approach is being used and the activity of different carotenoids. in the UK database (13). An analogous convention has been Once again it is desirable that the proposed for vitamin E where the databases contain the values for the aggregated Į-tocopherol equivalents are different forms so that users can apply being estimated by summing the their own consistent conversion weighted values of the different conventions. Should the compilers wish, tocopherols and tocotrienols. in addition, to give a total vitamin E Vitamin A Activity activity value in the database then the convention must be documented. The accepted convention is to take the values for the provitamin carotenoids Ŷ Development of Agreed and divide by factors which reflect the Conventions and Modes of Expression The interchange systems developed by per day so in practical terms great errors INFOODS (14) allow for the will not flow from combining these two documentation of data values within the types of values. exchange tags which incorporate units In the case of the vitamins the and details of method where values are biological equivalents will undoubtedly method-dependent. In the context of this change as more evidence emerges and paper “method-dependence” is it is essential to accumulate and give the restricted to those instances where actual analytical values for the different different methods give different values, vitamers in databases to permit not to differences arising because of reappraisal in the future. differences in the performance of the Finally documentation as part of the method. Values that have been database should be seen as the ideal. calculated by conventions, such as We really should be aiming for energy and protein are similarly imbedded information (15) so that the assigned different tags where this is user can easily establish what a nutrient appropriate. One has therefore within name implies and where analytical the INFOODS interchange scheme a methods may make the values method- system that should prevent the dependent. aggregation of incompatible data. The Conventions should be recognized as primary need at present for those such and the values they provide involved in merging data from recognized as approximations of databases where different conventions biological phenomena and treated as have been used is to resolve the one would treat all derived values, with differences so that they can bring the caution. values into a consistent compatible Ŷ References form. This depends on having access to complete documentation of all (1) Southgate, D.A.T. (1985) Ann. Nutr. databases, and at the primary data Metab. 29, Suppl., 49–53 source level, avoiding any tendency to (2) Greenfield, H., & Southgate, D.A.T. aggregate original analytical data. (1992) Food Composition Data: Thus if a database has values for Production, Management and Use, protein, fat and carbohydrate (however Elsevier Applied Science, London expressed) it is possible to recalculate (3) Allison, R.G., & Senti, F.R. (1983) A energy values de novo to avoid Perspective on the Application of incompatibilities. Similarly total nitrogen the Atwater System of Food Energy values give the possibility of calculating Assessment, Federation of protein values using either one or American Societies for several factors. Experimental Biology, Bethesda, At the present time some MD carbohydrate values are incompatible (4) Merrill, A.L., & Watt, B.K. (1955) without more detailed analytical Energy Value of Foods, US Dept of information than is currently available. Agriculture, Handbook 74, USDA, This is especially true for dietary fiber Washington, DC values because the crude fiber values (5) Paul, A.A., & Southgate, D.A.T still given in some databases are (1978) McCance and Widdowson's incompatible with dietary fiber values. In The Composition of Foods, 4th Ed., principle TDF values should be higher HMSO, London than NSP values but taken across the (6) Southgate, D.A.T., & Durnin, whole range of plant foods resistant J.V.G.A. (1970) Br. J. Nutr. 24, starch may be of the order of 3–6 g per 517–535. day and lignin intakes rarely exceed 1 g (7) Livesey, G. (1991) Proc. Nutr. Soc. (13) Holland, B., Welch, A.A., Unwin, Aust. 16, 79–87 I.D., Buss, D.H., Paul, A.A., & (8) FAO/WHO Expert Group (1973) Southgate, D.A.T. (1991) McCance FAO Nutrition Series, No. 52, FAO, and Widdowson's The Composition Rome of Food, 5th Ed., Royal Society of (9) FAO/WHO/UNU (1985). WHO Chemistry, Cambridge Technical Report Series 724, WHO, (14) Klensin, J.C., Feskanich, D., Lin, V., Geneva Truswell, A.S., & Southgate, D.A.T. (10) Southgate, D.A.T (1991) (1989) Identification of Food Determination of Food Components for INFOODS Data Carbohydrates, 2nd Ed., Elsevier Interchange, UNU Press, Tokyo Applied Science, London (15) Southgate, D.A.T. (1992) in The (11) Finglas, P.M., Faure, U., & Contribution of Nutrition to Human Southgate, D.A.T. (1993) Food and Animal Health, E.M. Chem. 46, 199–213 Widdowson & J.C. Mathers, (Eds.), (12) WHO (1967) WHO Technical Cambridge University Press, Report Series 362, WHO, Geneva Cambridge, pp. 369–378 Section III

Quality Control of Food Composition Data and Databases

his Session was chaired by Dr Dorothy Mackerras of the Department of Public Health, Sydney University. The keynote address entitled Food Classification and TTerminology Systems was given by J.A.T. Pennington. This was followed by papers on Nutritional Metrology: the Role of Reference Materials in Improving Quality of Analytical Measurement and Data on Food Components by J.T. Tanner, W.R. Wolf and W. Horwitz, Strategies for Sampling: the Assurance of Representative Values by J. Holden and C.S. Davis, and Assuring Regional Data Quality in the Food Composition Program in China by G. Wang and X. Li, and Quality Control for Food Composition Data in Journals — a Primer jointly presented by K.K. Stewart and M. R. Stewart. These papers are published on the following pages.

The paper by B. Perloff and S. Gebhardt, Building Data Quality in the Data Base Management Process, is not included. The authors can be contacted at the US Department of Agriculture, 4700 River Road, Riverdale, MD 20737, USA. Posters presented after Session III were:

ƒCriteria Used for Analytical Data Evaluation, Buick, D., Mottershead, R., & Scheelings, P., Australian Government Analytical Laboratory, Seaton, SA, Australia.

ƒEvaluation of Foods as Analytical Control Samples, Buick, D., Pant, I, Trenerry, C., & Scheelings, P., Australian Government Analytical Laboratory, Seaton, SA, Australia

ƒDevelopment of an In-house Nutrition and Food Science Bibliographic Database Using Micro CDS/ISIS, Chia, W.Y., & Greenfield, H., Department of Food Science and Technology, University of New South Wales, Sydney NSW, Australia.

ƒAPINMAP - an Integrated Database of Medicinal and Aromatic Plants, Henninger, M., School of Information, Library and Archive Studies, University of New South Wales, Sydney, NSW, Australia.

ƒFood Analysis Reference Materials for the Asia-Pacific, James, K.W., DSTO, Materials Research Laboratory, Scottsdale, TAS, Australia.

ƒInternational Survey on Dietary Fiber Definition, Analysis and Reference Materials, Lee, S.C., & Prosky, L., Kellogg Company, Battle Creek, MI 49016, and US FDA, Washington, DC 20204, USA. ƒDesktop Publishing of Food Tables, Mikkelsen, B.E., Danish Catering Centre, Institute of Food Chemistry and Nutrition, National Food Agency, Søborg, Denmark.

ƒInformation Sources in Nutrition and Food Science and Technology, Mobbs, S.L., & Siu, C.S., Biomedical Library, University of New South, Sydney NSW, Australia.

ƒInterface Standard for Food Databases, Pennington, J.A.T., Hendricks, T.C., Douglass, J., Peterson, B., & Kidwell, J., Center for Food Safety and Applied Nutrition, US FDA, Washington, DC 20204, USA.

ƒDevelopment of ASEANFOODS Reference Materials, Pustawien, P., & Sungpuag, P., Institute of Nutrition, Mahidol University, PO Box 31, Talingchan, Bangkok 10170, Thailand. Quality Control of Food Composition Data and Databases

Food Classification and Terminology Systems

Jean A.T. Pennington

Food and Drug Administration, 200 C Street, S.W., Washington, DC 20204, USA

Food classification systems organize foods in databases among groups and subgroups based on food type (e.g., grain products, fruits) and/or food use (e.g., beverages, main dishes). The food groups and subgroups vary among databases according to the number and types of foods in the database, the cultural uses of the foods, and specific decisions made by the database compiler. Terminology systems are structured methods of applying descriptive terms (e.g., terms relating to packaging, processing, color, maturity) to foods. Faceted terminology systems assign descriptive terms for specific characteristics of foods, allowing these characteristics to be considered independently. Eurocode is a food classification, coding, and terminology system. Langual/Interface Standard is a faceted food description system with standardized vocabulary, and the INFOODS system is a free-text faceted food description system.

he words used to classify, name, and describe foods are a mixture of traditional, fanciful, technical, and sensory terms. Sometimes these terms convey a clear Tpicture of what a food is, especially if one is already familiar with it. For a food that is not familiar, the mental image conveyed by the terms is important to understanding the food and the data associated with it. Food names and the terms associated with them are key to the use of information in food-related databases. There should be sufficient descriptive information about the food to clearly understand what the data represent. Table I. Food classification systems (number of groups in each database) References (1) (2) (3) (4) (5) (6) (7) (8) (9) Food Type Classifications Milk and eggs 1 Milk and milk products 1 1 1 1 1 1 1 1 Eggs 1 1 1 1 1 1 1 1 Meat, poultry, fish 1 1 Meat and poultry 1 1 1 1 1 1 Meat 3 Poultry 1 Luncheon meat & sausages 1 Fish and shellfish 1 1 1 1 1 1 1 Fats and oils 1 1 1 1 1 1 1 1 Grain products 1 1 1 1 1 3 1 1 1 Fruits and vegetables 1 Fruits 1 1 1 1 1 1 1 Fruits and nuts 1 Fruit juices/nectars 1 Legumes, nuts, seeds 1 1 Nuts and seeds 1 1 1 1 1 Vegetables 1 1 1 1 1 1 1 1 Legumes 1 1 Potatoes and roots 1 1 Food Use Classifications Beverages 1 3 1 1 1 1 1 1 Alcoholic beverages 1 1 Sugars/syrups/sweets 1 2 1 1 1 1 1 Special nutritional use 1 Herbs/spices/flavourings 1 1 1 1 Snacks 1 1 Soups/sauces/gravies/dressing 1 1 1 Fast foods 1 Baby food 1 Prepared products 1 Miscellancous/other 1 3 1 1 7 1 Number of major groups 13 19 10 12 14 21 10 29 14 References (1) Eurocode 2 (2) Germany (3) Sweden (4) Australia (5) Britain (6) USDA Agriculture Handbooks (7) USDA Nationwide Food Consumption Survey (8) Langual (9) Near East composition databases, except those Ŷ Classification Systems arranged alphabetically, are organized Classification systems refer to the by such groupings. These groupings groupings and subgroupings of foods in assist database users in locating foods databases, based on food type (e.g., and comparing the nutrient content of vegetables, dairy products) and/or food similar products. They also reduce the use (e.g., beverages, fast foods, repetition of group and subgroup snacks) (Table I). Most food headings. The number of major food groups found in nine selected databases equivalent to British “biscuits”) could be (1–9) ranges from 10 to 21 (Table I). classified under “grain products” or Foods in the major groups are “desserts”; “bouillon” (American term usually subgrouped by more precise equivalent to British “beef tea”) could be food names or by descriptive terms, classified under “soups” or “beverages.” creating hierarchies within each major Such products may be forced into one group. For example, meats may be group or listed in several. The latter subgrouped by beef, lamb, and pork, solution would result in foods with the and desserts may be subgrouped by same name being in different groups, cakes, cookies, and pies. Beef may be e.g., some French fries under the fast further subgrouped by specific cuts food group and some under the (brisket, loin, steak), grade (choice, vegetable group. This makes it difficult good, prime), and/or fat trim (0", ¼“, ½”). for users to locate similar foods and Cookies may be further subgrouped by compare their nutrient content. flavor ingredients (chocolate chip, Within each major food group, similar oatmeal, peanut butter, sugar) or source decisions about how to place foods (commercial, homemade). among subgroups must be made, As Table I shows, food type especially if there are rigid hierarchies. classifications vary somewhat among Some foods clearly fit two or more countries. For example, some subgroups. For example, “Irish coffee” databases group all vegetables (American name for coffee with together. Others have separate whiskey) is clearly a “beverage,” but is groupings for legumes and root both an “alcoholic beverage” and a vegetables; some group legumes and “coffee beverage”. Other foods seem to nuts together. However, there is be transitions between food groups or probably better international agreement subgroups. For example, a broth with for food type than for food use chunks of meat and vegetables may be groupings because the use of foods in a transition between a soup and a stew. daily diets varies among ethnic and In hard-copy databases with space cultural groups. Food use categories are constraints, subgroups are usually particularly useful to group together formed by identifying common products with common dietary use that descriptive terms and using them as could be “lost” among food type subgroup headings. This may lead to classifications. For example, under some inconsistences in a database as “snack foods” in the USDA Agriculture to how subgroups are formed. For Handbook No. 8–19 (6), one finds example, pancakes and waffles could vegetable-based products (potato be subgrouped under grain products chips/crisps), grain-based products first by type (frozen, frozen batter, (corn chips, tortilla chips, popcorn), and home-made, liquid batter) and then by fruit or nut-based products (trail mix, flavor (blueberry, cinnamon, plain, old banana chips). fashioned, strawberry, whole grain) or Food Grouping Problems the other way around. A fast-food fish sandwich might be subgrouped by For databases with both food type and entree type “fish,” by entrée type food use classifications, there may be “sandwiches,” or by restaurant name some difficulty in placing foods that fit (e.g., “McDonald's”). under two or more groups. For example, Decisions about groups and “French fries” (an American food subgroupings are usually made by the equivalent to British “chips”) could be database compiler after the data are classified under “vegetables” or “fast collected and sorted. At that point, the foods”; “cookies” (American food number of repetitive terms can be determined and minimized. The food name, but are linked to the food provision of an index assists users in name in a manual or computerized locating foods that might be placed in system. Faceted systems allow for multiple groups or subgroups or that different characteristics of food to be have inconsistent subgroup structure considered independently. To develop from group to group. such a system, one must identify the In computerized databases, one facets, collect the descriptive terms might view only one food name (and its belonging to each facet, and define the descriptors) at a time without benefit of terms. seeing the other foods and descriptors Faceted systems for foods are based in the classification hierarchy. It is largely on the faceted system developed necessary to repeat descriptive terms in in 1971 by the International Network of this case. For example, the terms Feed Information Centers (INFIC) for “breakfast cereal,” “cookie,” and “frozen international exchange and dinner” would need to be repeated with dissemination of information about feeds each listing for which it is appropriate. (10). Approximately 21,000 feeds have been described according to the facets: Ŷ Terminology Systems origin, part, process, growth stage, cut, Terminology systems refer to the and grade. The descriptive information systematic methods of applying (in English, French, and German) and descriptive terms to foods. These terms, numerical data associated with various which provide information about color, feeds can be stored, summarized, flavor, maturity, preparation, retrieved, and printed in various formats. preservation, brand names, etc., are Three unique terminology systems important because the nutrient contents are briefly discussed: Eurocode 2 (1), a of foods vary according to such terms. food classification/coding/terminology For example, the USDA database for system; Langual/Interface Standard (8, the 1987–88 NFCS (7) lists 18 entries 11, 12), a faceted description system for string beans, each of which has with standardized vocabulary; and the different nutrient values based on INFOODS system (13), a free-text descriptive terms for color, preservation, faceted description system. cooking, and/or added ingredients. Descriptive terms also provide insights Eurocode 2 about food safety (storage, Eurocode was originally developed in preservation) and nutritional quality the early 1980s as a common European (fortification, processing). system for coding foods consumed by The simplest type of terminology participants in dietary surveys (14, 15). system is one which orders the In this case, “coding” refers to the descriptive terms (as appropriate) assignment of alphanumeric codes to around the food name (linear foods in databases. The codes link the descriptors). Descriptors for most food food name to the data (e.g., composition names could be ordered in several or consumption data) associated with it ways. Database compilers generally try and allow for computer manipulation of to use consistent terms and ordering of the data. The Eurocode 2 manual (1) linear descriptors to facilitate the use of provides rules for coding single foods, the database. mixed foods, and foods as recipes. The Faceted terminology systems assign codes (as described in the manual) may descriptive terms for each food for to be applied to foods in manual or specific characteristics (facets). The computerized databases. terms are not necessarily a part of the Table II. The Eurocode systema Eurocode Fields with Examples Field 1. Field 2. Field 3. Field 4 Main group. Subgroup. Food name. Recipe (optional)

Meat and meat products (3) Mutton (3.4) Mutton, carcass meat (3.4.1) Mutton recipe prepared in Ireland (31E.4.1.2)b

Grains and grain products (6) Wheat breads (6X.1)c Rusks (6X.1.8)

Vegetables and products (8) Cabbages (8.2) Kohlrabi (8.2.6)

Eurocode Descriptors Examples T Thermal treatment at consumption T7 (deep fried) N Non-thermal treatment N4 (mashed) P Preservation method/packing medium P19 (frozen) A Component added A10 (fiber added) R Component removed R4 (skin removed) a Information adapted from Poortvliet and Kohlmeier (1) b IE indicates a recipe prepared in Ireland. The “2” in the fourth field indicates a specific recipe for a dish based on mutton, e.g., Irish stew c The X in the first field indicates that this food has been coded as a mixed food

The food codes have four fields mutton, 3.4.1 is for mutton carcass (Table II). The first field identifies one of meat, and the 2 at the end refers to the 13 main food groups, the second field recipe). identifies the food subgroup, the third Eurocode 2 provides an optional field identifies the food item, and the terminology system with descriptive fourth field, which is optional, provides terms for five facets: thermal treatment, reference to a recipe. For example, in nonthermal treatment, preservation and the code for rusks, 6×.1.8, 6 represents packing, components added, and grain products and 6×.1 represents components removed. Descriptors are wheat breads. The “X” in the first field identified with alphanumeric codes (e.g., indicates that a food is coded as a T7 for the thermal treatment “deep mixed food (i.e., a multi-ingredient food). fried”), and definitions are provided for A two-character country code replaces consistent coding. The authors of the the “X” to identify the country for a Eurocode 2 manual indicate that the national recipe. For example, 3IE.4.1.2 descriptors are designed for dietary is the code for a mutton recipe prepared surveys and do not attempt to satisfy the in Ireland (3 is for meat and meat degree of technical detail used in food products, IE if for Ireland, 3.4 is for technology (1). Table III. Langual factors and examples of factor terms Langual Factors Examples of Factor Terms 1. Product type Breakfast cereal 2. Food source (plant or animal) Leafy vegetable 3. Part of plant or animal Organ meat 4. Physical state, shape or form Semisolid 5. Extent of heat treatment Partially heat-treated 6. Cooking method Cooked by dry heat 7. Treatment applied Hydrogenated 8. Preservation method Pasteurized by heat 9. Packing medium Packed in gelatin 10. Container or wrapping Paperboard container 11. Food contact surface Plastic 12. Consumer group/dietary use Human food, low calorie 13. Geographic places and regions a. Area of origin (grown/produced) Zimbabwe b. Area of processing Italy c. Area of consumption Tennessee 14. Cuisine Chinese 15. Adjunct characteristics of food (examples) Color of poultry meat Dark meat Grade of meat, US Choice grade Plant maturity Ripe or mature Location of preparation Restaurant/fast food prepared

Langual assigned alphanumeric codes. Langual is currently used on a mainframe Langual is a faceted food description computer at FDA, but has been adapted language that has been under for personal computers in other development by the US Food and Drug locations. Administration (FDA) since the early Foods in various databases may be 1970s (8). It is a software system that searched or retrieved by one or more of may be applied to food-related the Langual descriptive terms. The more databases such as those of food accurate the descriptions of the foods, composition and food consumption. the more informative are the searches Each food is assigned a set of and retrievals. To facilitate retrieval and descriptors, using standardized aggregation, the descriptors within each language, from the following facets: facet are arrayed in a hierarchy from product type; food source; part of plant broader to narrower terms. The or animal; physical state, shape, or vocabulary includes definitions for the form; extent of heat treatment; cooking terms and explains when and in what method; treatment applied; preservation contexts they should be used. The method; packing medium; container or Langual thesaurus includes cross wrapping; food contact surface; references for synonyms and Latin consumer group/dietary use; names, and for preferred, broader, geographical places and regions; narrower, and related terms. cuisine; and adjunct characteristics An European Langual Working Group (Table III). If the factor term for a food is was established in the early 1990s to be not known or does not apply for a food, the focal point for Langual use in Europe the terms “unknown” and “not and to communicate needs to the US applicable” may be used. For internal Langual Committee. In May 1992, storage and processing, factor terms are Langual was evaluated for use in be used. Once the foods in a database European databases. Several European are described according to the interface, dietitians/nutritionists were trained in databases may be queried and Langual and were asked to code a information may be retrieved. The number of foods to determine the system will also allow for matching (or applicability of Langual to European finding the closest matches) of foods in foods. The results of this test indicated different databases. that Langual is an appropriate INFOODS terminology system for European foods (16). The International Network for Food Data The concept of an interface standard Systems (INFOODS) was organized to (a common communication link based improve the quality and accessibility of on the food name and descriptors) to food composition databases. It was allow international exchange of food- funded by US government agencies related data arose at a meeting of the from 1984 to 1987 with headquarters at Committee on Data for Science and the Massachusetts Institute of Technology (CODATA) in March 1990 in Technology. The Food Nomenclature Maryland, USA. Criteria for an and Terminology Committee (one of the international interface standard were three INFOODS committees) was drafted at this meeting, and FDA used charged with developing a proposal to those draft ideas to formulate an standardize the nomenclature and interface using Langual (11). The description of foods to allow for useful interface was further refined under a exchange of food composition data FDA contractual effort (Figure 1) (12), among countries (13). and the computer software for the The Committee met at several interface standard is expected to be international meetings and worked via completed in April 1995. mail to develop and refine a system for The aspects of the interface describing foods. The report that standard, which are linked to the food resulted from this work (13) provides for names, include food name synonyms, free text descriptors for specific Langual factor terms, other food characteristics of foods. The system, descriptors (agricultural and storage which was not specifically designed for variables), other descriptive coding computer implementation, includes six systems, ingredients and recipes, food major facets (Table IV): source of food standards, and reference files. The name and descriptive terms; name and reference files allow for the identification identification of the food; description of of substances administered or applied “single” foods; description of “mixed” during production and storage, the foods; customary uses of food organization that produced or prepared (optional), and sampling and laboratory the food, and the source of the data. handing of food. The INFOODS system As much information as possible is was not intended to supersede or provided about the food without making replace systems currently in use, but to questionable assumptions. Only those support and be compatible with them descriptors that pertain to a food need to (13). Figure 1. International interface standard for food databases Table IV. Major facets of the INFOODS system for describing foods1 A. Source of food name (5) and descriptive terms B. Name and identification of the food 1.Name in national language 2.Local name 3.Nearest equivalent name in English, French, or Spanish 4.Country/area where obtained 5.Food group and code in national database 6.Food group and code in regional database 7.Codex Alimentarius indexing group C. Description of “single” foods 1.(a) Food source (b) Scientific name (Latin) (c) Variety, breed, strain 2.Part of plant or animal 3.Country/area of origin 4.Manufacturer's name and address (batch or lot number) 5.Other ingredients 6.Food processing and/or preparation 7.Preservation method 8.Degree of cooking 9.Agricultural production conditions 10.Maturity or ripeness 11.Storage conditions 12.Grade 13.Container and food contact surface 14.Physical state, shape, or form 15.Color 16.Other descriptors 17.Availability and location of photograph/drawing of food D. Description of “mixed” foods 1.Ingredients and quantities 2.Recipe procedure 3.Place where prepared 4.Availability and location of photograph/picture 5.Manufacturer's name and address 6.Container and food contact surface 7.Preservation method 8.Storage conditions 9.Final preparation E. Customary uses of food (optional) 1.Typical portion weight and measure 2.Availability (frequency and season of consumption) 3.Role of food in the diet 4.Food users 5.Specific purposes of the food; special claims F. Sampling and laboratory handling of food 1.Date of collection 2.Weight(s) of sample(s) 3.Percentage edible portion; nature of edible portion 4.Percentage of refuse; nature of refuse 5.Place of collection 6.Handling between supplier and laboratory 7.Handling on arrival at laboratory 8.Laboratory storage and subsequent handling 9.Strategy for analyses 10.Reasons for doing analyses 1 Adapted from Truswell et al. (11). Examples of American food names that Ŷ Importance of Terminology Systems are implicit misnomers are corn dogs, grasshopper, hush puppies, rocky Terminology systems allow for mountain oysters, and sweetbreads. descriptive information about foods in a Examples of implicit misnomers from the consistent, standardized way that UK are Scotch woodcock, spotted dog, extends beyond the food name. Many toad-inthe-hole, and Yorkshire pudding. food names are not sufficient by Some implicit geographic food names themselves to identify foods. Descriptors imply an area of origin (Brussels are especially useful for implicit food sprouts, Danish (pastry), English names; different foods that have the muffins, Lima beans, and London broil), same name; foods that have different but have little to do with the identified names; and vague, generic names. areas. Terminology systems can address these problematic food names through Same Name, Different Foods descriptive terms relating to food Some foods share the same (or nearly source, food group, Latin name, the same) name, but are different foods. language of food name, maturity, “Tuna” in American English is a fish; in geographic region, cuisine, synonyms, Mexican Spanish, the term refers to a preferred terms, and/or other facets. prickly pear. “Rape” is a plant oil used in Implicit Food Names Mid-Eastern cookery, a Spanish fish, or a French cheese. In England and There are several types of implicit food France, “flan” is an open fruit tart in names. Some convey no meaning sponge cake or pastry crust; in Mexico without prior familiarity and do not or Spain, it is a baked caramel cream translate meaningfully to other custard. A terminology system which languages. Examples include bubble defines the language of the food name and squeak (British), and the cuisine is useful for kaerlinghedskranse (Danish “love distinguishing the correct usage of a rings”), hete bliksem (Dutch “hot food name. lightening”), himmel und erde (German There are many examples of the “heaven and earth”), scottadito (Italian “same name, different food” problem “burning fingers”), himmelsk lapskaus among American and British names for (Norwegian “heavenly potpourri”), brazo foods. “Half-and-half” in the UK is a de gitano (Spanish “gypsy's arm”), putt i beverage of half porter and half pale ale; panna (Swedish “tidbits in a pan”), and in the US, the food name refers to a the American foods baked Alaska, red mixture of cream and milk. “Mince” flannel hash, pigs-in-a-blanket, and could be chopped ground beef or succotash. Most of the commercial chopped fruit in the UK, but is chopped, names for alcoholic mixed drinks dried fruit (mainly raisins) in the US. (Bloody Mary, Rusty Nail, Screwdriver), “Silverside” is a beef cut in the UK and a ready-to-eat breakfast cereals fish in the US. A cordial is a soft drink in (Frankenberries, Froot Loops, Pebbles), the UK, but is a concentrated alcoholic and candies (Baby Ruth, M&Ms, Now'n' beverage in the US. A terminology Later, Payday) are fanciful, implicit system which identifies the language of names. the food name and specifically Some food names are implicit distinguishes between different forms of misnomers, i.e., the literal translation the same language (i.e., English in the may lead to the wrong food. If one is not UK, the US, Canada, and Australia) familiar with these food names, the would assist the data user. wrong conclusions may be drawn. Common usage of food names providing information on ingredients and (usually a tendency to shorten the geographic descriptors. name) may result in names that refer to Food standards (e.g., the definitions several different foods. For example, the for what constitute milk, butter, term “chili/chile” may refer to a chili margarine, beer, wine, ice cream), pepper (vegetable or spice), to chili nutrient fortification levels, and nutrient beans (beans with a chili pepper sauce), claims (e.g., low fat) are established by or a mixed dish made with beef, beans, government regulations and vary among and a chili pepper sauce. The term countries. A terminology system should “curry” may refer to the spice or to a rice allow for descriptive information relating dish made with the spice. The term to food standards, nutrient fortification “dressing” refers to salad dressing as levels, and claims and identify the well as to poultry stuffing (breading). A country associated with these legal terminology system can help clarify terms. these many uses of a food name Same Food, Different Names through food groups, food source, and homonym definitions. The “same food, different name” Some foods share the same name, problem can be handled by synonyms in but are prepared with different a terminology system. In many cases, ingredients and are not really the same the preferred food name varies by foods. For example, cocoa (hot geographical location or culture. There chocolate) is usually made with milk, but are different names for the same food some of the instant, dry cocoa products within a country, e.g., ocean perch is are reconstituted with water and contain known regionally in the US as rosefish, little or no milk. Similarly, “lemonade” redfish, snapper, sea perch, and may be made from lemons or with redbeam (17). There are different artificial flavoring. The nutrient data names for the same food in the same associated with various cocoas and language among countries, e.g., lemonades show clear differences in American molasses and British treacle; these products. Main dishes, soups, American oatmeal and British porridge; salads, and desserts may share the American raisin bread and British same food name (e.g., lasagne, currant loaf; American gelatin dessert gazpacho, carrot cake), but have and British jelly, and American jelly and different recipes. A terminology system British jam. should allow for information on Vague, Generic Names ingredients and recipes (how the ingredients are put together) and Food descriptions in databases are information on place of procurement often lacking for basic, traditional foods (e.g., restaurant, homemade, grocery such as fruits, vegetables, animal flesh, store). and grain products (breads, etc.). For Some foods have the same example “oranges” and “white bread” commercial product name (Kellogg's could be described by year of Corn Flakes, McDonald's Big Mac), but production and/or market share of are made from different ingredients in cultivars and brands, respectively. Such different countries. Different descriptors are especially important formulations may be due to different when database compilers are food standards, different nutrient aggregating data from various sources fortification levels, the local availability of and filling in missing values by matching ingredients, or local taste preferences. A food names. A terminology system terminology system may help by could allow for these types of descriptions through agricultural variables and information on sampling consistency in the use of defined terms; designs. access to a hierarchy of terms with Database users need to know if information on narrower, broader, and “generic” foods are market basket preferred terms and synonyms; retrieval samples and/or mixtures of cultivars or of food names based on descriptive maturity levels. If generic foods are not terms across food groups; and ability to adequately described, inappropriate or match foods in various databases based misleading conclusions may be drawn on identical or similar descriptive terms. about the data associated with them. Terminology systems must keep up For example, the vitamin A content of ½ with foods available in the marketplace grapefruit (120 g) is 318 IU for pink and which are changing to meet consumer red and 12 IU for white (6). The preferences for convenience, weighted vitamin A value of 149 IU for appreciation of ethnic foods, and the US market share product (6) does increased interest and knowledge of not reflect either the pink or white nutrition. Several types of foods in the product. marketplace are presenting challenges for food description systems. They Ŷ Current Status and Future Goals include products from newer or Food classification systems are changing plant cultivars and animal developed by database compilers breeds; foods previously used only by according to the number and types of select population groups that are foods in the database, cultural uses of becoming available in different foods, and/or intended users of the geographic areas (e.g., ugli fruit, database. Thus, each food composition jicama); synthetic foods made of database tends to have its own system. mixtures of refined ingredients (formula- The importance of food group type meal replacements, medical foods); classifications in databases depends on meat analogues; traditional foods made the types of terminology systems that with fat and sugar substitutes; and are present, i.e., what other sorting or traditional foods that have been retrieval mechanisms are available. If reformulated to meet special dietary there are no other mechanisms to claims. describe or locate foods in a database, The type and level of descriptive then classifications are very important, information needed about foods vary and must be carefully structured to among database users (i.e., place foods logically and consistently in researchers, epidemiologists, the hierarchy. If there are other means government agencies, educators). by which to describe (and hence However, it is possible that a retrieve) foods, then the classification terminology system can serve multiple system is of less importance. needs. It is important to note that current Because food classification systems systems are not incompatible and that are culture dependent, they are much knowledge and experience have probably best designed to assist been gained by the development of immediate (local) users of the database. several different systems. A universal classification system is not Foods in databases must be clearly necessary for the exchange and sharing and accurately described so that we can of information in food-related databases. better use the data associated with Faceted terminology systems, them. Descriptive information especially those with standardized associated with foods prior to laboratory vocabulary, have specific advantages analysis (e.g., information about for use with food composition sampling, preparation, and cooking databases. These advantages include methods and information from labels), should be recorded and carried with the Release 6, National Technical food composition data to the database. Information Service, Springfield, VA Countries need to work together toward (8) McCann, A., Pennington, J.A.T., flexible and compatible food description Smith, E.C., Holden, J.M., Soergel, systems for databases to increase the D., & Wiley, R.C. (1988) J. Am. capability to capture, exchange, share, Diet. Assoc. 88, 336–341 and retrieve information about foods. (9) Food and Agriculture Organization (1982) Food Composition Tables for Ŷ References the Near East, Rome (1) Poortvliet, E.J., & Kohlmeier, L. (10) Haendler, H., Neese, U., Jager, F., (1993) Manual for Using the & Harris, L.E. (1980) in International Eurocode 2 Food Coding System, Network of Food Information Federal Health Office, Institute for Centers, Pub. 2, L.E. Harris, H. Social Medicine and Epidemiology, Haendler, R. Riviere, & L. Berlin Rechaussat (Eds.), International (2) Souci, S.W., Fachmann, W., & Feed Databank System, Utab State Kraut, H. (1989) Food Composition University, Logan, UT and Nutrition Tables 1989–90, (11) Pennington, J.A.T., & Hendricks, Wissenschaftliche T.C., (1992) Food Add. Contam. 9, Verlagsgesellschaft mbH, Stuttgart 265–275 (3) Fettsyratabeller for Livsmedel och (12) Pennington, J.A.T., Hendricks, T.C., Matratter (1989) Statens Douglass, J.S., Petersen, B., & Livsmedelsverk, Produktion Kidwell, J. Food Add. Contam. (in Informako AB, Stockholm press) (4) English, R., & Lewis, J. (1992) (13) Truswell, A.S., Bateson, D.J., Nutritional Values of Australian Madafiglio, K.C., Pennington, Foods, Australian Government J.A.T., Rand, W.M., & Klensin, J.C. Publishing Service, Canberra (1991) J. Food Comp. Anal. 4, 18– (5) Holland, B., Welch, A.A., Unwin, 38 I.D., Buss, D.H., Paul, A.A., & (14) Arab, L., Wittler, M., & Schettler, G. Southgate, D.A.T. (1991) McCance (Eds.) (1987) in European Food and Widdowson's The Composition Composition Tables in Translation, of Foods, 5th Ed., Royal Society of Springer-Verlag, Berlin, pp. 132– Chemistry, Cambridge 154. (6) US Department of Agriculture (15) Kohlmeier, L. (1992) Eur. J. Clin. (1976-) Composition of Foods: Nutr. 46 (Suppl. 5), S25–S34 Raw, Processed, Prepared, Agric. (16) Deary, J. (1993) Langual Coding Handbook No. 8 series, USDA, Experiment, MAFF, London Washington, DC (17) FDA (1988) The Fish List. FDA (7) US Department of Agriculture Guide to Acceptable Market Names (1993) USDA Nutrient Data Base for Food Fish Sold in Interstate for Individual Food Intake Surveys, Commerce, US Government Printing Office, Washington, DC Quality Control of Food Composition Data and Databases

Nutritional Metrology: The Role of Reference Materials in Improving Quality of Analytical Measurement and Data on Food Components

James T. Tanner

Center for Food Safety and Applied Nutrition, Food and Drug Administration, Washington DC 20204, USA

Wayne R. Wolf

Food Composition Laboratory, Beltsville, Human Nutrition Research Center, ARS, US Department of Agriculture, Beltsville MD 20705, USA

William Horwitz

Center for Food Safety and Applied Nutrition, Food and Drug Administration, Washington DC 20204, USA

This paper discusses the role of reference materials (RMs) in improving analytical results in order to complement existing quality control procedures focused on processes such as standard methods and collaborative trials. Activities to improve the range of RMs available, and their incorporation into standard methods are also discussed. nalytical measurements of the content of food components are the foundation of nutritional science. Knowledge and application of the principles of metrology (the Ascience of measurement) are essential to improve and assure the quality of the data generated by these measurements. In the past, analytical methodology for nutrient measurements had focused primarily on the process of these analytical measurements, i.e. the emphasis on use of Official Methods of Analysis which have been collaboratively studied and evaluated through procedures established by AOAC INTERNATIONAL (formerly the Association of Official Analytical Chemists). These collaborative studies show the capability to achieve agreement of results among analysts using specifically defined analytical procedures. More recently metrology in general has chemical stoichiometry. The assignment focused on the result of the analytical of reference values by a certifying measurement process, i.e. the accuracy organization, based upon validation by of the data generated by the specific experienced laboratories faithfully application of the procedure. There is a following the details of the same well recognized need to build a method, produces the value which is to foundation for data validation through be reproduced by laboratories supplying establishment of accuracy based analytical values to nutritional science. measurement systems (1). In these Only if a reference value can be systems “routine” or “field” duplicated by an analytical laboratory methodologies are linked and traceable can any degree of confidence be through Reference Materials (RMs), ascribed to values produced by that Reference Methods, Certified Reference laboratory for the same nutrient in other Materials and Definitive Methods to the foods. basic measurement systems of national Indeed the foundation of the U.S. and international bodies. The use of Food and Drug Administration (FDA) RMs in conjunction with Official Methods regulatory process is a tested, reliable is necessary to build this foundation, not method combined with a reference only for establishment of an accurate material to validate accuracy of the database of food composition data, but resulting analytical data. This is a basic also for the monitoring of appropriate requirement of Good Laboratory regulations dealing with these types of Practices (GLP) and the corner-stone of data. good science. In its regulatory programs This concept of “traceability” is the FDA requires use of the analytical essentially important in nutritional methods of AOAC INTERNATIONAL, science because many of our essential which have been validated through nutrients are not single chemical interlaboratory methods performance entities, but are families of related studies to ensure that they are capable components. Chemical families of providing acceptable accuracy and ordinarily can not be analyzed by precision. This requirement does not methods designed for specific analytes. eliminate use of other analytical They require tailor-made methods that methods which have been evaluated try to include only components of through similar studies. Indeed the Code nutritional interest. Therefore, many of Federal Regulations (2), which nutrient measurements are method specifies that AOAC methods will be specific, requiring that the procedures used for regulatory purposes, requires be followed in exact detail to obtain that: “…if no AOAC method is available, repeatable answers. Such methods are by reliable and appropriate analytical even more dependent on reference procedures.” Other methods developed materials than are methods based upon by such organizations as the American Association of Cereal Chemists, the same pace as for inorganic American Oil Chemists Society, the components. International Standards Organization Reference materials are also (ISO), or other organizations may in necessary to determine the systematic some cases also be useful for regulatory error of new methods. Previously, some purposes. AOAC methods had used standard However, all of these methods additions for checking for the presence provide only half of the requirement. In of method bias, when a reference addition to a well-studied method, some material of known concentration was not means of determining that the method available. Although this technique is was performed correctly is also useful under some conditions, it really necessary. Obtaining acceptable results only measures the analyst's ability to with validated methodology for a recover analyte added at the reference material that has a known measurement stage and not the ability concentration of the analyte and is to determine the analyte that was similar in composition to the material endogenous to the matrix. For this being analyzed is presumptive evidence reason, the technique of standard that the method was performed correctly additions sometimes gives unreliable and that the results obtained for the test information. The determination of materials are correct. RMs, for which the precision or reproducibility is frequently true values are known, are important for used as a measure of the success of a this validation. From a regulator's point method because of the ability of a of view, the use of appropriate RMs is laboratory to obtain the same values as desirable for determining compliance well as to replicate the results of other with existing regulations. laboratories. This is an important part of Unfortunately, RMs are not available method evaluation but does not address for many products and analytes. Dating the accuracy question. The International back over 80 years, standard reference Standards Organization (ISO) has now materials (SRMs) have been developed broken down the concept of “error” as by such organizations as the National deviation from the true value into three Institute of Standards and Technology parts: 1) “Accuracy” is the deviation of a (NIST, formerly National Bureau of single value; 2) “Trueness” is the Standards, NBS) for products such as deviation of the average set of values; steel, in which the content of trace and 3) “Bias (or systematic error)” is the elements is very important. Building on deviation of the long term average (3). this expertise, RMs have been AOAC INTERNATIONAL recently developed within in the past 20 years for formed a task force to address the biological products such as flour, problem of the methods available to spinach, oysters and other food enforce regulations stemming from the products for which the main focus has Nutrition Labeling and Education Act of been the major and trace elements 1990 (NLEA) (4), which made nutrition rather than the various organic labeling mandatory as of March 1994 for compounds comprising the major retail foods distributed in the United components of food. One reason for this States. The purpose of the task force focus has been that some organic was to determine what methodology components may change with time and was available and whether existing are not shelf stable, therefore, the exact methods were adequate for the purpose “true” concentration at the time of use of nutrition labeling (5). In addition to cannot be assigned. Another reason is methods questions, the task force also that analytical expertise for organic examined the availability of RMs. It components has not progressed at the found a serious deficiency in the availability of RMs for organic nutrients of the nine different sectors, then it and recommended that action be taken should be applicable to all types of food. to improve that situation (6). Such an approach would also be useful Several problems must be addressed to AOAC Associate Referees and AOAC before reference materials for organic Official Methods committees in nutrient content can be made available. minimizing the effort required for The first is the selection of matrix collaborative studies while maximizing materials to represent many different the value of the resulting data to AOAC kinds of foods; the second is the Official Methods users. For example, the packaging and storage of these prospect of coordinating a collaborative materials to provide a useful shelf life. study involving 40 or more different Third is the characterization or foods may discourage many assignment of the “correct” or “best researchers from fully exploring the estimate” of the value for the scope of applicability of a particular components of interest. method. As a result, reseachers may The AOAC task force addressed the limit the scope of their study to a few question of matrix materials for different food groups to reduce the analytical foods in a creative way (7). Food is burden on the participating laboratories. composed of the basic components: However, as demonstrated by the protein, carbohydrate, fat, water and triangle, many of the 40 or more foods ash. Frequently, analysis of a food is not selected to represent foods for a successful because of interference or collaborative study may be very similar interaction from one or more of these to one another on a dry basis, and may components with the analyte of interest. behave chemically, and, thus, In any analytical procedure, water can analytically, in a very similar way. usually be added or subtracted to suit If a diagram such as Figure 1 were to the requirements of the method. Ash, in be used to select samples for a general, does not have a great impact collaborative study, two samples from a on the performance of analytical sector could be selected to account for methods for organic material in foods. variation in the type of protein, fat, or Thus, the behavior of a given food in an carbohydrate that may have an impact analytical method is primarily on the performance of the method. determined by the relative proportions of Examples of these variations within protein, fat, and carbohydrate. carbohydrates are high fiber foods A scheme has been proposed to versus high sugar foods. Other represent foods by first normalizing variations include fats containing content of these three components to significant amounts of short chain fatty 100 per cent of their sum (7). This acids versus those containing normalized food composition can then predominantly long chain fatty acids, or be plotted within a triangle with 100 per foods containing more hydrophilic cent fat, 100 per cent protein, and 100 proteins as opposed to those containing per cent carbohydrate at the respective predominantly hydrophobic proteins. In vertices with the concentration of each addition, two foods may be selected component decreasing to zero within a sector that vary according to the approaching the opposite side. This extent of processing each has schema can then be divided into nine undergone. different sectors, each encompassing a The logical extension of this same range of concentrations of the three approach would be to provide components (protein, carbohydrate, and appropriate reference materials for a fat) (Figure 1). If a method of analysis food type or category representing each were successful for foods falling in each of the nine sectors. By using the different types as part of a method- produced and made available through performance study and having a an organization such as NIST. The first reference material for each type, all priority would be to produce reference foods would have a method and a materials for the nine food sectors reference material, similar to the actual named above and for products in areas food, that could be used for regulatory where a critical need exists for reliable purposes. These RMs could be analyses, such as medical foods.

Figure 1. Schematic layout of food matrices by which all foods can be organized according to their relative proportions of protein, fat, and carbohydrate; the points of the triangle represent 100 per cent of the normalized content of these three major classes of food components (moisture and ash are excluded).

A reference analytical method of developed and collaboratively studied known reliability together with a stable because infant formula is the most RM to monitor analytical performance is highly regulated food in the United the most important requirement for a States today. It represents the sole regulatory agency. With results source of nutrition for a large segment of produced by using this combination, the the population, namely, infants. As part agency can proceed with appropriate of the Infant Formula Act of 1980, regulatory action that is based on sound companies are required to manufacture analytical science. formula within specified limits, and FDA This type of verification is part of the is required to monitor the formulas to infant formula program. Methods for the ensure that they are within those limits. analysis of infant formula have been Because of differences in methodology, many questions have arisen as to the AOAC Official Methods of Analysis. In “true” concentrations of some analytes addition this Technical Division will in the products. Currently, there are coordinate activities to assist in analytical methods for infant formula characterizing RMs and will conduct the that both industry and FDA have agreed International Symposia Series on are to be used for regulatory analyses. Biological and Environmental Reference These methods are now part of AOAC's Materials (BERM) (9). Official Methods of Analysis (8) and Ŷ References have been collaboratively studied by FDA, infant formula manufacturers, and (1) Uriano, G., & Cali, J.P. (1977) in several commercial laboratories. Validation of the Measurement However, no reference material is Process, ACS Symposium Series currently available for validating method No. 63, J.R. Devoe (Ed.), ACS, performance in each laboratory. One Washington DC, pp. 114–139 on-going NIST project is the (2) Code of Federal Regulations, (21 development of a spray dried Infant CFR 101. 9 (e) (2)) Formula material (SRM-1846) which is (3) International Standards being characterized for organic nutrient Organization (1994) ISO Standard content. SRM-1846 will serve as a 5725 reference material for Infant Formula, (4) Ellefson, W. (1993) in Methods of and will also provide a least one Analysis for Nutrition Labeling, D.M. reference material for validating Sullivan & D.E. Carpenter (Eds.), measurements that determine AOAC INTERNATIONAL, Arlington, conformity with the requirements of VA, pp. 3–26 NLEA. An infant formula has been (5) Sullivan, D.M., & D.E. Carpenter prepared, spray-dried, and packaged (Eds.) (1993) Methods of Analysis under nitrogen in individual packets for Nutrition Labeling, AOAC weighing approximately 30 g each. INTERNATIONAL, Arlington, VA These packets have been stored for (6) Wolf, W.R. (1993) in Methods of about two years and analyzed at Analysis for Nutrition Labeling, specific intervals. They appear to have AOAC INTERNATIONAL, Arlington, been shelf stable for that time period. VA, pp. 111–122 Further testing is still under way. If (7) Ikins, W., DeVries, J., Wolf, W.R., successful, this method of packaging Oles, P., Carpenter, D., Fraley, N., could be applied to other potential RMs & Ngeh-Ngwainbi, J. (1993) The to ensure that the nutrient content is Referee 17, 1, 6–7 stable for a reasonable time. (8) Official Methods of Analysis (1995) An AOAC international Technical 16th Ed., AOAC INTERNATIONAL, Division on Reference Materials has Arlington, VA been established in order to facilitate (9) Heavner, G. Fres. J. Anal. Chem. availability and use of RMs in the (in press) validation, implementation and use of Quality Control of Food Composition Data and Databases

Strategies for Sampling: The Assurance of Representative Values

Joanne M. Holden, Carol S. Davis

Food Composition Laboratory, Beltsville Human Nutrition Research Center, ARS/USDA, BARC-East, Beltsville, MD 20705, USA

Current interest in the relationship of diet to the maintenance of health has stimulated the demand for representative food composition data. Values for nutrients and other food components are required to calculate dietary intakes, to determine food policy, to monitor food safety, to formulate new products, and to facilitate trade. A specific estimate must be statistically representative of the population of all values for a component in the food product of interest. Serious bias in the estimate can lead to erroneous conclusions about diet-related issues. The Food Composition Laboratory has conducted research to develop statistically based strategies for sampling the US food supply to determine estimates for components in many foods. To determine a strategy for food sampling it is necessary to define project objectives and to determine analytical priorities for foods and components. Foods to be sampled should be described in terms of the product type, ingredients, preservation state, source, cultivar, and other factors which may influence component levels. Demographic and marketing data can be used to identify parameters which are potential sources of variability. In addition, protocols for sample handling and chemical analyses should be standardized to minimize the impact of errors which may arise during the measurement process. Results of sampling research for selenium, total fat, and cholesterol in several foods are presented and the impact of sampling results on the calculation of national estimates is discussed.

ince 1960 the assessment of food consumption patterns and their impact on health status has evolved, requiring food composition data for more foods and Scomponents (1). The recognition of food intake as one factor in the longitudinal development of complex, multifactorial diseases has occurred more recently (2, 3, 4). Not only are food composition data used to identify and monitor dietary trends but they are also used for hypothesis testing (5). Other uses of food composition data are equally important (e.g. trade, food safety, food manufacturing) (6). This increased interest in food composition data has stimulated the demand for improved data, including an indication of the number of analyses, the sampling plan, and the magnitude and sources of variability, as well as descriptive and quantitative information about the analytical method and quality control (7). The lack of data for foods and ingredients impedes the assessment of diet-health relationships and impacts on the production, regulation, and use of foods. Increased demand for more data can be attributed, in part, to the development of sophisticated instrumentation which permits the measurement of minute quantities of components in foods and in biological matrices more rapidly than ever before. Similarly, the development and accessibility of computers for data processing has improved the ability to manipulate large data files to investigate new hypotheses. In view of the importance of foods as vehicles for nutrients and other components, the generation of food composition data is not an isolated exercise but, rather, an integral part of the assessment of human health status and dietary effects.

Possible specific objectives for objective of this paper is to discuss the generating food composition data development of sampling strategies to include: provide estimates of central tendency ƒdevelopment of a national food and variability for component levels in composition database foods to be used in food composition databases and national dietary ƒdetermination of aflatoxin levels in a assessment projects. rail container of grain The average daily diet may contain ƒdetermination of pesticide levels in a 20–25 different items. It has been food product estimated that 4,000 different generic ƒquality control of food manufacturing products (e.g. beef, white bread, pizza) can be found in the American ƒdetermination of significant marketplace. Since a nation's food differences in the vitamin content of supply is a complex mixture of different animal muscles processed and non-processed products ƒbrand to brand (or region to region) each food item represents many brands, comparisons of component levels. formulations or styles, and geographical The generation of these data should sources. There may be as many as be based on a statistical sampling plan 50,000 products if one considers specific to the objective which will different brand names. For example, in indicate what to sample, where to the US there are hundreds of brands of sample, and how many units to select to white bread (9). Similarly, the diversity represent the food of interest. The of the population, personal preferences definition of the objective provides the for foods, and the availability of focus for the study and helps to sophisticated manufacturing and determine the most appropriate marketing schemes stimulates the sampling strategy. According to Horwitz nationwide distribution of new and a statistically based sampling plan unusual products. Due to the complexity should guide the selection of of a national food supply, the generation representative units from the population of accurate food composition data is a to provide component estimates “within difficult and expensive task. a specified degree of variability with a stated degree of confidence” (8). The Figure 1. Population and sample: the definition of representativeness

Figure 1 illustrates the statistical consumed in the US one would probably concept of the sample and its not sample such cultivars since they are relationship to the population of all not widely consumed. While it is not forms, brands, and units of a food (10). possible or desirable to analyze every The term population describes the package, unit, or lot of a food, the collection of relevant objects from which analysis of a subset of carefully selected a subset is chosen for analysis. units will provide the required data to Generally, the population of interest is draw inferences about the population of very large and can be considered infinite all available units (10) (Figure 1). relative to the size of the subset which is Using traditional survey sampling to be selected. For example, in Figure 1, theory, the term sample refers to that the population consists of all forms of subset or group of items or units which carrots usually consumed by individuals. are selected from the population of In this same example experimental interest to represent that population (10) cultivars of carrots lie outside of the (Figure 1). If the objective is to develop population but are part of the larger a nationally representative database of universe of all carrots. When estimating food composition values, then the levels of a nutrient contained in carrots sampling strategy must be carefully planned to construct a sample to include composition data to yield estimates of typical items or units in shares population parameters. Although the proportional to the sales volume or discussion of mathematical sampling consumption properties of the theory is beyond the scope of this paper population of those foods. The sample it provides the framework and point of will include units of predominant brands, reference for comments about the manufacturing locations, cultivars, etc. selection of foods, the number of units, relevant to the specific food as variability, etc. It is important to note that consumed by the individuals of interest. the usual statistical techniques which If one were to analyze all containers are used to evaluate the statistical or units for all available brands or characteristics of the sample subset and cultivars defined as the population for a to provide estimates of statistical food, e.g. carrots, then one could parameters for that subset assume construct a frequency distribution of all normality of the distribution of all analytical values. The distribution may possible analytical values in the subset. or may not be Gaussian or normal. In some disciplines various Since all units of a food which constitute mathematical transformations of the the population cannot be analyzed data are possible to permit the without destroying that population the evaluation of scientific hypotheses. concept of sampling, i.e., selecting a However, it is difficult if not impossible to representative subset of the population use transformation techniques to based on the probabilities of various estimate such parameters as the mean types has developed (10). The analysis and variance. More research is needed of all units in the subset will yield a for many components and foods to collection of values which can be used determine the statistical distributions for to construct a frequency distribution for food composition data and to evaluate that subset. If the sample is the robustness of statistical techniques representative of the population than as applied to such data. that distribution will be similar in its In general, most values for statistical characteristics and components and foods in a database subsequent “shape” to the distribution are calculated means of two or more for the population. If one were to take individual values. For analytical sources multiple samples, i.e., multiple subsets or files the data may have been of units, of the same size from the same generated in a single laboratory or in large population one could expect that several laboratories. Individual values the statistical characteristics of each may be the product of the analysis of an sample would be similar to those for the aliquot of a single unit or of a composite population. However, they will not be of several units. Each mean value in a identical since the collection of mean database is a point in the distribution of values for all samples taken from the sample means mentioned above and, population will form a frequency yet each mean also represents a distribution themselves. The degree of distribution of individual values or points similarity of the statistical characteristics for the sample subset from which it was between the population and the sample derived. Since a mean database value defines, in part, the degree of represents a sample subset selected representativeness of that sample for from the population a new analytical the population. While, in most cases, the value for another individual unit chosen true statistical characteristics of the at random from the food supply may not population can never be known fall within the confidence limits of the statistical sampling theory can be database value. However, the applied to the generation of food probability of any new value falling within limits defined by representative projects can be driven by the need to sampling and analysis will be high (10). estimate levels of a single component Thus, it is important to estimate the (e.g. selenium, ȕ-carotene, total fat) in mean composition and some parameter foods consumed by a population of of variability for the most important individuals. Conversely, the focus may food/component combinations in a be on a single food (e.g. beef, milk, database. carrot) and its major components. Recently, Greenfield and Southgate Ŷ What Components Should Be have published a discussion of the Analyzed? importance of sampling, including important definitions and approaches for The components of interest may be obtaining the representative sample set nutrients (e.g. protein, vitamin A, iron), (11). Analyses may or may not include additives, biological agents, or aliquots of all brands, types, or cultivars contaminants. Each component or class present in the population. In keeping of components represents a unique with fiscal and physical constraints, it sampling challenge. However, the may be necessary to take a subset of choice of components should be guided the brands or types available. One by the particular priorities or emphasis should seek statistical advice during the of the project or agency. In general development phase of the sampling three factors determine the selection of plan. Aliquots of single units (primary components: samples) may be analyzed. Conversely, ƒthe component should rank highly units can be combined or composited by relative to actual or suspected public brand name, geographic location, health effects cultivar, etc., as appropriate, before available analytical methods for the aliquots are taken to minimize the ƒ component(s) of interest should be number of analytical measurements and robust, valid, capable of producing yet represent the contribution of that unit accurate data, and economically to the estimate of central tendency. The feasible formulation of composites should be based on the statistical data about the ƒin view of fiscal and personnel collection of units representing brand limitations, analytical priorities names, geographic locations, etc. which should include those components for have been obtained from a pilot study or which available data are previous independent investigations. unacceptable or previously The impact of compositing on the unavailable (12,13). magnitude of variability should not be As an example, the scientific overlooked. The number of analyses to community has become interested in the be conducted will be determined by the possible effects of carotenoids intake on desired statistical power of the estimate, health, specifically the development of the observed variability in pilot tests, certain cancers (14). Since the 1930's, and such practical considerations as several carotenoids (Į and ȕ-carotene, physical and fiscal resources. More and ȕ-cryptoxanthin) have been known detail will be provided later in the text. to have significant vitamin A activity If the objective is to develop a (15). While vitamin A deficiency is still national food composition database, prevalent in many areas of the world, then two major questions need to be the broader role of carotenoids in answered: “What nutrient human metabolism has become the (s)/component(s) should be object of interest. However, until determined?” and “What foods should recently, no comprehensive assessment be selected for analysis?” Food analysis of carotenoid data for foods had been conducted (16, 17). In fact, for many although many foods may contain the foods carotenoid data are lacking. component of interest, the foods Analytical methods for measuring selected should be the major additional individual carotenoids in contributors of that component to the simple forms of foods have been diet. Frequently, a limited number of developed in recent years (18). While foods (5–100) contribute 50–90 per cent more work needs to be done in this area of a single component to the diet of the to release a robust field method, some population of interest (23, 24). Existing centers are using liquid chromatography data and/or pilot studies can provide (LC) while other centers are using valid preliminary estimates of the levels of open column chromatography (OCC) components in foods. Food methods (19). Finally, carotenoids are consumption survey data and/or data good candidates for further analyses from food balance sheets can be because sufficient high quality data are combined with preliminary food lacking (16, 20). As other less familiar composition data to provide a ranked list components (e.g. isoflavonoids, of the major contributors of specific flavonoids) have become the objects of components (20, 24). research, and as robust methods have Second, foods for which data are become available, their determination in unacceptable or unavailable should be foods will become important. selected. As an example, tomato Furthermore, as improved methods for products are the most important source recognized important components (e.g. of lycopene, an abundant carotenoid in folates) are developed, new analyses the US diet. However, after the will be needed. As new, more specific, assessment of carotenoids data quality analytical methods become available it for multi-component foods by Chug- is necessary to generate new data to Ahuja et al. (20), the authors determined replace the older outdated values. Data that analytical carotenoid data for for fiber content of foods is an example. popular commercial, tomato-based Crude fiber analysis has been replaced soups, sauces, and spaghetti sauce by other methods, including total dietary were nonexistent. A nationwide fiber (21). Today, carbohydrate values sampling plan for three cities was calculated by “difference” have been developed to select samples of these replaced by analyses of specific products to be analyzed for five of the fractions since carbohydrates, as a most important dietary carotenoids (25). class, contain diverse forms with New forms of foods are appearing in the different molecular weights and markets of many countries and are chemical structures and, therefore, gaining in popularity. Food composition different metabolic effects. Frequently, data for many of these foods are newer methods make it possible to nonexistent. For example, the influx of determine some components for the first many previously unknown fruits and time. A large nationwide study of fast- vegetables into the US food supply food chicken included the determination requires that these foods be sampled of levels of starch contained in the and analyzed to determine their seasoned flour coating (22). composition. Initially, new foods may be imported from other countries, prior to Ŷ What Foods Should Be Sampled? commencement of their local production The selection of foods is equally (e.g. kiwi fruit, Granny Smith apples). important. Stewart et al. (11) and Since climate, soil conditions and Beecher and Matthews (12) have stated geography affect levels of some that priorities for analyses should be components, geographical source and based on three considerations: First, variety/cultivar may be relevant to the sampling plan. Therefore, it would be analyses. Finally, advances in animal necessary to compare data for imported and plant breeding will require new fruits with data for fruit from domestic analyses to estimate changes in sources; values would be revised, if targeted components. For example, in necessary. For a recent study of human the US, recent advances in breeding carotenoid metabolism, a single and marketing practices have production lot of frozen broccoli was dramatically reduced the separable fat needed to assure uniformity of the trim on beef and pork. Nationwide retail product for all subjects over the entire studies were planned and conducted in course of the study. When a small collaboration with meat science regional company was contacted to departments at Texas Agricultural and procure the broccoli it was found that Mechanical (A & M) University and the the product was grown and processed in University of Wisconsin to assess the Guatemala. Analyses of carotenoid impact of these changes on the levels in the frozen broccoli revealed composition of beef and pork (27, 28). that values were significantly different Ŷ Food Description Effects from those for fresh broccoli procured in the retail market (26). This revelation After the foods and components have emphasized the importance of using been determined, the individual and analytical values for critical components specific products which represent a food in single lot foods used in human must be identified. For a single food metabolic studies. (e.g. beef, pizza, or eggs) the A third consideration is the need to investigator can define the analyze foods as eaten. As new forms characteristics of the product which may of important foods become popular they influence the composition and variability should be analyzed to generate up-to- for the component(s) of interest. date data more appropriate to eating Relevant characteristics will include the habits (12, 13). In many countries the primary food source and scientific name use of fully prepared commercial foods for a product (e.g. wheat v. corn, instead of home-prepared commodities coconut v. sunflower, beef v. pork), the has increased rapidly. Estimates for part of the plant or animal used, those prepared foods are more preservation state, food processing representative of what some segments treatments, added ingredients, etc. For of a population are eating than some components, geographical source estimates for foods prepared from the (e.g. broccoli, above) and ripening basic ingredients. Formulated foods practices will be important. As may contain different levels of fat, mentioned, carotenoid values for sodium, or other components than broccoli cultivars grown in Guatemala domestic recipes. A recipe calculation were different than values for fresh technique can be used for some broccoli grown in California (24). For components. However, formulations for others, packaging type, pH, or storage commercial products are generally conditions will be sources of variability. unavailable and are frequently different In recent years, several food description from home-prepared products. systems have been developed which Therefore, the composition of important provide classifications of important commercially-prepared foods will need descriptors (29, 30). The specific to be determined by analysis. New products and components of interest will ingredients such as fat substitutes, determine the preliminary list of gums and sweeteners alter the descriptors for the products. formulations of familiar foods, Following the definition of relevant necessitating the need for new descriptors for a food, it is necessary to identify the specific major sources of products will need to be selected as well that food which are consumed by the as the time and location for sampling. population of interest. In addition, the Ŷ Food Consumption Patterns distribution and marketing schemes need to be identified. For branded After marketing and distribution products, sales volume data and variables have been defined, product information are important to the consumption patterns should be selection of representative units (9). For assessed to determine where to select commodity products, such as meats, the samples. If the objective is to eggs, milk, etc., it is possible to identify determine estimates for foods in a the major breeds or cultivars, as well as national database then it is necessary to major commercial purveyors of the sample food products on the basis of products and an approximation of their the population distribution and product sales ranking (31). In regions where use. Several questions should be food production is localized the major answered: Is the food consumed outlets for products (butcher, bakery) or frequently and in significant amounts by ingredients (flour mill, refineries) can be the population of interest? In which identified. Some products may be regions or populations is the food manufactured in one location and consumed? Is the food consumed more distributed nationwide while others may in rural areas than towns? If the food is be formulated in many regions from widely consumed by many sub-groups, different sources of raw ingredients. In a what is the distribution of the population recent study of selenium levels in bread, in the country or region of interest? ninety samples of white bread were Major population centers within a selected in nine major population areas country can be identified and used as across the US (32). Bread is baked in locations for sample selection. In the US regional or local bakeries in or near the majority of the population is cities and towns. Yet most of the wheat concentrated in a number of in the US is grown in the north central metropolitan areas called Metropolitan plains area, an area of relatively high Statistical Areas (MSAs) and defined by selenium levels, and then transported to the US Office of Management and major metropolitan areas to supplement Budget as cities which have at least smaller regional supplies grown near 50,000 persons or an urbanized area of those areas. Selenium levels in bread in at least 50,000 with a total population of several of the cities varied as the local at least 100,000 individuals (33). The source of the flour was supplemented by top ten cities, their percent of the the supply grown in the north central population, and their respective area. The study demonstrated that proportion of grocery sales are given in selenium levels in breas were more Table I. The percent of the population closely related to the source of wheat represented by the top 100 MSAs as levels of selenium in soil where the grain well as the number of supermarkets is was grown than to the selenium levels in also given. In most major cities two to soils where the bread samples were four supermarket chains dominate each purchased. By defining the form of the city. Most are significant regional product and its sources, the investigator vendors. can begin to determine which specific Table I. Top 10 US MSAa markets by populationb % of US Totals Rank Market Area Population Supermarkets 1 Los Angeles-Long Beach, CA 3.59 2.39 2 New York 3.38 2.19 3 Chicago 2.40 1.76 4 Philadelphia 1.93 1.41 5 Detroit 1.72 1.41 6 Washington, DC-MD-VA 1.61 1.25 7 Houston 1.35 1.12 8 Atlanta 1.19 1.21 9 Riverside-San Bernadino, CA 1.14 0.90 10 Dallas 1.07 1.01 ------100 Youngstown-Warren, OH 0.23 0.28 Top 100 MSA Market 59.86 50.86 All Other U.S. 40.14 49.14 U.S. Total Figures 254,926,669 30,552 a Metropolitan Statistical Area b Adapted from Progressive Grocer's Market Scope (40)

While many sample mixes are contributors five to 25 analytical selfweighted–that is, the available samples were chosen. By choosing products are similar to the number and units of the highest volume brands kind needed to mimic sales volume, it is within the largest supermarkets in major possible to weight the sample estimates metropolitan areas it was assumed that after analysis of equal numbers of the most frequently consumed and individual units/per brand or region by representative products were selected applying pre-determined weighting for a specific food. factors (10). In view of the nationwide To determine selenium in beef, it was distribution and market share of many necessary to determine the major products and the concentration of the categories of beef products in the US population in major MSAs the USDA diet. Marketing and production data and others have selected representative obtained from the US Livestock and units of foods from retail grocery stores Meat Board, the private sector trade and/or restaurants in three to ten cities association for the meat producers, across the country (22, 27, 28, 34). For indicated that the per capita example, for the recent study of consumption of beef was 72.7 lbs. Fresh selenium in approximately 200 foods, beef cuts including steaks and roasts, sample units were purchased in two and ground beef, including bulk ground major supermarkets in each of nine beef purchased in supermarkets as well cities (Holden, unpublished data). Two as hamburger sandwiches sold in fast to three cities were selected in each of food restaurants were the major forms four regions of the country; two major of beef consumed (35). Using this supermarkets were sampled in each information a sampling plan was city. For each of the major contributors developed. Ninety four units of five of dietary selenium (beef, white bread, primal beef cuts were obtained from a pork, chicken, eggs) approximately 100 larger study of beef composition analytical samples were randomly conducted by Texas A & M (27). The selected and prepared. For minor samples had been collected from major retail stores in ten cities. In addition, 58 number of samples is an iterative samples of ground beef collected from a process which begins with an USDA nationwide study were analyzed approximation of the number of samples (34). Finally, 27 samples of hamburger determined by the investigator as a sandwiches were collected in nine cities “guess.” The “guess” can be based on from each of three prevalent chains preliminary cost estimates or capabilities (Holden, unpublished data). Mean of the analytical laboratory. After the values and standard deviations were initial calculation the estimate of number calculated and published in the recently of samples is further refined by released USDA Provisional Table of recalculation until successive trial values Selenium in Foods (36). of “n” yield similar values. The cost of sampling can be included in the Ŷ How Many Sample Units Are Needed? equation as well. Table II demonstrates the effect of increasing the coefficient of The number of sample units analyzed variation on the number of samples will determine, in part, the statistical required to obtain the same level of power of the estimate. Although confidence. The “t” value was set at statistical models for calculating the 2.00 while r=0.1 for the purpose of the required number of units can be illustration. Further information is given complex and multi-tiered, the following by Cochran (10). equation indicates the most important In the past, the mean or average facets of the computation for value has been used as the estimate of homogeneous populations (10): the level of a component in the food. However, the use of the mean The appropriate number of units is presumes that the statistical distribution based on four parameters. The first is of all values for that component in a “t”, the abscissa of the normal curve that specific food follows the Gaussian or cuts off an area “a” at the tails of the normal distribution (37). Recently, the distribution, indicating the desired USDA Food Composition Laboratory, in confidence level. The standard error of collaboration with the US National the estimate is denoted by “s” while the Cancer Institute, complied and sample mean is denoted by . This published a food composition table of mean and standard error can be the levels of five carotenoids in obtained from previously published data important fruit and vegetable or pilot studies, if available. Some contributors (16). The values were existing handbooks of food composition collected from published and data publish standard deviations or unpublished analytical sources. Due to standard errors of the mean and can be the apparent skewed distribution for used as rough estimates of sample size. several foods and the limited amount of Previously published estimates and the available data (one to 14 acceptable scientific objectives for the study should sources per food) the median value was serve as the basis for sample number used in the table. However, the use of calculations. The reader should note the median precludes the calculation of that the coefficient of variation, if known, a variance indicator. More research is needed to evaluate the characteristics of can substitute for s/ . The limit of the statistical distributions which result from desired relative error in the estimate is broad-based original sampling as well indicated by “r.” That is, the proximity of as those which result from compilations the estimate to the “true” mean, e.g. of data from different sources. within 10 per cent, is represented by “r.” Furthermore, the robustness of The calculation of the appropriate traditional statistical techniques should be evaluated to determine how drawn from dietary studies, must be appropriate these techniques are for tested. In particular, caution is required food composition data. The impact of when estimating food composition using means v. medians in food values from small data sets. composition databases on conclusions Table II. Effect of increasing the coefficient of variation on sample sizea If CV equals then n equals 12.5% 9 25%b 9 50% 100 100% 400 a Į = .05, t = 2.00, r=0.1 b If Į = .10 then n = 19

After the sample is defined individual inconsistencies or ambiguities and items or units within the sample can be subsequent misinterpretation of the identified and procured to be prepared results. The reader is referred to for analysis. Once the units, packages, reference (8) for further information. or containers arrive in the laboratory Ŷ How Good Do The Data Have To their handling (e.g. preparation, Be? homogenization) and the selection of aliquots must be carefully planned to Food composition data must be “good maintain the representativeness and enough” to permit the careful integrity of the material. Since the assessment of food consumption developer of the project design and patterns and their impact on the health sampling strategy may not be the of population groups and subgroups. laboratory analyst the importance of Similarly, the data must be “good communication between these enough” to accomplish other scientific individuals or groups cannot be and economic objectives defined by overestimated. At this point it is investigators. The quality of a specific important to emphasize the use of estimate is based, in part, on the standardized nomenclature with regard accuracy and precision of the to sampling at the laboratory level. measurement process. The generation According to the 1990 recommendations of accurate food composition data for nomenclature for sampling in requires that variability inherent to the analytical chemistry submitted to the food be accurately quantified while International Union of Pure and Applied variability inherent to the measurement Chemistry (IUPAC) Horwitz defines the process be minimized. In general, the “sample” as “a portion of material major sources of statistical variability in selected in some manner to represent a dietary estimates are the food larger body of material. The result consumption data captured by the obtained from the sample is merely an dietary assessment tool, and the food estimate of the quantity … of constituent composition data. Variability for food … of the parent material.” Previously, composition data includes all variability the term sample has often been used to attributable to sampling plans, sample refer to the portion (e.g. extract, diluted handling, analytical method, and or not) being analyzed at various points analytical quality control. Each of these in the analytical process. Other terms sources can be partitioned into the such as “test” or “analytical” should be sources of variability and can be used to describe those portions to avoid quantified by an analysis of variance (37). The assessment of the sources and magnitude of variability for food Perspective, W.M. Rand, C.T. composition data can indicate areas Windham, B.W. Wyse & V.R. where improvement in the measurement Young (Eds.), UNU Press, Tokyo, process needs to be made (38). While pp. 177–193 sampling is only one source of (8) Horwitz, W. (1990) Pure Appl. variability, the lack of representative Chem. 62, 1993–1208 sampling can increase the degree of (9) Nielsen, A.C. Co. (1990) Nielsen bias in the estimates of central tendency Scantrack Data, Northbrook, IL and cause errors in the estimates of (10) Cochran, W.G. (1977) Sampling variance. As previously mentioned, for a Techniques, 3rd Ed., Wiley, New specific component, a small number of York, NY, pp. 1–78 foods may contribute the majority of that (11) Greenfield, H., & Southgate, D.A.T. component to the diet of the population. (1992) Food Composition Data: Therefore, it is recommended that Production, Management and Use, sampling resources be dedicated to Elsevier Applied Science, London obtaining statistically sound estimates (12) Stewart, K.K. (1981) in Beltsville for those major contributors. Symposia in Agricultural Research IV Human Nutrition Research, Ŷ Acknowledgments Allenheld, Osmun Publication, The author wishes to express her Totowa, NJ appreciation to the First International (13) Beecher, G.R., & Matthews, R.H. Food Data Base Conference for (1990) in Present Knowledge in generous financial support to attend the Nutrition, 6th Ed., International Life conference. Sciences Institute, Washington, DC Ŷ References (14) Le Marchand, L., Yoshizawa, C.N., Kolonel, L.N., Hankin, G.H., & (1) Life Sciences Research Office Goodman, M.T. (1989) J. Nat. (1989) Nutrition Monitoring in the Cancer Inst. 81, 1158–1164 United States: An Update Report on (15) Moore, T. (1957) Vitamin A, Nutrition Monitoring, US Dept. of Elsevier, Amsterdam Health and Human Services, (16) Mangels, A.R., Holden, J.M., Hyattsville, MD Beecher, G.R., Forman, M.L., & (2) Steinmetz, K.A., & Potter, J.D. Lanza, E. (1993) J. Am. Diet. (1991) Cancer Causes and Control, Assoc. 93, 284–296 2, 427–442 (17) West, C.E., & Poortvliet, E.J. (1993) (3) Hegsted, D.M., & Ausman, L.M. The Carotenoid Content of Foods (1988) J. Nutr. 118, 1184–1189 with Special Reference to (4) Katan, M.B., Van Gastel, A.C., de Developing Countries, Rover, C.M., van Montfort, M.A.J., USAID/VITAL, Washington, DC & Knuiman, J.T. (1988) Eur. J. Clin (18) Khachik, F., Beecher, G.R., Goli, Invest. 18, 644–647 M.B., & Lusby, W.R. (1992) (5) Judd, J.T., Clevidence, B.A., Methods Enzymol. 213, 347–359 Muesing, R.A., Wittes, J., Sunkin, (19) Rodriguez-Amaya, D.B. (1989) J. M.E., & Podczasy, J.J. (1994) Am. Micronutr. Anal. 5, 191–225 J. Clin. Nutr. 59, 861–868 (20) Chug-Ahuja, J.K., Holden, J.M., (6) Vanderveen, J.E., & Pennington, Forman, M.R., Mangels, A.R., J.A.T. (1983) Food Nutr. Bull. 5, Beecher, G.R., & Lanza, E. (1993) 40–45 J. Am. Diet. Assoc. 93, 318–323 (7) Holden, J.M., Schubert, A., Wolf, (21) Schneeman, B.O., & Gallaher, D.D. W.R., & Beecher, G.R. (1987) in (1990) in Present Knowledge in Food Composition Data: A User's Nutrition, 6th Ed., International Life Eurofoods-Enfant Project Second Sciences Institute, Washington, DC Annual Meeting, Wageningen (22) Li, B.W., Holden, J.M., Brownlee, Agricultural University, Wageningen S.G., & Korth, S.G. (1987) J. Am. (31) Honikel, K.O. (1994) Report, FLAIR Diet. Assoc. 87, 740–743 Eurofoods-Enfant Project Third (23) Hepburn, F.N. (1988) Proceedings Annual Meeting, Wageningen of the 12th National Nutrient Data Agricultural University, Wageningen Bank Conference, The CBord (32) Holden, J.M., Gebhardt, S., Davis, Group, Inc., Ithaca, NY, pp. 31–33 C.S., & Lurie, D.G. (1991) J. Food (24) Schubert, A., Holden, J.M., & Wolf, Comp. Anal. 4, 183–195 W.R. (1987) J. Am. Diet. Assoc. 87, (33) Progressive Grocer's Market Scope 285–299 (1993) Progressive Grocer's Trade (25) Tonucci, L.H., Holden, J.M., Dimension Division, Maclean, Beecher, G.R., Khachik, F., Davis, Hunter Media, Inc., Stamford, CT, C.S., & Mulokozi, G. (1995) J. pp. 18, 348–549 Agric. Food Chem. (in press) (34) Holden, J.M., Lanza, E., & Wolf, (26) Micozzi, M.S., Brown, E.D., W.R. (1986) J. Agric. Food Chem. Edwards, B.K., Bieri, J.G., Taylor, 34, 302–308 P.R., Khachik, F., Beecher, G.R., & (35) Knutson, J. (1989) Meatfacts 88, Smith, J.C. (1992) Am. J. Clin. Nutr. American Meat Institute, 55, 1120–1125 Washington, DC, p. 17 (27) Savell, J.W., Harris, J.J., Cross, (36) Gebhardt, S.E., & Holden, J.M. D.S. Hale, D.S., & Beasley, L.C. (1992) Provisional Table on the (1991) J. Anim. Sci. 69, 2883–2893 Selenium Content of Foods, USDA, (28) Buege, D., Held, J.E., Smith, C.A., Washington, DC Sather, L.K., & Klatt, L.V. (1990) (37) Sokal, R.R., & Rohlf, F.J. (1981) Research Bulletin R-3509, College Biometry, 2nd Ed., W.H. Freeman of Agriculture and Life Sciences, and Company, San Francisco, CA University of Wisconsin, Madison, (38) Beaton, G.H., Milner, J., Corey, P., WI McGuire, V., Cousins, M., Stewart, (29) McCann, A., Pennington, J.A.T., E., de Ramos, M., Hewitt, D., Smith, E.C., Holden, J.M., Soergel, Grambsch, P.V., Hassim, N., & D., & Wiley, R.C. (1988) J. Am. Little, J.A. (1979) Am. J. Clin. Nutr. Diet. Assoc. 88, 336–341 32, 2546–2559 (30) Kohlmeier, L., & Poortvliet, E. (1992) Report of the FLAIR Quality Control of Food Composition Data and Databases

Assuring Regional Data Quality in the Food Composition Program in China

Guangya Wang, Xiaolin Li

Institute of Nutrition and Food Hygiene, Chinese Academy of Preventive Medicine, 29 Nan Wei Road, Beijing 100050, China

A nationwide collaborative project on the analysis of food composition for China was organized and conducted by the Institute of Nutrition and Food Hygiene between 1987 and 1990. In order to assure the quality of analytical data from all 20 participating laboratories, a quality assurance system was conducted involving five procedures: a written manual of analytical methods; technical training courses for laboratory technicians; the use of identical methodological protocols for sampling and handling of food samples; analytical duplicate or replicates for unknown samples; standard reference materials and quality control materials. The results were monitored by means of a control chart to check the reliability of technical performance. Data were evaluated by logic and statistical tests and then compiled into new Chinese food composition tables. The total number of food items is 1358, comprising 3280 separate food samples.

nationwide collaborative project to revise and update the food composition data of China was organized and conducted by the Institute of Nutrition and Food Hygiene A(INFH) in 1987–1990. In order to assure the quality of the analytical data provided by each of the 20 participating laboratories, an analytical quality assurance system was designed and carried out. The data obtained in this project were the basis of the new edition of the Chinese food composition tables (FCT) published in 1991.

1990 (1). A new edition of the food Ŷ Background composition tables has been needed The first edition of the Chinese FCT since the early 1980s, because food published by INFH in 1952 included only composition may have changed, due to 12 nutrients, crude fiber and energy the changes in crop cultivation and value for about 300 food items. In the animal husbandry as well as food following years, INFH updated the FCT storage, transportation and marketing with three editions having been during the recent decades; also newer published. The last printing was in 1981 and better analytical methods are now and its English version was published in available; and data on a number of important micronutrients (vitamins, trace Ŷ Working Procedures elements) were missing from the The following system was introduced to previous editions. In this project, both assure the quality of data generated by the nutrients and food items were all the participating laboratories. increased. The food items were selected based on the knowledge of frequency Validation of Analytical Methods and amount of food consumption A written manual of analytical methods obtained from several national dietary including food sampling and handling surveys, and newer methods were used was prepared by INFH to ensure in the laboratory analyses. All the laboratories adhered to the same nutrients were analyzed by AOAC methods. All the methods were methods (2) and official Chinese validated for accuracy and precision methods (3, 4). The analytical data were according to published guidelines (5, 6). categorized as follows: proximate Each analytical method was evaluated composition (moisture, energy, protein, by three to six selected laboratories fat, carbohydrate, dietary fiber and ash); using standard reference materials vitamins (ascorbic acid, thiamin, (SRMs), i.e. bovine liver (National riboflavin, niacin, retinol, carotenes, and Bureau of Standard, USA), bread tocopherols); minerals (calcium, iron, crumbs (a gift from Dr. Harry G. Lento, magnesium, phosphorus, potassium, Campbell Institute for Research sodium, zinc, copper, manganese and Technology, USA), and pig liver (China selenium); lipids (fatty acids and National Standard Bureau), as well as cholesterol); and, amino acids. Foods quality control materials (QCMs) items were divided into 28 groups prepared by the central laboratory in including cereals, dried legumes, fresh INFH, i.e. wheat flour, whole milk and sprouted legumes, roots, tubers and powder and carrot paste. Accuracy and stems, fresh leafy vegetables, melons, precision of analytical methods between squashes and gourds, fruit-bearing laboratories were determined daily to vegetables, pickled, salted and validate the methods. The three QCMs preserved vegetables, fungi and algae, were used to measure the level of fruits, nuts and seeds, meats, poultry, precision and recovery. The detectability milk and milk products, infant foods, and correlation coefficient of standard eggs, fish, molluscs and crustaceans, curves were used as additional indices fats and oils, confections and snacks, for method validation. tea and beverages, alcoholic beverages, sugars and sweets, starch and its Training Courses for Participating products, condiments and spices, edible Laboratories Chinese medicinal herbs, and In order to assure that the analytical miscellaneous items. The total number procedures would be carried out of food items analyzed in this project correctly and consistently by all the was 1358. Food composition analysis participating laboratories, several was performed by 20 laboratories technical training courses were located in 15 provinces. Among them, organized by INFH. The first training there were 11 provincial and five course was conducted in 1986 and municipal Institutes of Food Safety attended by more than 50 technicians Inspection, three provincial Medical from the 20 laboratories. The methods Institutes and one provincial Medical for determination of six vitamins, amino College. These provinces and acids, fatty acids, dietary fiber and municipalities covered half of the areas selenium were demonstrated by of China and more than two-thirds of the instructors and then practiced by the total Chinese population (Figure 1). trainees in the training laboratories. The second course was conducted in 1987. technologies in food nutrient analysis. Two specialists in food analysis, Dr During 1988 to 1989 secondary training Gary R. Beecher and Dr Joseph T. courses were organized at the local Vanderslice from USDA, were invited to level to train more technicians with the give lectures and to introduce new instructors from INFH.

Figure 1. Outline map of China showing provinces included in food composition program (shaded)

Sampling and Handling of Food of each piece was over 500 g, three Samples pieces were collected. The same variety of food was collected in three places It was critical to ensure that identical located in an urban district and/or protocols of sampling and handling of county area. After the food was foods for analysis were used in each homogenized, one-third of each participating laboratory in order to homogenate was pooled into one eliminate both intrinsic and extrinsic analytical sample. The analyses for sources of variation which could affect vitamins were carried out as soon as the the measured levels of nutrients in foods were collected. A set protocol for foods. The sampling scheme was homogenization, temperature control designed to reflect representativeness and other aspects of sample preparation of the food with regard to the brand or was followed. cultivar and geographic origin of the food as well as the differences in food consumption in the different areas. Interlaboratory Quality Control Foods were collected according to the priority of quantity consumed. The Any interlaboratory variation will affect sample size for each collection was 1.5 the variability of the compiled data. The kg by weight or by pieces. If the weight values produced by each laboratory were evaluated by using quality control The analytical data for foods from each materials (QCMs). The maximum laboratory were evaluated with respect acceptable relative standard deviation to their reliability. Some statistical tests (RSD or CV) was between 5 per cent were used such as the Dixon and and 10 per cent. The coefficient of Grubbs test to reveal the outlier values. correlation of the regression curve for The t test was used to determine each standard curve of an analytical whether data were significantly different, method should ideally be 0.999. and the F test was used to determine Recovery tests of fortified QCMs were whether the variances of the data were used as an index of accuracy. different (8, 9). Validated values were Recoveries between 90–110 per cent compiled into the new FCT. were defined as satisfactory. The Ŷ Results and Discussion analyzed mean value was expected to fall within plus or minus one standard SRMs are ideal tools for analytical deviation of the certified value. For the quality control, but they are too QCMs used in this program a mean expensive to be used throughout an certified value and standard deviation entire project. Therefore, QCMs was determined by six of the selected prepared by INFH were used by each laboratories. In general, values within laboratory. Whole milk powder and two standard deviations from the mean wheat flour were easy to obtain in large were acceptable. Data of QCM analyses amounts and very homogeneous, so produced by participating laboratories they were suitable for use as QCMs. On were evaluated using the Youden pairs the other hand, carrot paste proved to method (8, 9) to test whether the value be difficult to stabilize and was readily fell within the 95 per cent confidence spoiled during transportation and interval. The outlier data were examined storage. Results from carrot paste in order to identify problems. Sample showed large variations, and are not exchanges, replicate analyses, included in this paper. The certified calculation checks and further training of values of the QCMs were the mean technicians in INFH were carried out to values calculated from the individual improve analytical accuracy and values from six laboratories. Each precision. For unknown samples the individual value produced by a results of duplicate analyses had to be laboratory was the mean value within 10 per cent of their mean. calculated from six duplicate Otherwise, a third or further replication determinations on different days. These was required to re-determine the mean. six selected laboratories passed the quality control test. The certified nutrient Assessment of Analytical Data values of the QCMs for wheat flour and Reported from Different Laboratories whole milk powder are shown in Table I. Table I. Certified nutrient values of quality control materials per 100 g (mean±SD) Whole Milk Powder Wheat flour Moisture (g) 3.1±0.2 12.0±0.5 Protein (g) 24.8±0.8 11.7±0.6 Fat (g) 27.2±3.0 1.6±0.2 Ash (g) 5.8±0.1 0.84±0.02 Dietary fiber (g) - 2.4±0.2 Thiamin (mg) 0.18±0.05 0.35±0.02 Riboflavin (mg) 0.90±0.11 0.08±0.02 Niacin (mg) 0.80±0.11 2.12±0.19 Retinol (µg) 135±43 - Vitamin E (mg) 0.44±0.02 1.56±0.39 K (mg) 1010±115 202±17 Na (mg) 350.1±12.5 1.3±0.1 Ca (mg) 847±115 14.0±0.7 Mg (mg) 107±9 69±6 Fe (mg) 0.6±0.2 1.9±0.4 Zn (mg) 3.53±0.48 1.57±0.13 Cu (mg) 0.06±0.01 0.27±0.04 Mn (mg) 0.07±0.02 1.92±0.12 Se (µg) 8.90±0.73 28.7±1.10 P (mg) 770±45 195±16 - = not applicable Bovine liver, bread crumbs and pig around 2–7 per cent and the CV of liver SRMs were used to validate the reproducibility between each laboratory analytical methods. The accuracy and was larger (Table III). The CV for percent recovery of analyses were used proximate analyses (not shown in Table as indices for evaluation. For example, III) was between 1 per cent to 8 per the accuracy of the fluorometric method cent, but for vitamin analyses there were for selenium (Se) analysis is shown in larger variations. In general, the Table II. The reported mean values methods for vitamin determination had were close to the certified values and somewhat lower precision than mineral the coefficients of variation (CV) of the and proximate analyses. The above analytical values were between 2.7 per results showed that all the methods cent and 6.3 per cent. The percent were satisfactory. In order to monitor recoveries of the analysis were between analytical performance, the data for 95.2–99. 1 per cent and the CVs of the nutrient analyses of QCMs were results were between 2.8 per cent and collected and evaluated using three 6.2 per cent. The data in Table II statistical tests. A simple method was indicate that the fluorometric method the control chart test (10). All the QCMs was a valid method for Se. data from each laboratory were plotted Other indices for evaluating analytical on the control chart. The certified value methods were also used and the results (X) of a given nutrient was assigned as are shown in Table III. Using the data the central line (CL), the mean value collected from the selected laboratories, plus or minus one standard deviation (S) all the methods were evaluated. The as the upper and lower auxiliary lines recoveries of these methods ranged (XS), respectively, the mean 2S as the from 87 per cent to 110 per cent, most upper and lower warning limit line, and of them being in the range 90 per cent the mean 3S as the control or to 110 per cent. The analytical precision confidence limits. Because the QCMs of the methods shows that the CV of are biological materials and are repeatability within each laboratory was unstable, their composition could change with time and be affected by acceptable limit. The percentages of factors such as oxidation, temperature acceptable values from the participating and light etc. So we preferred to use the laboratories are shown in Table IV and mean value 3S as the largest Figures 2 and 3.

Table II. Accuracy of the fluorometric method for determination of selenium Standard reference material Certified value Reported mean value Reported recovery (X±S,µg/g) (X±S,µg/g) CV% (X±S,%) CV% Bovine Liver 1577a 1.1±0.1 1.04±0.03(11) 2.7 Pig Liver 0.940±0.028 0.960±0.028(8) 2.9 Milk Powder 0.089±0.007 0.094±0.003(6) 3.1 99.1±2.8 2.8 Wheat Flour 0.287±0.011 0.298±0.019(6) 6.3 95.2±4.6 5.0 Rice Flour 0.083±0.007 0.082±0.005(6) 6.0 95.9±5.9 6.2 Numbers in parentheses are the total number of determinations

Table III. Indices and results for methods validation Nutrientananlysed Recovery RepeatabilityReproducibility Linearity Limit of Method of std detection curve % CV% CV% CV% (r) Atomic Ca 93.7–108.33.8–5.1 2.0–5.7 1.1–5.7 0.9996 0.1µg/ml absorption Fc 95.0–108.53.4–5.3 5.2–7.2 4.6–11.0 0.9996 0.2µg/ml spectometry Mg 94.9–105.12.8–4.7 3.8–7.0 1.8–7.9 0.9998 0.05µg/ml Mn 94.1–109.04.5–6.1 6.4–9.6 2.4–8.2 0.9991 0.01µg/ml Flame K 97.9–104.82.6–2.8 1.4–2.8 0.3–10.4 0.9998 0.05µg/ml photometry Na 96.4–103.82.4–3.1 2.6–5.1 2.0–5.8 0.9997 0.3µg/ml Microbiology Riboflavin 98.3–110.02.6–2.8 2.2–5.2 11.3–11.9 NA 0.05µg Niacin 93.6–110.03.3–4.7 2.4–4.0 6.4–8.3 NA 0.05µg Paper Total carotenes 88.8–102.9 5.2 1.7 5.9 0.9996 0.1µg chromatography Fluorometry Thiamin 91–100 7.4 6.8–8.5 17–22 0.9993 0.05µg Riboflavin 92–109 3.2–6.2 4.5–5.7 11.7–15.1 0.9998 0.002µg Ascorbic acid 99.5–107.1 6.0 2.7–7.3 - 0.9996 0.022µg Selenium 87.4–104.52.8–6.1 3.1–6.2 9.6–10.2 0.9998 3 ng LC Retinol 92–105 9.0 11.0 5.7–8.8 0.9981 0.04µg/µl Tocopherol, Į 92–105 9.0 10.0 5.4–8.7 0.9996 4.59ng/µl Ȗ+ ȕ 97–108 3.6 13.0 3.1–7.4 0.9918 1.83ng/µl ȗ 87–107 4.1 11.1 11.1 0.9910 1.03ng/µl Spectrometry Phosphorus 94.9–105.44.4–4.8 2.1–6.4 1.1–6.4 0.9999 1.5µg/µl Gravimetry NDF NA NA 2.5–7.6 3.8–15.5 NA 1.1mg NDF = Neutral detergent fiber - = Not determined NA = Not applicable Table IV. Percentages of acceptable values from participating laboratories Nutrient Wheat flour Whole milk No. of labs No. of labs Acceptability No. of labs No. of labs Acceptability accepted % accepted % Moisture 17 16 94.1 16 13 81.2 Protein 17 17 100 18 17 94.4 Fat 16 16 100 19 19 100 NDF 16 15 93.8 - - - Ash 15 13 86.8 16 14 87.5 K 18 17 94.4 16 15 93.8 Na 16 14 87.5 17 16 94.1 Ca 17 10 58.8 16 15 93.8 Mg 17 15 88.2 16 15 93.8 Fe 18 18 100 17 17 100 Zn 18 17 94.4 17 16 94.1 Cu 17 17 100 17 15 88.2 Mn 16 15 93.8 16 16 100 P 17 17 100 16 15 93.8 Se 14 13 92.8 12 11 91.7 Thiamin 16 8 50.0 16 16 100 Riboflavin 15 15 100 17 15 88.2 Niacin 12 12 100 13 13 100 Retinol -- -12 12 100 Tocopherol 12 12 100 - - - NDF = Neutral detergent fibre - = Not applicable

Some laboratories failed to submit identified by means of replicating the the results to INFH in time and their data analyses, making new standard curves were not included in Table IV. According and checking the calculations. To to the results in Table IV, 87–100 per calculate the representative value for cent of the laboratories passed the each analyte in each food item, two quality control tests, except that around standard deviations from the mean half of the laboratories failed in the value after deleting the suspect data determination of thiamin and calcium. were used to eliminate the values Most of the calcium values of wheat outside the range limits and then the flour were much higher than the certified mean value was recalculated. This values. The errors came from technical mean value was used for the food mistakes such as not adding the 8- composition table. Some results hydroxyquinoline to eliminate considered as unreasonable were interference from reagents. Some checked for the causes. In some cases, thiamin values of wheat flour were re-analysis of foods was carried out higher and some were lower than the through exchanges with other central line. Problems included low laboratories or IFNH. Some recovery after column filtration or unreasonable data which could not be interfering substances from reagents. validated were eliminated during data Most of the niacin and protein values of compilation. In practice, some values both wheat flour and milk powder fell were difficult to judge based on the within UWL and LWL (X2S) (shown in current knowledge of food and nutrition, Figures 1 and 2). The over-range data and were, therefore, retained in the were questioned and the problems FCT. Figures 2 and 3. Examples of a quality control chart for two QCMs with different mean values of protein obtained from collaborative laboratories methods, the availability and use of Ŷ Conclusion reference materials and the training of Quality control is costly and time the technicians. Large variations existed consuming, but it is essential. We have in the conditions of the collaborating conducted an efficient analytical quality laboratories as well as in the technical assurance system in a nationwide background of the technicians. There project of food composition analysis of were some inadequacies in this 1358 food items, and involving 20 approach, for example, analytical collaborating laboratories. According to methods for minerals were not included our experience, the critical parts in this in training courses, except for selenium, analytical quality assurance system and a few technicians were not familiar were the validation of analytical with the LC and GLC techniques. The question of how to ensure the People's Medical Publishing House, comparability of the Chinese food Beijing composition data with those of other (5) Uriano, G.A., & Cali, J.P. (1977) in nations is still an unresolved problem. Role of Reference Materials and Reference Methods in the Ŷ Acknowledgments Measurement Process, J.R. DeVoe This project was supported by National (Ed), ACS Symposium Series 63, Science Foundation of China and American Chemical Society, Ministry of Public Health and Jia Li Bao Washington, DC, Chap. 4 company. (6) Holden J.M., Schubert A., Wolf, Ŷ References W.R., & Beecher, G.R. (1987) Food Nutr. Bull., Suppl. 12, 177– 193 (1) Ershow, A.G., & Wang, Chen, K. (7) People's Republic of China (1990) J. Food Comp. Anal. 3, 191– Standard G.B. 6379–86 (1986) 442 Precision of Test Methods for (2) Official Methods of Analysis (1984) Determination of Repeatability and 14th Ed., AOAC, Arlington. VA, Reproducibility for a Standard Test secs 14.002–14.004, 31.005– Method by Interlaboratory Tests (in 31.008, 7.009, 43.275– 43.277, Chinese), Chinese Standard 24.037–24.040, 7.093– 7.103, Publishing House, Beijing 43.024–43.038, 43.069– 43.081 (8) Pan, X.R. (1989) Assurance and (3) People's Republic of China Evaluation of the Accuracy of Standard GB 12388-12399-90 Chemical Analysis (in Chinese), (1990) Methods for Determination Chinese Measurement Publishing of Nutrient Composition in Foods (in House, Beijing Chinese), Chinese Standard (9) Gerrit, K., & Frans, W.P. (1981) Publishing House, Beijing Quality Control in Analytical (4) Institute of Nutrition and Food Chemistry, John Wiley & Sons Inc., Hygiene (1990) Methods of Food New York, NY Analysis, 3rd Ed. (in Chinese), Quality Control of Food Composition Data and Databases

Quality Control for Food Composition Data in Journals — A Primer

Kent K. Stewart, Margaret R. Stewart

Virginia Polytechnic Institute & State University, Blacksburg, VA, 24061-0308, USA

Scientific journals are a primary vehicle for the transmission of original food composition data and critical reviews of food composition data to the scientific community. Publication of composition data in a scientific journal implies that the data are accurate, precise, and meaningful. To publish data meeting these attributes it is necessary to establish criteria for data quality control. Quality control is achieved by critical evaluation of all aspects of a scientific manuscript by expert reviewers. The key attribute of a good quality control in a manuscript is adequate documentation. In a good publication, those items that should be documented include the purpose of the study; description of the sampling plan for selection of the food items to be assayed; descriptions of the food items; descriptions of the sample preparation, homogenization, and storage; description of the analyte extraction; descriptions of the identification and measurement of the analyte; and description of the analytical quality control measures used to validate the data sets. Reviewers also evaluate the quantitative data including their statistical components and the discussion of how the new data relate to existing knowledge on the composition of foods.

t is almost an article of faith in the scientific community that “good” data will aid in the development of wise decisions and that “bad” data will lead to the development of Iunwise decisions. In cases of conflicting data, the perceptions of which are good data and which are bad data may well be as important as the actual fact of the quality of the data. These are not just issues of academic concern to those working on food composition data. The current public concerns about the impact of diet on health will inevitably lead to the promulgation of new policies and regulations on the composition of foods. The public in many countries is concerned about the possibility of inadequate intakes of essential nutrients, problems related to inadequate or excessive intake of energy, the possibility of intakes of toxic levels of man-made chemicals such as pesticides and herbicides, and the perceived dangers of the use of biotechnology in the production and processing of the food supply. Given the current level of knowledge on the composition of foods, a great deal of new food composition data will be needed if wise policies and regulations on the issues of diet and public health are to be made. The discussions in this paper about Once the reviewers ascertain that the quality control for food composition data documentation is present, then they in scientific journals are extensions of should determine that the appropriate opinions from editorials originally techniques were used for the acquisition published in the Journal of Food of the food composition data. A review Composition and Analysis (1–10). of the appropriateness of various Scientific journals are a primary vehicle techniques for food composition data for the transmission of original food acquisition is very complex and requires composition data and critical reviews of a great deal of technical discussion food composition data to the scientific beyond the scope of this paper. Finally, community. Publication of composition it is the reviewer's responsibility to data in a scientific journal specifically determine whether or not the implies that the data and their attributes composition data are accurate, precise, have been evaluated and reviewed prior novel, and credible. These are primarily to publication by independent experts in issues of quality control and are the the field. The responsibilities of main topic of this paper. publishing credible, good quality The editor's primary responsibility is composition data are spread among the to ensure to the readers of a journal that authors of the manuscripts, the the data published therein are accurate, reviewers of the manuscripts, and the precise, and meaningful. The goal of a editors of the journals. It is the authors' journal is to have the data review and responsibility to carry out the study publication done in an authoritative properly and then to provide adequate manner so that the burden of proof will documentation on how the study was be on those who challenge the done. published assay data of the composition While the editor selects the reviewers of specific foods. Thus in many ways, and ensures that conflicts between while authors are the source of scientific authors and reviewers are resolved, it is knowledge, scientific journals can be the reviewers who are the key to quality viewed as “gates” for transmission of control of journal articles through critical knowledge, and the reviewers and evaluation of all aspects of a scientific editors can be viewed as “gatekeepers”. manuscript. Given the chemical The primary focus of this paper is complexity of foods and their matrices, documentation needed in a food the enormous size of the food composition paper. The underlying distribution system, and the frequent theme is that adequate documentation technical complexity and difficulty of is required for good quality control of modern analytical assay techniques, published food composition data. reviewers of food composition data need Ŷ Documentation Needed in special expertise as well as significant Manuscripts knowledge of a broad range of subjects. The reviewers' first responsibility is to The documentation for a food determine whether or not adequate composition data manuscript includes documentation (a key aspect of food the sampling plan, a description of the composition data quality control) was foods and the laboratory samples, a provided in the manuscript reporting the description of the assay methodologies, study. Without adequate documentation the actual composition data, the quality there can be no critical evaluation of the control information for those data, and a science, and its lack is a fundamental comparison of the data presented with failure of manuscript quality control and, those in the published scientific if not rectified, should ultimately result in literature. rejection of the manuscript. Sampling Plans documented. In most cases the total A primary goal in the analysis of food number of analyses to be performed is samples is the description of the nutrient strongly influenced by economics and content of the foods that are each laboratory sample should be encountered in the real world. This assayed a minimum number of times. It seemingly trivial and almost tautological might seem that a general statement is unfortunately not often implementation of this design strategy followed to its logical conclusion: that would be to assay each laboratory the design of the analytic protocol sample only once. However, in general, (especially the choice of food items and the need for protection against major the number of analyses run) should be blunders in the assay of an individual directed towards gathering as much sample leads to the suggestion that information as possible about the each sample be assayed in duplicate. distribution of the nutrients in food in the Food Descriptions real world. Users of food composition There are a very large number of foods data need information about the in international and domestic market average or “usual” level of that nutrient places. Different species are used as and the range of values that would likely food sources; various growing be encountered. conditions are used; the processing, A key issue is thus which food item packaging, and storage technologies should be assayed? Another way to put vary. Cultural differences in food recipes this is the question “Were the assayed are common. At the same time, there is items representative of the foods for a great commonality within some which composition information is foodstuffs due to the worldwide presented?” It should be intuitively availability of some brand name items obvious that representative composition (e.g., soft drinks and fast foods). Given data can only be obtained from the the complexity of the world food supply, assay of representative lots. Thus the the readers of papers on the determination of which lots to assay composition of foods need to be given may well be the most important of all the the information to identify the foods for questions facing the analyst in the which the composition data are being design of a food composition assay presented. For even with the best of program. The goal of a good sampling analytical techniques, food composition plan should be to have a protocol which data are no better than the description indicates how many lots should be of the products or foods being analyzed. sampled and when and where they The analyst should describe the foods should be obtained, and which provides so that another professional in the field other details on how individual food can identify the foods. The sources and items should be selected. unique descriptions of the foods should Once a good sampling plan is be given. Those foods which were selected and documented, then the enriched and/or fortified should be so means used for distinguishing the identified. Identification of market share sources of variability in analytical data, can be useful. The analysts should i.e., those arising from inherent biologic specify numbers of items collected and variability that reflect differences in the number of units in a composite. The genotype, phenotype, environment, dates of food acquisition should always processing, etc., and those arising from be given. Frequently, most of the analytic variability, introduced in the needed description of manufactured process of preparation and assay of the foods can be provided by identifying the laboratory sample, should be brand name and place and date of purchase. The post-selection non-enzymatic oxidation, to various transportation and storage of the food enzymatic actions and to other items should be described. Any further destructive reactions. Many fractionation of samples such as homogenizing techniques do not yield trimming and draining should also be homogenous material with mixed foods described. When the food is cooked, the and diets. In such cases, representative cooking processes should be described. sub-sampling is difficult and the References to published cooking precision of the results deteriorates. procedures should be given whenever Much more work is needed on the possible. techniques for validation of the Compositing and Homogenizing appropriateness of homogenizing techniques. Compositing is the process of preparing Even when many samples are used a single representative composite to make a composite, once they have sample from several food samples. been composited, the analyst has only Homogenization is the process of one test sample. Thus a primary feature reducing a food sample to equally of compositing is that all information on distributed particles of uniform size. the variation between lots is lost once Compositing and homogenization are the individual samples are composited. the invisible components of food assay There are several other issues that need systems. Homogenization is almost resolution when doing compositing. For always necessary to transform the large example, given the purposes of the bulk of heterogeneous foods and diets assays: How many units should be used to homogenous representative material in a composite? How much information suitable for sub-sampling. The process on real variance is lost when a given has been described as transforming a 5 compositing procedure is used? pound meat roast into an analytic sample which can be introduced by a 5 Assay Methods mL syringe into a chromatograph. The A generalized diagram of an ideal food procedures used for composition assay system is shown in compositing/homogenizing can have a Figure 1. The key feature of this schema significant impact on the accuracy and is that a food composition assay should precision of the final results. The issues be viewed as a whole. Each part of the inherent in compositing and assay system must fit with the rest of homogenizing are crucial to the the assay system. Inappropriate use of production of accurate, representative, any one technique can invalidate the and precise food composition data. It is accuracy and/or precision of the entire important that adequate documentation assay. Basically what is needed is a of these processes be given in food holistic approach to food composition composition papers. assays. In these homogenization processes there is potential for analyte loss due to Figure 1. An ideal assay system

Published Assay Methods When new food assay methodologies are presented, it is important that these In most cases published assay methods be validated for their use in methodologies are used for food obtaining food composition data. What composition assays. The authors should follows is a general description of what provide complete references to is necessary to validate the use of a published assays. In those cases where new assay methodology for a given methods manuals are used, e.g., an matrix. The underlying premise is that AOAC methods manual, the edition of the validation of an assay method is a the book and the assay number should process by which the assay method is be given. Some indication should be demonstrated to be capable of given as to how the selected methods producing the desired analytical results were determined to be appropriate for when used with the matrix of interest the assays at hand. Most published (i.e., assay methodology validations are methods are not appropriate for the matrix specific). Usually such a assay of every food matrix. If there are validation is done in some authoritative known potential interferences in a given manner so that the burden of proof will matrix, then the authors should use the be on those who challenge the assay method of standard additions to verify method or the data from such a method. that the assay is appropriate for the food For most assay methods, the desired matrix. Failure to demonstrate results include acceptable accuracy, quantitative recovery should raise precision, and sensitivity. The definition serious concerns of the appropriateness of acceptable accuracy, precision, and of the choice of assay. sensitivity of an assay is a function of New Assay Methods the end use of the assay results. Different end uses will change the to produce the same results with the perception of what is acceptable. The method following criteria are presented as an ƒthe individual data set validation idealized list for assay validation. While processes should be made an it is understood that not every criterion integral part of any method will be met in every new method description. validation, each individual criterion should be considered when doing new Analyte Extraction assay method validations. The idealized Recent research studies in analytical assay validation criteria are: chemistry have focused on the ƒthe entire assay method must be development of new instrumentation for consistent with the chemical the separation and measurement of properties of the analyte and the analytes. Similar significant advances matrix have been made in analytical biochemistry and molecular biology in it should be demonstrated that there ƒ the development of highly selective and are no obvious matrix interferences sensitive probes such as immune- for any stage of the assay method, reagents and DNA and RNA probes. or that subsequent or previous Significant improvements have been stages have eliminated the made in the use of enzymes as interferences reagents. Almost all the advanced ƒthe method should give quantitative techniques described above require recovery of pure standards carried clean extracts free of interfering through total assay method compounds. For the most part, it is best ƒthe method should give quantitative if the analytes are dissolved in solutions recovery of analytes to matrices of which are themselves compatible with concern the separation and measurement components of the assays. These the method should give acceptable ƒ advanced techniques are perfect for the results on composition of standard assay of pure standards or the assay of reference materials of matrices mixtures of pure standards. similar to the foods to be assayed Unfortunately, most foods do not come ƒthe method should give acceptable in tidy packages free of assay precision of replicate assays (five interferences. Rather, most foods are replicates) of the analyte in matrices complex mixtures of multi-phase of concern materials with extremely complex ƒthe limit of detection (LOD) for chemical matrices. Analyte levels are analyte determination in matrices often low and assay interferences are should be defined and acceptable common. If analysts are to properly use today's marvelous array of analytical the linear range for analyte ƒ tools for the assay of most food determination in matrices should be components, they first should isolate the defined and acceptable analyte from the food matrix. ƒthe method should give the same In an ideal extraction procedure the results as a validated (or accepted) analyte is quantitatively removed from method or there should be an the food, no analyte remains with the adequate explanation for the residue, and no analyte is altered by the observed differences extraction procedure or by the inherent ƒmore than one analyst and more biochemical and chemical activities of than one laboratory should be able the matrix. The extract should not contain compounds that would interfere with the separation and measurement presenting the necessary components of the assays. For documentation for system evaluation. example, if chromatographic Chromatography systems have separations are used, the extract should several common problems including not contain components which coelute drift, difficulty in confirming peak with the analytes or those which alter identifications, and the difficulty of the chromatographic behavior of the obtaining reproducible sample analytes. If immune reactions and/or injections. The use of internal standards enzyme measurement systems are helps to reduce the problems of drift and used, components which alter those sample injections. The use of internal enzymatic and/or immune reactions are standards is now considered to be unacceptable. Given the complexity and almost mandatory and the lack of their variability of food matrices, quality use is considered to be a serious flaw in control procedures for the extraction the methodology and often leads to steps should be required parts of most rejection of a manuscript. Peak assays. Certainly, the analyte isolation identification can be difficult in some procedures should be carefully food matrices and care should be taken documented and critically reviewed to document the proper identification of before they are used or published in the analyte peaks convincingly. reputable scientific journals. Analyte Identification, and Analytical Separation of the Analyte Quantification The current state of the art in analytical Today's analyst has an enormous array separation techniques such as gas of detection techniques for analyte liquid chromatography (GLC), and liquid quantification including atomic chromatography (LC), super-critical fluid absorption and plasma emission chromatography, and capillary spectrophotometry, mass spectrometry, electrophoresis is impressive. These diode array spectrophotometry, various sophisticated separation tools have the electrochemical detectors, fluorescence capability of separating very complex detectors, the highly selective and mixtures in relatively short time periods sensitive probes such as immune- and as such each of these techniques reagents, DNA and RNA probes, can be very useful for the food analyst. various chemical detection systems, the Accompanying the power and use of enzymes as reagents and of sophistication of these techniques is enzyme amplifier systems, and enzyme- their complexity. Reproducing assays linkedimmunosorbant-assays. Many of using these techniques requires detailed these assay systems have been information on the entire analytical automated through the use of system including the manufacturer and continuous flow systems, flow injection model of the instrumentation, the systems, and robotics. Recently, there column used, the solvents or carrier have been significant advances in the gases used, the flow rates of solvents or use of hyphenated techniques such as carrier gases, and the temperatures GLC-mass spectrometry or other used at the injection port, the column combinations such as enzymelinked- oven, and the detector. Adequate immunosorbant-assays (ELISA) evaluation of any given analytical systems using electrochemical detection separation system requires extensive automated through the use of flow documentation of the system. Each sub- injection analysis. All of these systems discipline in analytical separations has are powerful and the analyst has an developed its own shorthand mode of impressive array of quantification tools to draw upon. However, as in the case of analytical separations, the with no more than three significant sophistication is usually accompanied digits. by significant increases in complexity. The moisture contents of individual Reproducing given assays using these foods are highly variable and thus most techniques requires detailed information composition data should be presented on the entire analytical system. The on a dry weight basis. The composition analyst needs to provide significant data for beverages are obvious detailed documentation on the exceptions. Many believe that data quantification system. Many detection presented on a dry weight basis should systems do not provide unique be accompanied by a moisture value to identification; verification of analyte enable calculation to “as consumed” identification is often necessary. Even basis. the highly selective assay systems such Data Set Validation as immune-reagents and enzymes do not give totally unique identifications and It is our observation as editors, that are often quite sensitive to interferences mistakes in composition data are a in the quantification reactions. relatively frequent occurrence. Their Verification of the appropriate use of frequency should be substantially these systems is usually necessary. reduced. The challenges in the production of good data are that while Computation of Results there are a very large number of useful Unfortunately, computational errors are assays, their implementation is often still one of the most common sources of complex and mistakes are relatively errors in food composition data (or any easy to make. Even experts can get other assays for that matter). The incorrect results and generate incorrect introduction of computerized data when using “good” methods. computation systems, such as Therefore, to produce good data that spreadsheets and black box analytical are credible, the food composition instrumentation, has not alleviated these community should develop protocols for problems. Thus it remains important for data set validation. Given the complexity analysts to check the process by which of the problems of assay and the wide they do their computations. Since variety of methodologies which are computational errors are still so available today, we believe that the common, it is prudent for authors to validation of individual data sets is document their quality control necessary and that all food composition procedures for the computation of the data sets should be individually analytical results. validated. The concept that each Composition Data individual data set be validated specifically implies that some type of Authors should report the means and quality control sample was assayed standard deviation of the composition along with the samples that were values and the number of lots assayed. assayed. Furthermore, it also implies Replicate assays on one test sample or that the data set results underwent an one composite yield a single value and internal quality assurance check prior to as such are usually not sufficient for the acceptance of the results. There are journal publication. The significant digit many ways to validate data sets convention (reporting of only all digits including the use of common sense- known with certainty and the first digit of consistency observation, standard uncertainty) should be used in reporting laboratories, standard instruments, all data. Most food assays yield results certified analysts, certified algorithms, standard reference materials, internal standards, audit trails, and in-house Significant efforts by journal editors, reference samples, (i.e. pool samples). reviewers and authors are needed if we The choice of the data validation are to improve the comparability of data procedure depends upon the laboratory, between studies. the food samples, and the component More composition data need to be being measured. The addition of the published in refereed journals. The concept of validated individual data sets current practice of directly publishing the will be of significant help in efforts to results of food composition studies in provide “good” food composition data databases rather than referred journals that are also perceived to be “good means that the documentation behind data”. those new composition data are usually Comparison of the New Composition not placed in the public domain. Thus Data with Existing Information on the the end users can not evaluate the Composition of Foods appropriateness of the analytical quality control used in those studies. The failure One aspect of almost all data quality to publish composition data in refereed control operations is a comparison of journals prior to placement in a the new data with the existing body of database is a worst case scenario. The knowledge. There are very few totally data are available but the user has no unique food composition data and idea of their quality. Ignorance is not reviewers will normally evaluate a new bliss in such cases. set of data by comparing it to the More analysts need to incorporate existing knowledge on food composition. more analytical control into their assays Extreme departures from existing and to better document those quality knowledge are usually rejected by control procedures. These actions can reviewers unless significant justifications be accomplished by both the editors and are presented for the acceptance of the reviewers having an absolute new data. Authors are well advised to requirement for documentation of good make such comparisons within the quality control procedures in all manuscript and provide a rationale for manuscripts accepted for publication. those data which appear to conflict with Currently, almost all textbooks and previously published data. courses on analytical chemistry, Ŷ Future Actions analytical biochemistry, food analysis and nutritional biochemistry contain While significant improvements in the little, if any, discussion of or instruction quality control of food composition data in assay quality control. (An exception is published in journals have been the recent book by Greenfield and accomplished in the recent years, a Southgate (12)). This is a fundamental great deal of work still needs to be done. failure in our training of future analysts A comment by Jorhem and Sundström and it should be corrected. We strongly in a recent paper (11) made the point advocate that all analytical courses and clearly: text books in these areas contain a During the last decade the thoughtful section on the basics of application of analytical quality assay quality control. control measures has gradually been Adoption of these actions will have intensified. However, since analytical several benefits. Authors will increase quality control activities are not yet in the documentation of quality control general use or standardized, it is procedures already in use in their often still difficult to compare results laboratories. Authors will increase the from different studies. use of acceptable assay quality control procedures in their studies. The existence of published papers with IUPAC (1978) Compendium of appropriate assay quality control will be Analytical Nomenclature, H.M.N.H. useful as good examples to those in the Irving, H. Freiser, & T.S. West field who wish to improve the quality of (Eds.), Pergamon Press, Oxford. their own composition studies. Finally, Klensin, J.C., Feskanich, D., Lin, V., the existence of papers with good Truswell, A.S., & Southgate, D.A.T. quality control procedures will permit the (1989) Identification of Food users of food composition data to better Components for INFOODS Data evaluate the appropriateness of each Interchange, UNU Press, Tokyo food composition data set for the Official Methods of Analysis (1995) 16th purpose at hand. Ed., AOAC INTERNATIONAL, Arlington, VA Ŷ References Rand, W.M., Pennington, J.A.T., (1) Stewart, K. K. (1987) J. Food Murphy, S.P., & Klensin, J.C. Comp. Anal. 1, 1 (1991) Compiling Data for Food (2) Stewart, K. K. (1988) J. Food Composition Data Bases, UNU Comp. Anal. 1, 291–292 Press, Tokyo (3) Stewart, K. K. (1989) J. Food Rand, W.M., Windham, C.T., Wyse, Comp. Anal. 2, 91–92 B.W., & Young, V.T. (1987) Food (4) Stewart, K. K. (1990) J. Food Composition Data: A User's Comp. Anal. 3, 103–104 Perspective, UNU Press, Tokyo (5) Stewart, K. K. (1992) J. Food Stewart, K. K. (Ed.) (1980) Nutrient Comp. Anal. 5, 1 Analysis of Foods — The State of (6) Stewart, K. K. (1992) J. Food the Art for Routine Analysis, AOAC, Comp. Anal. 5, 99 Washington, DC (7) Stewart, K. K. (1992) J. Food Stewart, K.K., & Whitaker, J.R. (Eds.) Comp. Anal. 5, 183 (1984) Modern Methods of Food (8) Rand, W.M. (1992) J. Food Comp. Analysis, AVI Publ. Co., Westport, Anal. 5, 267 CT (9) Stewart, K. K. (1993) J. Food Stewart, K.K. (1985) in Methods of Comp. Anal. 6, 105–106 Vitamin Assay, 4th Ed., J. Augustin, (10) Stewart, K. K. (1993) J. Food B. Klein, D.R. Becker, P.B. Comp. Anal. 6, 201–202 Venugopal, P.B. (Eds.), Wiley, NY, (11) Jorhem L., & Sundström, B. (1993) pp. 1–15 J. Food Comp. Anal. 6, 223–241 Wernimont, G.T. (1985) Use of Statistics (35) Greenfield, H., & Southgate, D.A.T. to Develop and Evaluate Analytical (1992) Food Composition Data: Methods, W. Spendley (Ed.), Production, Management and Use, AOAC, Arlington, VA Elsevier Applied Science, London, Wolf, W.R. (Ed.) (1985) Biological pp. 127–138 Reference Materials, Wiley, NY Quality Assurance Ŷ Additional Reading Garfield, F.M. (Ed.) (1980) Optimizing Chemical Laboratory Performance General Topics Through the Application of Quality Beecher, G.R., & Mathews, R.H. (1990) Assurance Principles, Proceedings in Present Knowledge in Nutrition, of a Symposium, AOAC, Arlington, 6th Ed., M.L. Brown (Eds.), ILSI- VA Nutrition Foundation, Washington, Garfield, F.M. (1991) Quality Assurance DC, pp. 430–443 Principles for Analytical Laboratories, AOAC, Arlington, VA Taylor, J.K. (1987) Quality Assurance of Borman, S.A. (Ed.) (1982) Chemical Measurements, Lewis Instrumentation in Analytical Publ., Chelsea, MI Chemistry, Vol. 2, ACS, Modern Assay Techniques Washington, DC Harlow, E., & Lane, D. (1988) Becker, J.M., Caldwell, G.A., & Zachgo, Antibodies: A Laboratory Manual, E.A. (1990) Biotechnology, A Cold Spring Harbor Laboratory, Laboratory Course, Academic New York, NY Press, San Diego Strobel, H.A., & Heineman, W.R. (1989) Boehringer Mannheim, GmbH (1987) Chemical Instrumentation: A Methods of Biochemical Analysis Systematic Approach, 3rd Ed., and Food Analysis, Mannheim, Wiley, New York, NY Germany Section IV

Information Needs and Computer Systems

his Session was chaired by Ms Karen Cashel of the University of Canberra. A keynote address was presented by C.E. West entitled The Future Information TNeeds for Research at the Interface between Food Science and Nutrition. This was followed by papers on Food Database Management Systems—a Review by W. Becker and I. Unwin and Data Identification Consideration in International Interchange of Food Composition Data by J.C. Klensin. These papers were followed by a computer demonstration Food Data: Numbers, Words and Images by B. Burlingame, F. Cook, G. Duxfield and G. Milligan. These are all published in this Section.

The following posters Computer Construction of Recipes to Meet Nutritional and Palatability Requirements by L.R. Fletcher and P. Soden (presented by D.A.T. Southgate) and Requirements for Applications Software for Computerized Databases in Research Projects by D. Mackerras are published at the end of this Section. Information Needs and Computer Systems

The Future Information Needs for Research at the Interface Between Food Science and Nutrition

Clive E. West

Department of Human Nutrition, Wageningen Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands and Program Against Micronutrient Malnutrition, Center For International Health, Emory University School of Public Health, 1518 Clifton Road, NE, Atlanta GA 30322, USA

Nutrition and food science are disciplines at the interface between agriculture and health. Therefore, their information needs encompass both those of agriculture and health, and in addition extend into the realms of other disciplines such as the basic physical sciences, mathematics, the social sciences from economics to anthropology, the behavioral sciences, and history. In this paper, attention will be directed to the narrow interface between nutrition and food science, addressing information needs such as food naming and description, food intake, attributes of foods, and nutritional status.

n order to be certain about the identity of foods being consumed or traded, general agreement about food names is needed backed up by an adequate food description Isystem containing a sufficient number of terms to describe foods in an unambiguous way. For this purpose, several systems, or types of systems, have been developed including Langual (1) the INFOODS system (2) and Eurocode 2 (3) as discussed more fully by Pennington (4). The most important use of Langual in population. Food balance sheet data are Europe and in other regions outside the useful in monitoring trends in food United States may not be in its consumption over time and in making comprehensive use but in the series of rough comparisons between countries. definitions which it provides for the Often, such data are the only data which description of food attributes. In order to can be readily obtained for rapid ensure that the Langual system being evaluation of new problems. The used in Europe does not develop continued need for such data was independently of that continuing to be highlighted by the resolutions of the developed by the Food and Drug International Conference on Nutrition Administration, a joint USEuropean held in Rome in December 1992 (5). Committee has been established. The Countries attending gave a commitment long-term success of Eurocode 2 and/or to meet the Nutrition Goals of the Fourth Langual in Europe depends on adoption United Nations Development Decade: in major Europeanwide epidemiological ƒto eliminate starvation and death studies and not just on endorsement by caused by famine projects such as FLAIR Eurofoods- Enfant. ƒto reduce malnutrition and mortality among children substantially Ŷ Food Intake ƒto reduce chronic hunger tangibly It is possible to measure food to eliminate major nutritional consumption or intake at three levels: ƒ diseases. the national level using food balance The first three goals are directed sheets; at the household level using essentially to problems in developing household budget surveys; and at the countries, while eliminating nutritional individual level using individual food diseases also refers to the problems consumption surveys. The data associated with the excess consumption obtained from these approaches enable of particular foods and nutrients. Food the availability or consumption of foods, balance sheets will be one instrument in and therefore of nutrients, to be monitoring the food and nutrition monitored. They can be used for a situation in countries throughout the variety of purposes such as the world and reacting to it. There are a development and monitoring of number of challenges associated with agricultural, food and nutrition policies the provision of food balance sheet data and for studying the relationship which are peculiar to various areas of between diet and health. The three the world. approaches for measuring food intake European Union. Ways have to be are complementary since all have their found to collect data, at the national advantages and disadvantages. At all level in countries in the European Union levels, challenges are emerging. after the creation of the single market. National and Regional Level Traditionally, national food balance data Food balance sheets provide a picture have been compiled largely from data of food disappearance within a country collected for customs purposes. during a specified reference period. The However, with the creation of the single term “food disappearance” refers to market, customs data are no longer “food available for human consumption” available. Unfortunately, this has also and not to “food actually eaten”. It can come at a time when statistical offices in be calculated not only for the whole Europe are undergoing reorganization population but also on a per capita basis and budget cuts. The problem will be by dividing the quantity of food by the exacerbated with the enlargement of the European Union. The FLAIR Eurofoods- Developing Countries. In developing Enfant Project has held discussions with countries, especially in Africa where the the three organizations responsible for per capita availability of food remains publishing food balance sheets up until low, there is a continuing need to collect this time: with Eurostat, which is the data. However, often the data are of low Statistical Office of the Commission of quality because of the inherent the European Union, FAO and with problems in collecting and analyzing OECD. The purpose of the discussion is information on food provided through to explore whether other survey non-commercial channels such as that techniques can be used to complement produced at the household level or or even replace the data collected in the obtained by hunting, gathering or conventional food balance sheets. FAO fishing. FAO and a number of and OECD are keen to maintain food governments provide assistance to balance sheets but Eurostat will provide some countries to improve their data data only for the countries of the collection and analysis procedures and European Union as a whole because capabilities. However, more needs to be other Directorates in the Commission done especially because of the need to have no interest in food intake in plan external assistance when the food individual countries in Europe. This is situation in countries deteriorates. because the Commission has no direct Coordination. In addition to mandate for health and nutrition matters maintaining and improving the but only an indirect mandate through its collection, analysis and dissemination of involvement in social issues. Thus the data from food balance sheets, there interest of the Commission in food should be more coordination of other intake will probably be restricted to the surveys designed to build up a picture of household level. It remains to be seen food consumption at the national level. whether FAO and OECD can continue Such data are often collected at the to collect food balance sheet data for household or individual level. As the countries of the European Union. mentioned later, there is a need to Eastern Europe. The increasing improve the quality of food composition number of newly emerging countries of data associated with food balance Eastern Europe and the established sheets. countries in transition have an even Household Level more pressing problem in providing data on food intake at the national level. At the household level, there are three Many surveys have shown that the main challenges: to improve the quality amount of food available in these of the data on food purchases generally; countries is declining rapidly but the to obtain comparability between data available to monitor such changes countries; and to determine food are often poor and not comparable over consumption outside the home. time or among countries. The situation Household budget surveys were is exacerbated by the lack of designed to measure household infrastructure for the collection, analysis expenditure, often for determining retail and dissemination of the data. OECD is price indices, and not to measure food providing help to many countries in the intake for nutritional purposes. Although region to improve the provision of such household budget surveys are data but more needs to be done. This is coordinated among countries of the important in order to maintain stability in European Union, the scope for the countries and for making improving the usefulness of data international arrangements concerning collection for nutritional purposes is trade and external assistance. somewhat limited because of the priorities of those collecting the data, the Secondly, countries not collecting data need to maintain the comparability of at the individual level should be the data over time, and the problem of encouraged to do so. This will enable converting expenditure on food to food countries to compare themselves with consumption. As yet, surveys designed each other. Any coordination will to measure food purchases at the probably have more effect on new household level are not coordinated surveys than on those already within Europe or among other countries. established such as the National Food This is unfortunate because, for Survey in the UK (7) and in the food nutritional surveys, household food intake components of the NHANES surveys will provide better data on surveys in the US (8). Thirdly, more household food consumption than will attention should be paid to collecting household budget surveys which are data required to answer such research designed with another function in mind. questions as the bioavailability of However, unless those who wish to nutrients and the etiology of cancer and coordinate national household food other diseases. surveys can come up with money to Ŷ Food Composition improve or modify surveys, there is very little chance that national household One of the first priorities when food surveys will be coordinated in the INFOODS was established (9) was to foreseeable future. Measuring food produce guidelines for the production, consumption outside the home is very management and use of food difficult because the person responsible composition data. These guidelines for purchasing food for the family as a (11), which were published with the whole is often not aware of food assistance of the FLAIR Eurofoods- purchases by individual household Enfant project, have now become the members. definitive work in the area. However, for nutrition research, there are a number of Individual Level problems which need to receive Data at the individual level, particularly if increased attention in the future. for a sufficient number of people, Data on More Foods provide the best information for nutritionists especially for examining the Developed Countries. In developed relationship between diet and health or countries, the main gap in our diet and disease. It is not appropriate to knowledge is for data on prepared and discuss here all that needs to be done to processed foods, especially those improve dietary intake information at the prepared in the home. Much of the individual level as this topic has been information on these foods is derived by discussed in detail at other meetings calculating the nutrient content from that such as the Dietary Assessment of the ingredients and the proportion of Meetings, the first of which was held in the various ingredients given in recipes. 1992 (6). However, a number of points This can give rise to errors because of should be noted. Firstly, there should be the imprecision of the recipe and more coordination to improve because of the losses and gains of consistency of data among countries. individual constituents during the This can only be achieved when the process. Often, for example, fat added coordinating agency can offer funds to during preparation is not eaten while the those carrying out the work. Otherwise removal of water by evaporation during countries are reluctant to change their cooking or drainage after cooking can systems because such changes can increase the concentration of many affect the continuity of the data. constituents in a food. Minor components such as water-soluble simple because non-provitamin A minerals, trace elements and vitamins vitaminoid activity is not a single can be discarded with the cooking water function but includes a range of while fat-soluble vitamins can be antioxidant activities and activity in discarded with cooking oil. Some food modifying the immune response which components, such as vitamin C are to some extent is independent of destroyed during food preparation. It is antioxidant function (12). Individual also important to know the degree of carotenes differ in their ability to carry nutrification of processed foods. out the various non-provitamin A Developing Countries. In developing vitaminoid activities attributed to them. countries, data for many nutrients and Thus the following classification based energy not only for prepared and on that of Olson (13) uses singlet processed foods but also for oxygen quenching activity as the non- unprocessed foods are lacking. If data provitamin A activity: are available, they are often derived ƒType 1: Provitamin A and non- from data from “comparable” foods provitamin A vitaminoid activity (ȕ- elsewhere which may not be carotene) appropriate. ƒType 2: No provitamin A activity but Analytes of Interest non-provitamin A vitaminoid activity When deciding which substances to (canthaxanthin) analyze in foods, priorities have to be ƒType 3: Provitamin A activity but no set because analytical chemists can non-provitamin A vitaminoid activity produce information on a very large (ȕ-apo-14'-carotenal) number of food components. Thus, it Type 4: No provitamin A nor depends on the nutritional problems ƒ nonprovitamin A vitaminoid activity being investigated. However, this means (phytoene) that a chicken-and-egg situation It may well be that the non-provitamin A develops because nutritionists often do vitaminoid activity of carotenoids is not know which food components are overemphasized because important if they do not have information measurements of carotene intake on the concentration of the components usually reflect consumption of dark in the food. For example with dietary green leafy vegetables and fiber, it was necessary to have data on orange/yellow-colored fruits. Other different classes of fiber before their minor components of such foods may nutritional significance could be have greater nutritional significance. For investigated. In the past, one total value example, it has recently been reported for a vitamin or a value for a particular that quercetin in plants is associated vitamer was regarded as adequate but with lower rates of heart disease (14). now, many nutritionists would like Such components are generally referred separate data on all individual vitamers. to as non-nutrients, a class of Carotenoids are an interesting case in substances with a wide range of point. In the past, only provitamin A structure and function. As discussed for activity was considered with 6 mg of ȕ dietary fiber, it will be difficult to set carotene or 12 mg of other provitamin A priorities in the analysis of non-nutrients carotenoids being equivalent to 1 mg of because, without composition data, retinol. However, it is now thought that epidemiological studies will not be able carotenoids also have non-provitamin A to show whether their intake is important vitaminoid activity. Thus it is possible to or not. classify carotenoids based on the activities they possess. This is not There are a numbers of ways which usefulness of the data. If values are food components can be classified. tightly distributed, they would be of use Apart from classifying them as nutrients to a wide audience but if they are or nonnutrients, we could also classify specific to the batch of food in question, them as favorable, neutral or they would be of use only to people with unfavorable components (often an interest in that food. Food naming depending to a large extent on the and description will be very important in content in a particular food or the diet as determining the degree to which the a whole). However, a more useful data can be used more generally. The classification may be into intrinsic extent to which the data are widely substances, non-intentional food applicable will be important in additives and intentional food additives. determining the policy on making the Intrinsic Substances. These are data available. absorbed from the environment or Analytical Methods produced by the plant or animal from which the food is derived. The content of Many of the basic methods for food some of the components is reasonably analysis were established about one constant while the content of other hundred years ago and there has been components, such as of trace elements very little change in the principles of the (essential; nonessential but non-toxic; methods since then, even though the and toxic) would depend on their apparatus used may have been content in the food chain and the automated to some extent. However for environment. Important intrinsic non- some components, the introduction of nutrients in foods are tannins and phytic new techniques has been essential for acids which affect the bioavailability of obtaining reliable data. Such techniques iron (15, 16). include chromatography, both gas-liquid Non-intentional Food Additives. chromatography and liquid These are neither intrinsic to the food chromatography, and atomic absorption nor added intentionally. They include spectrophotometry, a technique which microbial metabolites, such as aflatoxin was developed in Australia. Through and some B vitamins, hormones, their use, it has been possible to antibiotics, and components derived generate data on the content in foods of during storage, preparation and amino acids, fatty acids and a wide transport including components derived range of vitamins and minerals. There from packing materials. Thus, generally, are a number of tasks facing analysts the content of these components in today. foods is very variable. Development of Techniques for the Intentional Food Additives. This Analysis of Food Components for Which group comprises substances added to No Adequate Methods Exist. Such food give the desired physical appearance or components include not only those structure, organoleptic properties or which are well recognized, such as nutrient value and include emulsifiers, vitamin K, heme iron and non-heme colors, flavors and also nutrients. iron, but also compounds which are just Generally, but not always, the content of being recognized as having nutritional these components is reasonably importance such as the flavonoids (17). constant for a given food. Development of Techniques Suitable The way in which values on the for Use in Laboratories in Developing concentration of components in a Countries. In western countries, database are handled depends not only equipment has become sophisticated on the distribution or range of values and sometimes highly automated encountered but also on the general because of the high cost of labor and the ready availability of funds for are a combination of nutrient equipment and expendables. In composition values with factors developing countries, often labor is analogous to the extraction rate of relatively cheap but limited funds are nutrients from cereals but the system available for equipment, parts and was not well documented. Thus they reagents. In addition, provision of suggested that the factors be separated constant power and water is often a so that each component could be problem. Since the need for data on checked and revised if necessary. food composition in developing Preliminary work has also been done on countries is even more pressing than in food composition tables for use with developed countries, the development household budget surveys. of methods suitable for use is a pressing Ŷ Physical Properties problem. Such development will have to be accompanied by the establishment of Although much work has been done on suitable quality control procedures. the physical properties of foods, such as Since two of the most important on viscosity, elasticity, tensile and shear nutritional problems in developing strength, and water-holding properties, countries are vitamin A and iron the information is not as readily deficiencies, methods for the available as that on the content of determination of provitamin A various constituents. This is an area carotenoids, tannin and phytic acid which should receive more attention in should receive high priority. the future. The data are not only of Quality Control of Analyses and interest to food processors but should Determination of the Quality of Data. also be of interest to nutritionists The use of reference materials has been especially those involved in discussed by Tanner et al. (18). Their bioavailability. proper use is essential for producing Ŷ Bioavailability good quality analytical data. Evaluation of data is a difficult task and it needs to An area which must receive much more be made less subjective. Mangels et al. attention in the future is the (19), have made some progress in this measurement of the bioavailability of area by developing expert systems for food constituents. A start has been the evaluation of data on the carotenoid, made with a number of vitamins and copper and selenium content of foods. It minerals such as calcium, iron, zinc and is essential if data in databases are a number of B vitamins but very little going to be widely distributed that has been done with respect to uniform criteria for data are adopted. bioavailability of other nutrients such as Levels of Data Required. When the carotenoids. Since bioavailability nutritionists consider food composition depends to a large extent on the meal in tables, they generally think of them for which the food constituent in question is calculating nutrient intake from food consumed, this means that we will need intake (or vice versa) at the individual information not only on daily food level. However, as mentioned above, consumption but also on intake of other data on food consumption are also constituents at individual meals. collected at the household and at the Recently, I have developed a series national or regional level. As part of the of carotene bioavailability indices (or FLAIR Eurofoods-Enfant Project, Carbi indices) to correct carotene intake Belsten and Southgate (20) reviewed for bioavailability (West, in preparation). the so-called “conversion factors” for Carbi-1 Index. This provides a converting food disappearance data to measure of the absorption of provitamin nutrient data. Up until now, the factors A carotene from a given matrix relative to the absorption of the same amount of However, the extent of conversion of carotene dissolved in oil. Based on the these carotenoids to ȕ-carotene varies. work of Hume and Krebs (21), the The idea of such indices is not new. following is a provisional list of Carbi-1 Monsen et al. (15) have developed a indices: method by which the amount of iron ƒȕ-carotene dissolved in fats/oils, which is bioavailable can be estimated 1.00 from the intake not only of iron but also of enhancers and inhibitors of iron ƒȕ-carotene in cabbage and spinach, absorption. It is just as important for the 0.53 Carbi concept to be used in order to ƒȕ-carotene in carrots, boiled, sliced, assess whether the intake of carotene- 0.33 containing foods meets the vitamin A ƒȕ-carotene in carrots, domestic requirements of individuals. For puree, 0.33 example, a child consuming boiled sliced carrots, with a fat intake of less ƒȕ-carotene in carrots, homogenized, than 3 g/d, and infected with Entamoeba 0.73. histolytica would need to consume 24 Carbi-2 Index. This provides a measure times more of the food in order to meet of host-dependent reduction in requirements than the content would carotenoid absorption and/or conversion suggest. to retinol and would be related initially to the intake of fat (Carbi-2a index) and the Ŷ Nutritional Status degree of parasitemia (Carbi-2b index). Nutritional status with respect to a Based on the work of Jayarajan et al. particular nutrient depends to a large (22), the Carbi-2a index would be 0.5 extent, but not entirely, on the intake of when the intake of fat in children was the nutrient in question. Bioavailability, less than 3 g/d. Similarly, based on work concurrent ingestion of other nutrients, from our laboratory on the absorption of physiological factors, and environmental iodized oil (23), the Carbi-2b index in and genetic factors also play a role in Entamoeba histolytica-infected children determining nutritional status. Be that as would be 0.25. Other Carbi-2 indices it may, there is a need to examine the could be developed to take into account relationship between nutrient intake and factors such as the effect of various status and to collect more information types of dietary fibre on carotenoid on the nutritional status of people absorption and of zinc deficiency on the especially at the national or regional conversion of carotenoids to retinol. level. Using such data in conjunction Carbi-3 Index. This provides a with food and nutrient intake data, it is measure of the effect of carotene intake possible to develop and monitor on the rate of conversion of ȕ-carotene strategies for controlling nutritional to retinol as suggested in the FAO/WHO imbalances. recommendations (24). For the purposes of calculating the Carbi-3 Ŷ Priorities for the Future index, carotene intake should first be Providing Data on Food Composition corrected by applying the Carbi-1 and for Developing Countries Carbi-2 indices. Carbi-4 Index. This provides a In developed countries, increased measure of the extent of conversion of resources for generating data on food various carotenoids to retinol. With the composition will require a reallocation of Carbi-4 index for ȕ-carotene set at 1.00, resources within the countries the Carbi4 index for other provitamin A themselves (including via the carotenoids is generally set at 0.50 (24). Commission of the European Union). Forums such as INFOODS, FLAIR data (28), and on procedures for Eurofoods-Enfant and the National transferring data between nutrient Nutrient Databank Conferences in the databases (29). In addition, they have US, and meetings such as the present made a start in establishing regional one will play an important role in the centers throughout the world. This effort exchange of ideas. However in has been strengthened by INFOODS developing countries, Eastern Europe joining forces with FAO (11). In Europe and the former countries of the Soviet a similar organization, which has worked Union, the needs for data are being met closely with INFOODS was established. only poorly and the countries need Initially, this was referred to as assistance from outside to improve the Eurofoods but was later incorporated situation. For example, in Africa, the into the Food-Linked Agroindustrial most comprehensive source of data on Research (FLAIR) Programme of the the composition of foods was published Commission of the European Union as by FAO in conjunction with the US Eurofoods-Enfant. These organizations Department of Health, Education and have been working towards the Welfare in 1968 (25). This book, as well improvement of the quality and as those prepared for a number of other compatibility of data on food world regions, is now hopelessly composition and consumption in outdated and inadequate in terms of the Europe. There work has led to a marked number of foods, nutrients and other improvement in the quality, food components on which data are comparability and accessibility of data available, food naming and description, on food composition in Europe (30). The analytical methods used, and the quality contract supporting Eurofoods-Enfant control of the data. For example, many finished at the beginning of 1994 but a of the methods available at the time for new project is planned to commence at the determination of nutrients were the end of 1994 through the COST poorly developed. This is particularly mechanism of the Commission of the true for the determination of provitamin European Union. One activity evolving A carotenoids so estimates of the out of the Eurofoods-Enfant Project is amount of vitamin A which can be the series of biennial Postgraduate provided from the diet are Courses on the Production and Use of overestimated, probably by a factor of Food Composition Data in Nutrition. The two (26). For non-nutrients, practically Second Course, held in October 1994 no data exist. This is particularly under the auspices of the Graduate important for those factors influencing School VLAG (Advanced Studies in bioavailability such as phytic acid and Food Technology, Agrobiotechnology, tannin referred to above with respect to Nutrition and Health Sciences) at iron. Wageningen Agricultural University in Since 1982, a number of groups have conjunction with UNU, FAO and the been active in stimulating international International Union of Nutritional cooperation on improving the quality Sciences was attended by over 30 and availability of data on food people from about 20 countries. Such composition. The INFOODS project of courses will help to increase expertise in the United Nations University has the area of food composition tables and examined the needs of users (27) and nutrient databases worldwide. developed guidelines in a number of There is no doubt that in order to areas such as on the description of meet the goals of the International foods (2), definition of names of Conference on Nutrition, it is essential nutrients with appropriate tag names that a program of action should be which can be used when transferring instigated to produce and disseminate data on the composition of foods in been made with pictures of foods being developing countries and in Eastern stored on compact disk. Perhaps food Europe. All nutrition-related programs texture will be recorded as the sound of depend on the availability of such data a standard person biting into a standard in the same way as traffic depends on carrot. But what about smell and taste of maps. This work will require an input of foods and the ethereal atmosphere in resources from industrialized countries which foods are eaten? Is computer especially the US and those in Europe. technology up to storing these data yet? New Developments in Computer Use Computers are becoming faster and Ŷ References cheaper, data storage is also becoming cheaper and software is becoming more (1) McCann, A., Soergel, D., Holden, sophisticated. All of these developments J., Pennington, J., Smith, E., & mean that computers will be more able Wiley, R. (1980) Langual to serve the needs of nutritionists and Vocabulary, Users' Manual, US food scientists. However, it is becoming Food and Drug Administration, more and more important to develop Washington, DC systems and practices to ensure the (2) Truswell, A.S., Bateson, D.J., quality of both input and output from Madafiglio K.C., Pennington, J.A.T., computer systems. It is all too easy to Rand, W.M., & Klensin, J.C. (1991) think that more is necessarily better. J. Food Comp. Anal. 4, 18–38 Users should remember the computer (3) Poortvliet, E.J., & Kohlmeier, L. adage: “garbage in means garbage out”. (1993) Manual for Using the The expert systems described by Eurocode 2 Food Coding System, Mangels et al. (19) need to be improved Wageningen Agricultural University, and extended to nutrients other than Wageningen copper, selenium and the carotenoids in (4) Pennington, J.A.T. (1995) in Quality order to ensure the quality of data being and Accessiblity of Food-Related entered into nutrient databases is Data, H. Greenfield (Ed.), AOAC adequate for the use envisaged. Similar INTERNATIONAL, Arlington, VA, systems will need to be developed for pp. 85–97 dietary intake data and for ensuring the (5) FAO/WHO (1992) International quality of food names and descriptions Conference on Nutrition. World as well as for monitoring the quality of Declaration and Plan of Action for data output. There is a lot of pressure Nutrition, FAO, Rome for online systems to be developed for (6) Buzzard, I.M., & Willett, W.C. (Eds.) supplying food composition and (1994) Am. J. Clin. Nutr. 59, 143S– consumption data. In my opinion, the 306S need for on-line systems is over- (7) Ministry of Agriculture, Fisheries emphasized because most of the food and Food (1953-) Household Food composition data in nutrient databases Consumption and Expenditure, in the world can be accommodated on HMSO, London one compact disk. Perhaps some (8) Briefel, R.R., & Sempos, C.T. (Eds.) entrepreneur would like to consider an (1992) Dietary Methodology annual version of “World Nutrient Data” Workshop for the Third National on compact disk although such a Health and Nutrition Examination venture may not be economically viable. Survey, US Government Printing As far as computers are concerned, Office, Washington, DC the biggest challenge is to present non- (9) Rand, W.M., & Young V.R. (1984) alphanumeric information. A start has Am. J. Clin. Nutr. 39, 144–151 (10) Greenfield, H., & Southgate, D.A.T. Adults: an Experimental Study of (1992) Food Composition Data: Vitamin A Deprivation in Man, Production, Management and Use, Medical Research Council Special Elsevier Applied Science, London Report Series No. 264, HMSO, (11) Lupien, J. (1995) in Quality and London Accessibility of Food-Related Data, (22) Jayarajan, P., Reddy, V., & H. Greenfield (Ed.), AOAC Mohanram, M. (1980) Indian J. INTERNATIONAL, Arlington, VA, Med. Res. 71, 53–56 pp. 3–9 (23) Furnée, A.C. (1994) PhD thesis, (12) Bendich, A. (1992) Voeding 53, Wageningen Agricultural University, 191–195 Wageningen (13) Bendich, A., & Olson, J.A. (1989) (24) FAO/WHO (1988) Requirements of FASEB J. 3, 1927–1932 Vitamin A, Iron, Folate and Vitamin (14) Hertog, M.G.L., Feskens, E.J.M., B12, FAO Food and Nutrition Series Hollman, P.C.H., Katan, M.B., & No. 23, FAO, Rome Kromhout, D. (1993) Lancet 342, (25) Wu Leung, W.T., Busson, F., & 1007–1011 Jardin, C. (1968) Food Composition (15) Monsen, E.R., Hallberg, L., Table for Use in Africa, US Layrisse, M., Hegsted, D.M., Cook, Department of Health, Education J.D. Mertz, W., & Finch, C.A. (1978) and Welfare, Bethesda and FAO, Am. J. Clin. Nutr. 31, 134–141 Rome (16) Hallberg, L., & Rossander-Hultén, (26) West, C.E., & Poortvliet, E.J. (1993) L. (1993) in Bioavailability '93: The Carotenoid Content of Foods Nutritional, Chemical and Food with Special Reference to Processing Implications of Nutrient Developing Countries, USAID- Availability, Part 2, U. Schlemmer VITAL, Washington, DC (Ed.), Bundesforschungsanstalt für (27) Rand, W.M., Pennington, J.A.T., Ernährung, Karlsruhe, pp. 23–32 Murphy, S.P., & Klensin, J.C. (17) Hertog, M.G.L., Hollman, P.C.H., & (1991) Compiling Data for Food van der Putte, B. (1993) J. Agric. Composition Data Bases, UNU Food Chem. 41, 1242–1246 Press, Tokyo (18) Tanner, J.T., Wolf, W., & Horwitz, (28) Klensin, J.C., Feskanich, D., Lin, V., W. (1995) in Quality and Truswell, A.S., & Southgate D.A.T. Accessibility of Food-Realted Data, (1989) Identification of Food H. Greenfield (Ed.), AOAC Components for INFOODS Data INTERNATIONAL, Arlington, VA, Interchange, UNU Press, Tokyo pp. 99–104 (29) Klensin, J.C. (1992) INFOODS (19) Mangels, A.R., Holden, J.M., Food Composition Data Beecher, G.R., Forman, M.R., & Interchange Handbook, UNU Press, Labuza, E. (1993) J. Am. Diet. Tokyo Assoc. 93, 284–296 (30) Castenmiller, J., & West, C.E. (20) Belsten, J.L., & Southgate, D.A.T. (Eds.) (1994) Report of the Third (1992) Review of FAO Food Annual Meeting of the FLAIR Balance Sheet Nutritional Data, Eurofoods-Enfant Project, AFRC Food Research Institute, Wageningen Agricultural University, Norwich Wageningen (21) Hume, E.M., & Krebs, H.A. (1949) Vitamin A Requirement of Human Information Needs and Computer Systems

Food Database Management Systems — A Review

Wulf Becker

Nutrition Division, National Food Administration, Uppsala, Sweden

Ian Unwin

The Opas Centre, Cambridge CB4 4WS, UK

The use of database management systems (DBMS) for handling food composition data is reviewed, together with some basic concepts underlying database design and current developments in DBMS support for food data. The results of a survey of system users in Europe, USA and Australasia indicated that facilities supporting the identification and description of foods, as well as methods for specifying compositional data, need to be extended and harmonized. Most systems are unilingual, but include synonyms for foods, while some support multiple languages. Generally, a single grouping or classification system for foods is used; it is often based on food source and built into the system of food codes. Facilities for calculating and storing measures of variation in a compositional value, and for describing the quality of a value are frequently lacking. Computer- readable composition data are usually exchanged as text files on floppy disk. Although most food information handling DBMS have been developed for the needs of a specific organization, more sophisticated software tools and international standardization (e.g. INFOODS, FLAIR Eurofoods-Enfant and multinational epidemiological studies) are encouraging collaborative development. The New Zealand Food Composition Database, the Swedish NUTSYS system and the EuroNIMS collaboration are briefly described as examples of recent developments in this area. n increasing number of countries is compiling and publishing food composition tables. Inventories of food composition tables and nutritional software in Europe A(1, 2, Slimani & Poortvliet, unpublished) showed that many organizations responsible for publishing tables use a computerized database management system (DBMS) for the handling and management of food composition data and related information. The systems were either developed in-house or based on commercial software packages, operated in various computing hardware and software environments, and were generally designed for the specific needs of the individual organization. Few of these systems were commercially available for other users. There are high costs involved in the production of high quality food composition data as well as in the development of FDBMS (Food DBMS) for handling the data. In view of this, efforts have been made, during the last decade, to improve the availability of national food composition data and to develop the means to achieve the international exchange of data. The purpose of this review is to outline modern database management techniques, including the relational model, and to give a brief overview of some existing database systems used for handling food composition data and of some recent international developments. More detailed descriptions of the handling of bibliographic information (3) and food composition information (4) using the relational model have been published.

diagram (7) may be used to model the Ŷ Databases — Basic Concepts data conceptually. This model may then Although alternative data structures be implemented using a logical data based on hierarchical or network models structure based on the relational model may be used in database management (8, 9). Some attention needs to be paid software, much attention is at present to the terminology which derives from paid to the relational approach in the several origins including “traditional” file design of food information handling processing data description, data systems. The first of two main reasons modeling methodology such as E-R for this is that commercially available diagrams, and relational theory. Further relational DBMS (e.g. Oracle, Sybase, care will be required as terminology Ingres) provide the software of choice from object-oriented data modeling for many organizations. The second is becomes wide- spread. In referring to that food composition data appear well data, whether conceptual or logical, suited to the application of relational associated with basic data structures, it principles, consisting of data values is convenient to use E-R terminology. which relate to a food, a component, Thus in the following description, “entity and to various other entities such as type” refers to a single data structure analytical method, analytical laboratory and “entity in stance” for an individual and literature reference. Therefore very occurrence of that type of item. briefly we shall review the main All items of data in a relational DBMS concepts underlying relational DBMS are held in data structures constructed and note some possible limitations as “tables” (in the more formal literature which might impinge on food-related often referred to as “relations”). Each information-handling facilities based on table deals with a separate subject or them. entity type, e.g. a nutrient value, journal Database analysis and design (5, 6) article or author. In a table each row involves, inter alia, the building of a refers to a different instance of the conceptual data model and its entity, an individual value, article or translation into a logical data model. For author, and each column to a particular example an Entity-Relationship (E-R) property (or “attribute”) of the entity for which data are held, e.g. the journal, instance to another of the same entity volume, issue and pagination data for a type. A common example is an entity journal article. Rows are frequently type in which instances have a referred to as records and columns as hierarchical organization, as with fields. employees in a management structure It is a requirement of the relational or in a facet of a food description model that each row of the table is language. Usually a relationship is uniquely identified by the data for one of between two (binary), three (ternary), or its attributes (or the combined data of occasionally more entity types. A further more than one attribute); the identifier is property of a relationship is its known as the table's “primary key”. A “cardinality” which expresses the primary key may include meaningful number of instances which can partake data, e.g. food name (or food group) for from each side of the bidirectional a food item, or have no meaningful relationship. Cardinalities are often content, being for example a expressed as 1:1 (a “one-to-one” sequentially assigned number. Care relationship), 1:N (a “one-to-many” must be taken in selecting a meaningful relationship) or M:N (a “many-to-many” key since it must remain unique over relationship). However further detail is any valid items which may need to be important, in particular whether the added to the table and must remain cardinality is mandatory or optional, i.e. constant for the given item; so-called whether for a one or many cardinality “intelligent keys” are usually avoided. the minimum requirement is one or zero Any identifier whose assignment rules, occurrences. A maximum or minimum assignment or use are external to the number of occurrences may apply to system under design should be treated “many” cardinality; this is less significant with similar caution. For example, for data structure but important for data neither Chemical Abstracts Registry validation. Numbers nor ISBN (for example as for The relational data model provides the 4th Edition of The Composition of the means to eliminate the redundant Foods (10) are unique for substances or storage of information in the database books, respectively, at the level which would result in possible appropriate to an FDBMS. Equally for inconsistencies during data insertion, meaningless keys, adequate checks deletion and amendment procedures. In must be made whether incoming data the process of “normalization” the belong to a new row (a new entity structures of tables are subjected to a instance), or update an existing row. series of steps in which dependent Although this may not be a significant attributes are removed to separate problem for real-life discrete objects and tables. For example, a table of food events such as employees and sales, component values may include less discretely defined entity instances, attributes concerning the analytical as in food items, may provide some laboratory. However these details difficulty. should not be repeated in each row The “relationships” of an E-R model corresponding to a value generated by are associations between instances of that laboratory. Instead details of the one or more entity types, the “degree” of laboratory such as its name, address the relationship being the number of and contact person are removed to a entity types involved. A “unary” or separate table linked by a one-to-many “recursive” relationship is a link from one laboratory-to-values relationship.

Table I. Component value table FOOD_ID COMP_ID SOURCE_TYPE SOURCE_ID VALUE UNIT_ID 1234 CARTBEQ L 4321 1540 (µg/100 g) 1234 RETOL L 4321 16500 (µg/100 g) 1234 VITA L 4321 16760 (µg/100 g) 1234 VITAA A 6789 54000 (IU) 1234 VITA- L 3456 20000 IU The relationship links between tables to be differentiated through additional are made through a “foreign key”, an entity types such as literature reference attribute which records the primary key or analytical detail which may include for a row in (except for a unary identification of the laboratory. To avoid relationship) another table thus pointing having separate attributes pointing to to related data. A direct link can be literature and analytical source made between tables for 1:1 information, a separate attribute could relationships and for 1:N relationships, be defined for “source type” to point to since more than one table row on the N different tables (e.g. Literature or side of the relationship can hold the Analytical). This is then required in the same foreign key. Since this can only unique key since duplicate values of the apply in one direction, an M:N single attribute (SOURCE_ID) might relationship is stored by breaking it into appear in both the literature and two 1:N links through the insertion of an analytical information tables. Thus, with additional table. This has an attribute the columns constituting the unique key column for the primary key of each delineated in bold, the component value entity table of the relationship. Although table might be constructed as in Table I. these keys can be repeated as In Table I, components are identified necessary, any one row of the relation using INFOODS tagnames (11, 12). The table has a unique combination of the first three represent ȕ-carotene keys from the entity tables, and indeed equivalents, retinol and vitamin A (as the combination serves as its primary retinol equivalents). A separate key. In practice, data can be associated tagname, VITAA, identifies vitamin A with the M:N relationship and stored as determined by bioassay and another, attributes in the table. For example the VITA, indicates a value for vitamin A relationship between authors and whose method of determination is published papers is M:N, with details on unknown. The attribute UNIT_ID each held in separate tables. However identifies the unit in which the value is data concerning the authorship of a expressed. Those in parentheses are given paper must be held in the linking default values for the corresponding table. In particular, the position of a component. This information could be given author must be recorded here if held for the component and might be the ordered list of authors needs to be omitted in this column. Note the default reconstructed, for example in the unit for VITAA is IU. However it is formatting of references. µg/100 g for VITA- and thus the unit of A central component of a food IU must be held explicitly for the final composition database, the component row. value record, can be considered further. An important aid to the A table for such data must include standardization of development in foreign keys at least to identify the food relational database applications has and the component. It will also have been Structured Query Language (SQL; actual data such as the value, its units, often pronounced “Sequel”) which is and perhaps reports of quality, based on work done by IBM in the precision, status, etc. Values for the 1970s. SQL provides a concise set of same food-component pairing will need commands to support the definition, display and updating of relational tables predefining items in the data collection (a broad interpretation of “query”!). at the level normally considered an These facilities include the handling of individual food item. Instead every “views” providing in a convenient distinguishable instance or sample could derived table the data based on a be stored as separate records, with data subset of rows and columns selected being aggregated on search criteria from underlying existing tables. Views (given that an adequate system of food are dynamic; changing their data indexing is available) or through links changes the corresponded stored data between equivalent items as identified and any change in the underlying data is by data managers and stored in a food reflected in the values displayed in a correlation table. The latter approach view. would allow separate sets of aggregate Basing the design of an FDBMS on items to be maintained, for example to the relational model allows similar data support food table production and non- structures and data management nutrient work. procedures, using SQL, to be Further circumstances where the “all implemented in a wide range of or nothing” separation of food items is hardware and software environments. unsatisfactory include seasonal data However various shortcomings in the sets which vary in only one or two use of the relational model for FDBMS components and the need to apply have been noted (13) and a more taxonomic and alternative names to extensive review, particularly with each item derived from one raw food. respect to the limitations of SQL for the Although relational solutions can be management of statistical data, has envisaged, new developments, in been published (14). The relational particular an object-oriented model, may model needs each instance of an entity provide an approach which more closely type to be clearly distinguishable from represents the real world. In object- all other possible instances. This is often oriented programming a key concept is not the case with entity types required in the “class” object which can hold a an FDBMS such as food components, number of objects of different types analytical methods and particularly, as (such as various types of variable, data we note below, food items. Also structure and function), together with relational systems may not be adept at functions for manipulating the objects handling the textual information needed and mechanisms to control inheritance to document such entities and the and access from other classes. statistical descriptions associated with Variables and functions declared for the component values. class are known as class members. A A fundamental problem for food derived class can be declared, inheriting information management is the members from one or more base underlying assumption, apparently (parent) classes. In the derived class, required if the relational model is to be new members may be declared and applied, that a food item must be existing ones redefined. It may be that uniquely defined and distinguished from an object-oriented data model will all other food items. This may provide handle the characteristics acquired (for difficulties, inter alia, with variants of example, in processing or cooking) by a composite foods and when switching derived food item more effectively, contexts, for example between including the multiple inheritance from composition and consumption records the various ingredients implicit in and between nutrient and non-nutrient composite foods. studies. A possible alternative staying Ŷ Facilities of Existing Food DBMS within the relational model is to avoid In order to obtain additional information compositional data. The system users on facilities available in DBMS were also asked to give examples of implemented for handling food improvements they would consider to be composition data, a questionnaire was most important. The purpose of the sent to a selection of 20 system users in survey was to compare how systems Europe, USA, Australia, New Zealand differed in the handling of food and and the Pacific. The questionnaire component information and thus to focused on aspects of information identify areas that could be considered handling of foods and components, problematic in relation to international recipe calculation, and storage of exchange of food information.

Table II. Food information technology Food A general term, sometimes used more specifically for a basic (e.g. unprocessed) food item. Food item A specific term for a unique entity which can be differentiated from all other food entities with which it may be compared. Food identification The decision whether two food entities can be considered the same food item. This may be achievable within defined objectives or particular data collections, but such decisions may not be valid in any broader context, e.g. data exchange. Food identifier Any tag (code, name, etc.) which is unambiguosly associated with (but not necessarily unique for) a food item. Food code A code (which may be a sequential number based on an ordered list of items or incorporate some degree of hierarchical classification) used to identify a food item unambiguously. Food name A name assigned to a food item considered sufficient to distinguish it from all other items which may also occur in the data collection. The names for existing items may need to be made more specific when new, similar items are added. Food description Information on a food item which may be relevant to the data (e.g. on composition) associated with it; the information may reside in the food name, in an overlying classification or as additional descriptive detail. Food descriptor A terms included in a more or less formal set used as a food description system. The system may be faceted where the descriptors are organized in subsets according to the attribute described, e.g. preservation method, cooking method. Food grouping A categorization of food items based on an individual attribute or a selection of attributes which groups the items usefully within a given (broad or narrow) context. Food classification Any grouping system for food items (often using hierarchical categories) which attempts to assign a single “correct” (i.e. unique) locationa for any food item. Food Group The primary food classification, based on a single hierarchy, which (note capitalization) a food information system uses; generally based on food type and/or source. Parallel _ _ _ systems Where more than one independent system of either food codes, names, description, grouping or classification coexist in an implementation, these are referred to as parallel systems. a If unique locations are defined down to the level of each individual item, the location may also be considered unambiguous with respect to that item. However this approach to food identification will only be canonical if incontrovertible rules can be defined for assigning items to locations. This is highly improbable for foods.

The organizations covered were allow food and component names to be mainly those responsible for official held in multiple languages. national food composition data Generally, a single grouping or compilations but others mainly involved classification system for foods is used; it in dietary surveys and epidemiological is often based on food source and may research were also included. Answere be built into the coding system. Half of were received for 17 systems. Fifteen the FDBMS used for national food systems were developed in-house using composition data allow the use of various programming languages or multiple grouping or classification commercial DBMS (e.g. Oracle, systems in parallel. Advanced Revelation, dBASE III). No Preferably, a FDBMS should allow clear preferences were evident for the the use of multiple coding systems, e.g. hardware environment, programming parallel management of national food language or DBMS; it appeared these codes and Eurocode 2 (16), and of were determined by practise and parallel grouping and classification availability within the organization. systems based on different criteria. For Information Handling of Foods the use of international data it is also necessary to be able to handle in Food databases contain records on food parallel multiple food names, including items and the information detailing these different language versions. is crucial. Thus the identification, description, grouping and classification Information Handling of Recipes of foods, the ways of representing A recipe system for calculating the information on the items, are key areas composition of dishes or mixed foods in the handling of food information. from their ingredients is usually needed However, there are still no universally in a FDBMS. Apart from calculating the accepted definitions and taxonomy of composition of cooked dishes, such a these terms, which would be desirable system can be used to estimate the when using them in an international composition of mixed foods from the context. Table II shows suggested proportions and compositions of their definitions for a number of terms used in constituents. A recipe system is the paper. included in fourteen of the systems According to the questionnaire reviewed, especially in those mainly responses, foods are usually identified intended for processing of data from by their names and a code. Some dietary surveys. About half of the recipe systems (one-third of those used for systems allow for the use of alternative data compilation but none used for preparation methods, portion sizes, and dietary survey work) can include several yield factors for change in weight/water coding systems in parallel. Two systems content during preparation but facilities use Langual (15), and six a less formal for adjustments in fat content were but similar type of descriptor, for a more included in five systems. Ingredient detailed food description. Otherwise the substitution (e.g. between different types name is used for describing the food, of fats) in recipes is possible in some sometimes supplemented with systems. The application of retention additional free-text description. Most factors for components (e.g. vitamins) systems can include synonyms for foods has been included in the calculations but are unilingual, although six systems with varying degrees of sophistication in eight of the systems. Although there are many difficulties in accounting for losses from other organizations, together with a and gains during food preparation, reference to their source, could be facilities to allow for these are in many included, but generally the imported situations essential. values were not stored in the original The ability to break down recipes by data format (thus, for example, not ingredient and to calculate the necessarily retaining the original contribution of each ingredient to each precision). Indication of the period for component was possible in five which an official value is or was valid, systems. Such a function is of possible in four systems, is a useful importance, e.g. when calculating the facility, e.g. for reconstructing previous contribution of various foods or food databases used for dietary surveys or groups to various nutrients. Flexibility in earlier editions of food tables. input of recipe data is desirable and The compositional data for a given ideally it should be easy to enter recipes component in a given food can be expressed in both household measures stored in various forms, ranging from and grams. It should also be possible to individual results for each individual modify a recipe, e.g. exchange analytical portion to only a single alternative liquids, fats, etc. derived or imputed value. Ideally, a Information Handling of Component database would contain mainly Data analytical data based on verified methods. Food analysis is, however, Various considerations are important in costly and requires large resources. the storage and handling of component Therefore many organizations also use data, especially in an international data from the literature and also context. These can be subdivided into alternative methods for calculating or two main areas, identification of the estimating component levels. A component reported and the details of comprehensive FDBMS should support the compositional value stored. facilities for recording literature In the systems reviewed, references and also the means to components are usually identified by a indicate the quality and method of code and/or by a name or abbreviation. derivation of a value. Two systems also use the INFOODS Half of the organizations use their tag system (11, 12) in parallel. The FDBMS for the production of food INFOODS tag system was developed to composition tables. Three systems uniquely identify components, especially directly output data formatted for in data exchange. publication of food tables, while six use Several systems have separate commercial word-processing software databases for the compilation of “raw” such as WordPerfect or Microsoft Word data (the “working” database) and for for editing the data. The underlying the “official” values used for publication principle in using a relational DBMS is of food tables, calculation of intakes that a given item of data is only entered from dietary surveys, etc. Component and stored once, but repeated as data are stored as single analytical necessary on output, for example when results for individual samples (in the a food name appears in the main food working database) or as mean values table and in the food index. derived from analytical data or data from other sources. There was generally a Other Aspects of Information lack of facilities for calculating and Handling storing measures of variation in a value Exchange of computer-readable as well as for reporting the assessed composition data was common, with text quality of a value. Component values files on floppy disk being the most be based on modern computing and widely used format. informatics techniques and standards. User friendliness is important for any This should allow a flexible design which software handling complex information made the system easy to enhance and like food composition data, a graphical modify. The interface should be user user interface (GUI) being preferable. friendly and preferably be a GUI. The This should provide the ability to design and operating environment interchange data with commercial should allow for data exchange, both software, e.g. spread-sheets, word with other applications such as spread- processing, statistical packages, since sheets and with other FDBMS. each of these support specialist facilities Currently the computing techniques which it is not practical to implement in a most appropriate to FDBMS involve DBMS. Another aspect of user relational databases accessed through friendliness is the inclusion of individual SQL, although as noted earlier these profiles for users so that their working may not prove a perfect solution and environment is customized when they potentially better alternatives may log on, e.g. by setting their preferred become available. They do, however, working language, code system. provide a basic standard, making Suggested Improvements practical collaborative developments to create transportable systems The systems users were asked to state implementable in the current hardware the three improvements that they would and software environments of many consider the most important if they were organizations. Such developments about to enhance their system. should also encourage the Responses included improved facilities implementation of compatible data for the calculation and storage of structures and the application of compositional data (including measures standard policies to the food-related of data quality), for recipe calculation data stored, key aspects in improving and the handling of multi-constituent the effectiveness of data exchange. foods, and for food classification and Until recently, most FDBMS were aggregation. Better user-friendliness in developed specially for the needs of an general was cited, as well as specifically individual organization, in part because a GUI, and there was a requirement for the development and use of common multilingual support, particularly of food software had been limited by names. A need for greater flexibility was compatibility and portability problems. expressed, for example in allowing extra Generally the systems have not been components to be included, the available for purchase on a commercial modification of existing recipes and the basis. However there are high costs handling of user-specific data (e.g. “own involved in production of high quality foods”). In general the results seem to food composition data as well as imply a considerable agreement on the development of DBMS for handling the overall facilities which are required in a data. The sharing of development costs comprehensive system when the to produce a highly functional system resources are available to implement would enable the most effective use to them. be made of the analytical data obtained. Ŷ Current Developments in Increased standardization, more International Food Databases sophisticated software tools, and international cooperation (e.g. In addition to supporting the facilities INFOODS, Eurofoods-Enfant and required for the handling of food multinational dietary and information, a modern FDBMS should epidemiological research) have stimulated interest in a DBMS capable ƒrecipe calculation system of handling high quality food data, which would allow the use of multiple ƒsystem for compositional data languages, coding, description and source references classification systems for foods and ƒmodules for print-out of food components (17–19). composition tables Recent FDBMS developments ƒsystem for handling data from include the New Zealand Food dietary surveys Composition Database, the Swedish NUTSYS system, and the EuroNIMS ƒsystem for menu planning. collaboration. The system was designed to contain New Zealand Food Composition functions that allow for: Database ƒstorage of an “unlimited” number of The New Zealand Food Composition foods, recipes, components Database is designed to handle data indication of the origin, quality, from different countries in a flexible way ƒ source, etc. of a value (20). It has been developed in-house using Advanced Revelation DBMS and ƒindication of the period during which its programming language Rbasic and is a value is valid operated on a PC network. It is well ƒindication of the origin (country, suited for easy data interchange with region) of a food other countries and institutions, while ƒindication of the method of maintaining the ease of information and preparation and processing of a data output, in both electronic and food hardcopy formats. In addition to using various facets for describing and naming ƒindication of the density, portion foods, the system includes images of weight, etc. of a food foods (color photographs), which are ƒgrouping of foods and components linked to the compositional data. The according to different criteria system is now installed in Latin America ƒuse of different names, synonyms, (INCAP in Guatemala), the South languages, codes, measures, etc. Pacific Commission (New Caledonia) and for ASEANFOODS (at INMU in ƒbreakdown of recipes to ingredients Thailand). and exchange of recipe ingredients NUTSYS ƒuse of yield and retention factors in recipe calculation NUTSYS is the name of a prototype easy communication with other FDBMS developed at the Swedish ƒ systems. National Food Administration (21). It is A data model was outlined with a the result of a project to develop a number of entities and concepts. Based modern, flexible DBMS for handling food on the model, a prototype was composition data. A number of functions constructed using the Ingres 4GL tool and modules were identified by a project Vision. About 70 programs were group that ideally should be included in generated and completed. a fully developed system. Some of the most important were: EuroNIMS ƒregisters for foods, nutrients and Development of a new system other components (EuroNIMS, European Nutrition ƒdatabase for nutrients and other Information Management System) components began after the start of the NUTSYS project. The EuroNIMS cooperation is a EuroNIMS Version 1.0 includes most result of an initiative from the Belgian functions defined in NUTSYS. Features NUBEL Foundation, responsible for the of particular interest include: management of national food ƒmultilinguality both at the user composition data, and NIMS interface and data storage levels representatives. NIMS (NUBEL Information Management System) is a ƒinternational food identification software package for management of (country, organization, sequential nutrient composition data currently and version number) being used by the NUBEL Foundation in ƒparallel management of different Belgium and was developed by coding and classification systems Logimed, a software development ƒregistration of food manufacturers company. NIMS supports some of the and distributors and of analytical key functions defined in NUTSYS, e.g. laboratories multiple languages, coding and classification systems for foods. ƒregistration of items as aggregated Representatives from about a dozen or representative foods European countries and one ƒa range of algorithms for the international organization (IARC) have calculation and conversion of values participated in the discussions on ƒrecipe storage with link to EuroNIMS. spreadsheet calculation using yield The design of EuroNIMS is based on and retention factors a client-server software architecture in which networked PCs or workstations, ƒfacilities for Langual encoding. as “clients”, access data held on a central machine, the “server” (although in practice a single, powerful PC could Ŷ Conclusions support both the client and server The use of up-to-date computing functions). The database is held on the techniques allows FDBMS currently server using a proprietary DBMS such under development to support more as Ingres or Oracle, perhaps one comprehensive facilities than hitherto, already installed by the user. EuroNIMS for example in the handling of interacts with the DBMS through an documentary information and images, ODBC (Microsoft Open Database accessed through user-friendly Connectivity) interface and this uses a interfaces. In addition to providing an single dialect of SQL. At the client end, effective operational environment for the data exchanged with the server will be compilation of food composition and processed through an object-oriented related data, the systems are DBMS to be presented to the user with increasingly being developed and a graphical user interface (GUI). In the implemented on the basis of first EuroNIMS software release international cooperation. This (Version 1.0), client machines use complements the efforts of the past Windows 3.X as the GUI and the server decade in establishing guidelines for runs under Windows NT. As a result, such data collections and should prove EuroNIMS uses 16-bit Unicode data to be an important step in facilitating the storage, but with 8-bit images in parallel use and exchange of high quality food to accommodate operating composition data. environments using current character Acknowledgments storage conventions. Ŷ The authors thank the following for kind assistance in completing the DBMS questionnaire and supplying further (10) Paul, A.A. & Southgate, D.A.T. information: D. Buss and M. Day, UK; (1978) McCance and Widdowson's M. Buzzard, USA; F. Cook, New The Composition of Foods, 4th Ed., Zealand; K. Day, UK; M. Hoke, USA; D. HMSO, London Douglass, USA; J. Ireland-Ripert, (11) Klensin J.C., Feskanich, D., Lin, V., France; J. Klensin, INFOODS; J. Lewis, Truswell, A.S., & Southgate, D.A.T. Australia; B. O'Shea, Ireland; J. Taylor, (1989) Identification of Food UK; A. Trichopoulou, Greece; A. Turrini, Components for INFOODS Data Italy; L. Valsta, Finland; A. Walker, UK; Interchange, UNU Press, Tokyo C. E. West, The Netherlands. (12) Klensin J.C. (1992) INFOODS Food Composition Data Interchange Ŷ References Handbook, UNU Press, Tokyo (1) West, C.E. (Ed.) (1989) Inventory of (13) Klensin J.C. (1991) Trends Food European Food Composition Tables Sci. Technol. 2, 279–282 and Nutrient Database Systems, (14) Klensin J.C., & Romberg, R.M. National Food Administration, (1989) Lect. Notes Comput. Sci. Uppsala 339, 19–38 (2) Loughridge, J.M., Walker, A.D., & (15) Hendricks, T.C. (1992) World Rev. Towler, G. (1993) Inventory of Nutr. Diet. 68, 94–103 Nutritional Software, FLAIR (16) Poortvliet, E.J., Klensin J.C., & Eurofoods-Enfant, Wageningen Kohlmeier, L. (1992) Eur. J. Clin. Agricultural University, Wageningen Nutr. 46 (Suppl. 5), S9–S24 (3) Crawford, R.G. (1981) J. Am. Soc. (17) Truswell, A.S., Bateson, D.J., Inf. Sci. 32, 51–64 Madafiglio, K.C., Pennington, (4) Feinberg, M., Ireland-Ripert, J., & J.A.T., Rand, W.M., & Klensin J.C. Favier, J-C. (1992) World Rev. Nutr. (1991) J. Food Comp. Anal. 4, 18– Diet. 68, 49–93. 39 (5) McFadden, F.R., & Hoffer, J.A. (18) Greenfield, H., & Southgate, D.A.T. (1994) Modern Database (1992) Food Composition Data: Management, 4th Ed., Benjamin- Production, Management and Use, Cummings, Redwood City, CA Elsevier Applied Science, London (6) Jennings, R. (1993) Using Access (19) Simopoulos, A.P., & Butrum, R.R. 1.1 for Windows, Special Ed., Que (1992). World Rev. Nutr. Diet. 68, Corporation, Carmel, IN 1–160 (7) Chen, P.P-S. (1976) ACM Trans. (20) Cook, F., Duxfield, G., & Database Syst. 1, 9–36. Burlingame, B. (1992) Proc. Nutr. (8) Codd, E.F. (1970) Comm. ACM 13, Soc. NZ 17, 204–207 No. 6 (21) Becker, W. (1993) NUTSYS — a (9) Date, C.J. (1981) Introduction to Food and Nutrition Composition and Database Systems, 3rd Ed., Information Management System, Addison-Wesley, Reading, MA National Food Administration, Uppsala Information Needs and Computer Systems

Data Identification Considerations in International Interchange of Food Composition Data

John C. Klensin

INFOODS Secretariat, United Nations University, PO Box 500, Charles St Sta, Boston, MA 02114-0500, USA

Correct use of food composition tables and databases outside the country of origin requires identification of the values in those tables. In addition to the problems of adequate nomenclature, identification, and classification of foods, problems also exist in adequate description of laboratory samples, identification of the food components being reported, and identification of the accuracy, precision and representativeness of the data values themselves. This paper reviews the procedures for identifying food components developed by INFOODS in collaboration with IUNS and their increasing use around the world. The paper then discusses the issues associated with data value identification and, in particular, methods of reporting accuracy and precision that provide maximum information to sophisticated users and compilers of food composition tables.

hen data are exchanged among countries, or even among researchers within a country, the recipients must have adequate identification, or at least Wdescription, of those data to make intelligent use of them. Some of that identification is provided implicitly, by the conventions of the field. For example, scientists doing cryogenic studies always use degrees Kelvin to report temperatures. The use of Fahrenheit, or even Celsius, degrees would be odd indeed, so the scale is almost never explicitly reported. A peculiarity of food composition data is that there are so many different aspects of the data that must be identified, and yet few established international conventions that would permit this implicitly. Adequate identification of the data variance) and, to the degree values depends on the purpose for possible, how closely the value is which they will be used, but typically related to the nutrient levels that requires describing: would be encountered in the food ƒthe food involved in terms of what it as found in nature is called, since we typically want to (representativeness). match foods-that-are-analyzed with In addition to being issues of the foods-that-people-eat or report description, many of these items bear eating on data quality both the quality of the data values themselves and, in the ƒthe food involved in terms of its presence or absence of appropriate biological or recipe origins, since we description, the overall quality of the often need to know how one food is tables or databases in which they are related to another to compare embedded. values Many other papers, including some at ƒhow the food was sampled, stored, this conference, have focused on the packaged, prepared, etc., since first of the above elements and these factors can greatly affect particularly on the issues associated magnitudes of nutrient values and with attempting to describe or classify the degree to which the values foods accurately. Accurate and reported actually represent the standardized description of sampling quantities present in the food as methods has been discussed a great eaten deal and identified as important (1, 2), ƒhow the food was handled after but there are no known specific selection but prior to analysis, since proposals for how to do this that are this, too, can greatly affect the applicable to food composition data resulting values work as actually practiced. The last three elements in the list above the ƒthe nutrient or other food identification of the data values component being reported, since a themselves are the topic of this paper. value given without indication of what it represents is useless ƒthe analysis method used and how Ŷ Identification of Food Components the “nutrient” was defined, since and Analysis Methods different methods and definitions, and even different conversion Many of the nutrient values reported in factors, where required (energy, food composition tables actually are the protein, vitamins A and E, and so result of (sometimes local) standards for on), can produce different values conversion factors, conventions about that cannot be compared directly the relationship of one value to another, or differing assumptions about the ƒthe statistical (distributional) relationship of measurable properties to properties of the value reported, bioavailability, rather than things that since it is useful to know both how can be uniquely and unambiguously similar the values are from different determined in the laboratory. For analyses and samples (precision or example, while energy measurement by putting people into calorimeters is well- commonly-reported nutrients. For understood, it is rarely done today. others, definitions have changed over Instead, conversion factors are applied time and sometimes remain to other nutrients, but those conversion controversial: a value that is simply factors differ over time and from one identified as “fiber” may be nearly country to another. So having a value in useless. And for still others, differences a food composition table labeled in methods of analysis produce “energy” is rarely sufficient to permit differences in results, i.e., not exactly comparing that value to others. Similar the same things are being analyzed. issues arise for a variety of other

Table I. Some recent additions to INFOODS Food Component Identification Tags Bromide Cyanide Fatty acid 10:1 Fatty acid 18:1 Ȧ-7 Fatty acid 18:1 Ȧ-9 Fatty acid 22:1 Fatty acid 23:1 Fatty acid 18:2 Fatty acid 18:3 Fatty acid 22:3 Fatty acid 22:5 Ȧ-6 Fatty acid 24:6 Fiber, acid detergent method, Clancy modification Fiber, total dietary; Wenlock modification Antimony Total tocopherol Fatty acid 10:1; expressed per quantity of total fatty acids Fatty acid 18:1 Ȧ-7; expressed per quantity of total fatty acids Fatty acid 18:1 Ȧ-9; expressed per quantity of total fatty acids Fatty acid 23:1; expressed per quantity of total fatty acids Fatty acid 18:2; expressed per quantity of total fatty acids Fatty acid 18:3; expressed per quantity of total fatty acids Fatty acid 22:3; expressed per quantity of total fatty acids Fatty acid 22:5 Ȧ-6; expressed per quantity of total fatty acids Fatty acid 24:6; expressed per quantity of total fatty acids Nitrogen-protein ratio

These issues were examined from also contained abbreviated names for the standpoint of food component each component-method pair. These identification a few years ago. That work names can be used in electronic data resulted in publication of a listing of food interchange and abbreviated table components and value-affecting headings and, using the terminology of methods for analysis that could be found the International Standard (4) on which in the various food composition tables the associated data interchange system and databases of the world (3). That list (5) is based, are called “tags” or “tagnames”. The list is now being [email protected]. Several incrementally updated, using an new definitions, especially of fatty acids, electronic mail distribution list as the have been added recently (see Table I). primary mechanism for suggesting and As additional nutrients of interest are reviewing new proposals. To subscribe identified and incorporated into tables, to that list, send Internet mail to food- the list is likely to be extended further.

Figure 1. Small-sample normal distributions with 5% confidence intervals It is important to note that these food substantially more information than component identification “tags” are not today's food tables and databases normative and are not associated with provide. This added detail is intended to any concept of good or desirable provide a target for improvement so that practice. There are only two no one assumes that the tags represent requirements for something being listed: as much information as might be (i) a national or regional food desired. It should also encourage the composition table compiler, somewhere, recording of more detailed information, thought that the value was important as it becomes available and is enough to include in his or her table, appropriate in the view of table and (ii) there is an adequate definition compilers, in databases. available. The second requirement was Ŷ Data Value Description waived in the original publication for commonly-occurring under-identified Just as the choice of analytic methods values (e.g., “energy” with no further can have a significant impact on the description, is tagged as ), but particular value that is produced for a future registrations are expected to be nutrient, decisions about the statistic to adequately defined. use to represent the result of multiple At the other extreme from “unknown analyses, estimates, or methods of method”, some tags have provision, imputation, may make a considerable through sub-elements and keywords, for difference in the value placed in the table. Means cannot be readily estimates of variability are even more compared to medians and, especially difficult to compare in a reasonable way. with small sample sizes, different

Figure 2. Small-sample normal distributions: medians and fences To an even greater degree than with zero and standard deviation of one. nutrient identification, the nature of Figure 1 shows “boxplots” from these numeric data values is typically not successive samples, with the white bar reported to a degree specific enough to representing the median and the shaded make them usefully comparable (6). area representing the hinges or “fourths” Values reported are typically not (approximately quartiles—the middle identified as to whether they are means, half of the data). The “whiskers” on the medians, or some other estimate of two plots to the right extend out to the location, nor is the type of data “fences” or outlier cutoffs, calculated as censoring (e.g., “outlier elimination”) 1.5 time the hinge-spread past the reported and discussed. Standard errors hinges. These types of plots, widely or variances are often reported with introduced after Turkey's “orange book” sample sizes as small as two or three. (7) and explained in detail in Hoaglin et Even with normally-distributed data, al. (8), usually provide a better overview such small sample sizes tend to yield of small-sample data than more confidence limits broad enough to make traditional scatter plots or histograms. this type of variance reporting almost It is interesting to observe with this useless. group that the second and third random The relationships between sample draws produced values all of which fell size and confidence limits are illustrated below the known population mean of in Figures 1 and 2. The plots show zero. While this is clearly a random repeated samples, using a good random event, it illustrates the dangers of number generator, from a Gaussian making statistical estimates that are distribution with the traditional mean of designed for the large sample case with only two points. The three-point samples one would hope for. Things begin to are better, as one would expect, but the stabilize at 30 points (the value at the medians fall well away from the very top is an outlier when the fences expected mean (especially in the third are used to set the criteria) and the 100 case), and the hinge spreads are quite point sample looks quite reasonable. wide relative to the standard deviations

Figure 3. Two-hundred point sample from Normal (1,1) distribution Figure 2 exhibits these same data Some food composition tables try to and plots in more traditional confidence avoid the difficulties with small samples interval terms (shown by crosshatching): by reporting the range, i.e. the maximum while some food composition tables and minimum values actually obtained. report medians for small samples rather But, since they represent extremes, than means, none that INFOODS has those values are exceptionally sensitive discovered report hinge spreads or to sampling and experimental error: it is similar robust measures. The almost impossible to create a statistical confidence intervals for the sample size estimate of the reliability of an extreme of two are artificially small due to the value. nature of the computation. But those for Worse yet, empirical evidence is sample sizes of three illustrate the accumulating that the distributions of problem: 5 per cent confidence intervals many nutrient values are asymmetric. extending out past two standard Rand and Pennington discussed the deviations of the universe being issues two years ago (9); Pennington sampled. It is nearly impossible to make provided an update and some additional statements about values with these data in a more recent paper (10). Those types of confidence intervals: they could efforts attempt to examine variability in be used to “prove” almost anything. foods, but, when quantities of nutrients Things start to become acceptable at 30 are being measured that are close to the points: the 5 per cent confidence detection level of the instrumentation, intervals actually fall within the hinge inherent censoring of trace levels also spread. causes asymmetry in the values actually obtained. Instrumentation censoring occurs corresponding frequency histogram when nutrients exist in foods at levels appears as Figure 4. In both cases, it is below the detection thresholds of the easy to observe that the distribution is measurement methods being used. For approximately Gaussian with a mean at illustration, one possible situation was 1, as one would expect. (If the negative simulated by drawing 200 points at values are bothersome, mentally shift random from a Gaussian distribution the graphs by adding about 3 to all of with mean 1 and variance 1. When the “measurements” in Figure 3 and the those points are sorted into ascending “data values” in Figure 4. That shift, of order to make an easy-to-understand course, has no impact on the analysis.) plot, they appear as in Figure 3. The

Figure 4. Frequency histogram corresponding to 200 point Normal (1,1) sample: data values Figure 5. Truncated sample from Normal (1,1) distribution: small values removed at Y=0.2

Figure 6. Frequency histogram corresponding to truncated sample: data values Now suppose that the method Since the mean value is very involved is incapable of detecting any sensitive to extreme values and the values smaller than 0.2 (marked as shape of the distribution, it may not be “presumed detection limit” in Figure 3). very useful when the data are severely One would then observe plots that look asymmetric or when trace-censoring more like Figures 5 and 6 instead of the eliminates very small values without a “true” plots in Figures 3 and 4. The new corresponding impact on the higher tail. histogram is especially interesting, since The median is often considered a cure it shows not only significant asymmetry, for fussy data problems, but asymmetry but the mean of the values actually due to tracecensoring can distort it even detected has shifted from 1.0 to about more than the mean, moving it well 1.4. A different assumption, that all the away from the subjective “center” of the undetectable values were actually at the data. theoretical minimum (somewhat below - Combination of means and medians 2 if one judges from the sample in a single table, or comparison of them, illustrated in Figure 3), would shift the is rarely appropriate, especially where mean considerably in the other central limit assumptions may not apply. direction. Their use together is usually confusing. The combination of distributions that Neither of them can easily be compared represent asymmetric natural with the more sophisticated measures of phenomena and instrumentation location that are appropriate for censoring is worse than additive in distributions that are known to be terms of the degree to which it tends to asymmetric. In particular, it is not force measured distributions into non- possible to compute a “weighted normal form. average” of a mean from one report with a median from another, even if the sample sizes are known. It is, in challenges encountered in analysis that general, not even possible to combine might bias the results. If much of this two medians this way since substantially type of information were provided, it all of the distributional information is would pose a serious challenge to discarded when half of the data are database management systems, since eliminated from each side, leaving only few of those are designed to handle a single point. When data are to be re- data with these types of used and re-evaluated by others, as in interrelationships. However, the interchange situations and reference advantages to those trying to do serious databases, and only the usual small evaluation or quality assessment of data numbers of data points have been values under consideration for use in determined by analysis or combining studies, calculation of imputed food values, it is perhaps better to list the values, or for inclusion into other tables actual values themselves, rather than would make it worth the trouble. using marginally appropriate, or Ŷ References inappropriate, statistical summaries. (1) Greenfield, H., & Southgate, D.A.T. Ŷ Tagging the Data Values (1992) Food Composition Data: As with nutrient identification, while Production, Management, and Use, doing things correctly is important, it Elsevier Applied Science, London may be even more important that (2) Truswell, A.S., Bateson, D., whatever is done be identified Madafiglio, D., Pennington, J.A.T., accurately so that a recipient or Rand, W.M., & Klensin, J.C. (1991) evaluator of data can determine if they J. Food Comp. Anal. 4, 18– 38. are suitable for his or her purposes. Just (3) Klensin, J.C., Feskanich, D., Lin, V., as it provides for identification of food Truswell, A.S., & Southgate, D.A.T. components and methods by the use of (1989) Identification of Food “tags” with exact definitions, the Components for INFOODS Data INFOODS interchange system provides Interchange, UNU Press, Tokyo tags to identify data values and (4) Standard Generalized Markup descriptions of variability. As with the Language (1986) ISO 8879 nutrient tags, these tags provide more (5) Klensin, J.C. (1993) INFOODS information than appears in any known Food Composition Data food composition table today. At the Interchange Handbook, UNU Press, same time, and again like the food Tokyo component tags, the data tags are not (6) Rand, W.M., Pennington, J.A.T., normative: tags are provided for values Murphy, S.P., & Klensin, J.C. that are reported in tables even if they (1991) Compiling Data for Food are useless from a statistical point of Composition Databases, UNU view. The system for data values Press, Tokyo extends beyond labeling of simple (7) Turkey, J.W. (1977) Exploratory measures of location (e.g., mean, Data Analysis, Addison-Wesley, median, trimmed mean, or the “X Reading, MA percent of RDI” values that appear on (8) Hoaglin, D.C., Mosteller, F., & food labels in some countries) and Turkey, J.W. (1983) Understanding variability (e.g., standard deviation, Robust and Exploratory Data standard error, quartiles, range, or Analysis, John Wiley and Sons, percentage points) to permit description New York, NY of distribution-based statistical filtering (9) Rand, W.M., & Pennington, J.A.T. procedures applied to the laboratory (1991) Proceedings of the 16th data and description of particular National Nutrient Databank Conference, The CBORD Group, the 18th National Nutrient Ithaca, NY, pp. 179–182 DDatabank Conference, ILSI Press, (10) Pennington, J.A.T. Albert, R.H., & Washington, DC, pp. 155–158 Rand, W.M. (1993) Proceedings of Information Needs and Computers Systems

Food Data: Numbers, Words and Images

Barbara Burlingame, Fran Cook, Graham Duxfield, Gregory Milligan

Nutrition Programme, New Zealand Institute for Crop & Food Research, Private Bag 11030, Palmerston North, New Zealand

Food composition databases are generally collections of numeric and descriptive data in various formats with a variety of limitations related to proper documentation. Current technologies make it feasible for databases to go beyond words and numbers now to include images and graphical representations of foods. Presently there are over 130 food images in the New Zealand Food Composition Database, ranging in size from 25 KB to 1.3 MB each, and occupying a total of about 33 MB of disk space. The process at Crop & Food Research involves digitizing photographs of the actual food samples using an optical scanner at 400 dpi resolution. Advanced Revelation 3.0, the development environment system used, does not deal with images yet, but can call DOS-based programs which convert and display digitized images in several different formats such as PCX and GIF. To date, several important uses for food database images have emerged. These include sample validation where common name could relate to several different scientific names; data validation where intensity of the orange color led to accepting ȕ- carotene values outside the expected range; food intake surveys where food descriptors were insufficient due to language or cultural differences or where children were subjects; and for international interchange of food composition data.

any problems arise as a result of poor, incomplete or ambiguous descriptions of foods listed in databases and as a result of confusion over the interpretation of Mcommonly used names for foods. Many solutions have been recommended to deal with these problems (1,2). These solutions typically rely on words, alphanumeric codes, position-specific facets, etc, and go some way toward alleviating some problems. These systems will never solve all the problems.

A picture is worth a thousand words, descriptor files could contain a field or as the old adage goes, and technologies an accompanying file of digitized images have advanced to the stage that all food or series of images so that barriers of language, culture, and the limitations beverage record represents a composite and subjectivities of our vocabularies of three different brands of powdered are minimized (e.g., how lean is lean drink mix. The packaging scanned in meat? how do you describe the depth 256 colors occupies 630 KB; this same and intensity of the color of an apricot? file compressed with PKZIP and what is a muttonbird?). (compression format by PKWare) occupies 416 KB; and as a GIF Ŷ Documentation by Image (Graphics Interchange Format) file, 93 In the New Zealand Food Composition KB. The same information contained on work, the process of documentation the packaging, when entered into the begins at the sample preparation stage. database as text, occupies a mere 30 Food samples are collected and then bytes. Table I shows disk space prepared in the laboratory. Samples are required by other images and plain text. photographed intact, raw and after More and more software products are consumer-type preparation (e.g. allowing the inclusion of digitized processed by cooking). Each sample is images. Advanced Revelation 3.0 photographed as prepared for (ARev), the development environment consumption and with a scale definition used for the New Zealand Food (e.g. metric ruler), and lately with a color Composition Database, does not index (see Figure 1). Food packaging incorporate graphics procedures. and labels are also routinely Presently, however, we associate photographed (see Figure 2). All this is scanned images of food via other done in addition to the recording of word software. ARev is programmed to call descriptors and detailed text containing DOS-based programs — we use SVGA the standard documentation details (age (Super Video Graphics Adaptor) and of sample, date of sampling, geographic Color View — which can display region, common and scientific name, digitized images stored under various physical state, processing, packaging, formats. etc). Using a number of different software The photos are then digitized into package and shareware, images stored PCX formats (IBM PC Paintbrush in PCX format can be transferred to Picture File) using an optical scanner at media as PCX or other less byte- 400 dpi (dots per inch) resolution. Much consuming formats such as GIF. This is higher resolution is available, but there important because users will have is a trade-off between resolution and different hardware and software space required to store the image. products available to them. GIF and Presently, there are 130 PCX images in TIFF (Tag Image File Format) have the NZ Food Composition Database, become industry standards, and JPEG occupying about 33 MB (megabytes) of (Joint Photographic Experts Group) with disk space. The size of the individual the ISO (International Standards files ranges from 25 KB (kilobytes) to Organization) and CCITT (Consultative 1.3 MB each. Compression would Committee on International Telegraph significantly reduce amount of disk and Telephone) backing (3) is becoming space required. popular for compressing still images for Disk space requirements vary storage. Exchanging of images will be depending on size of the image, number facilitated by having image format of colors, and the image resolution. flexibility. Various manipulations can be done to achieve efficient storage. One NZ Table I. Disk space requirements for food record images in bytes compared with non-graphical text Powdered drink Powdered drink Storage format Bread bag packaging; composite packaging of 3 .PCX (color)a 1,925,289 805,234 2,238,904 .PCX (grey)a, b 704,053 277,092 780,405 .ZIPa, c 1,240,687 471,347 1,327,914 .GIFa, d 317,399 110,505 295,897 .JPGa, e 116,613 52,925 116,716 Non-graphical text; 250 401 946 database descriptors a scanned at 16 million colors; saved as Zsoft PC paintbrush format b converted and stored by PhotoMagic as greyscale Zsoft PC paintbrush format c compressed and stored by PKZIP d converted and stored by PhotoMagic as Graphic Interchange Format e converted and stored by PhotoMagic as JPEG format Third party software will allow Ŷ The Hardware integration of compact disks and The ability to view images is dependent proprietary technologies such as Photo- on the hardware available. Images CD with food composition databases. require, as a minimum, a Super VGA Many information systems have been (Video Graphics Adapter) monitor which developed using CD-ROM technology. can display 1024 × 768 pixels in at least Conventional information retrieval 256 colors. Some images require a 1 techniques including full-text searching MB video card capable of displaying and relational databases were 32,000 colors from a palette of over 16 integrated for accessing information million colors. Graphical printers are stored on the CD-ROM for agricultural also now readily available, with as little extension information (4). as 300 dpi resolution and 24 bit color. Ŷ Lossy Compression Flopticals Lossy compression is so named Flopticals have been used already in the because redundant or otherwise exchange of images between New unnecessary data are deleted in the Zealand and INFOODS. Floptical drives compression process. Two compression are inexpensive and can use both types that can be used for lossy floptical disks and normal 3.5" floppy compression are accepted as current disks. Floptical disks are 21 MB in size, standards for still images: JPEG (Joint compared to the 1.44 MB size of Photographic Experts Group) and standard 3.5" disks. This capacity is Fractal compression. JPEG was important because some high resolution designed as a digital image images can be 20 MB and would require compression standard for continuous- fifteen standard floppies for a single tone, gray scale and color still images image. Most of the images for the New (3). It is based on a generic Zealand Food Composition Database mathematical function known as forward are between 25 KB and 1.3 MB each. DCT (Discrete Cosine Transform), Other Media which basically transforms the image into a different form which takes up less obtained values which were significantly space. Its compression is very fast, but lower. We examined details of methods, the JPEG-compacted image files are compared sampling and sample larger for the same quality than files preparation protocols, and finally compressed by other methods. Fractal resolved the problem by comparing compression uses a mathematical images of the actual samples used. The transformation called an affine map images showed that the earlier samples which identifies all patterns that can be had a much deeper, darker, orange matched even if it means rotating, color than the more recent samples (see stretching or squashing the pattern. It is Figure 3). This, of course, raises more resolution-independent. Lossy issues about the introduction and compression used by both these widespread adoption of modified compression types involves a trade off cultivars, which is often done without between information and compressed consideration of the nutritional size. Both methods intentionally discard implications. parts of the data (5, 6, 7). Food Intake Surveys Ŷ Limitations The NZ Food Composition Database There are some limitations and has over 100 beef records. For many of problems with using images in food these records, the word descriptors are composition databases. These include identical up to the facet containing ratios hardware and software restrictions of separable lean and separable fat related to storage, compression, (e.g., there are five records for beef, decompression, image resolution and rump steak, grilled, having different faithfulness to the original. Additionally, ratios of separable lean and fat: 80:20, an image cannot be searched in the 85:15, 95:5, separable lean only and same way as text files. For example, the separable fat only). Once the database bread wrapper images will identify is searched for the words grilled rump ingredients, one of which may be steak, and four records are presented, a potassium bromate. However, the judgment is required which many people graphics files of bread wrappers cannot cannot make without the benefit of be searched for potassium bromate the visual examples. It is far easier for most way a descriptor text file or a Langual people, nutrition professionals and lay file can, and therefore will not substitute alike, to select a picture of meat which for documentation by words or looks like what they would consume, alphanumeric codes. rather than to say with confidence that their grilled rump steak was 95 per cent Ŷ Uses of Images separable lean and 5 per cent separable Data Validation fat. Language differences present a Verification of information has become challenge which is dealt with by the most valuable use to date of the including an alternative names facet in effort to document by images. each food descriptor file. Still, with Sometimes we have reason to question international interchange and our own data, and images have on international trade in agricultural many occasions allowed us to make the products, some descriptors, however decision about accepting or rejecting the comprehensive and however many results of some nutrient analyses. For language translations are provided, will example, we obtained some very high never be enough. For example, the New values for ȕ-carotene in apricots in our Zealand kumara, with the alternative 1989 work. In some later work we name sweet potato, is quite unlike the North American sweet potato; the New habitats. The right nutrients in the right Zealand pumpkin is unlike the typical sorts of foods will improve the well-being North American pumpkin. The of the wildlife, including enhancing the differences seen in the nutrient potential for reproduction (9). We composition are not so surprising when experience problems even within New the physical differences are shown with Zealand, where endangered bird a picture of the food (see Figure 4). species must be relocated from their Communication barriers exist within native habitats in the South Island, to countries and with the rest of the world. small off-shore island sanctuaries. Their Language, culture, age, are just a few. traditional foods are not all available, so In a clinical setting, it is often necessary the nutrient content, as well as physical to determine the nutrient intake of similarities of the native food, are patients. In New Zealand there are considered when designing the several Polynesian languages in use, as supplementary feeding program. well as Maori and English. Children are International Interchange often subjects in nutrient intake surveys. Food images can assist overcome these What is a feijoa? What is a pukeko? communication barriers. What is a karaka berry? Most people outside of New Zealand would have no Wildlife Feeding Programs idea at all what these foods are. Even A recent project involves providing the alternative names would be useless, nutrient data to an aquarium in New as these are (almost) uniquely New Zealand. This organization will soon Zealand foods (see Figure 5). bring in penguins for exhibition which INFOODS has considered the issue have been successfully bred and reared of images in food composition for many generations on fish from the databases (10), and an image element Northern seas. We are assisting them in is included in the interchange model determining what locally available foods (11). The structure for interchange using could substitute for their present diet. the INFOODS' model requires elements The task of matching the nutrient that indicate the picture encoding type composition of our Antarctic finfish with as well as providing the actual image. A Arctic finfish would be easier if the comment element may also be used. nutrient data, a plethora of which is The images are subsidiary to the available in the USDA Standard classification element, which is the first Reference 10 (8), were accompanied by immediate subsidiary of the food images. This would be particularly element. Images associated with a cut useful where the sample numbers are of meat record might include a carcass only one or a few, where the information diagram showing the position of the cut presented does not specify different and a photograph of the cut itself. These stages of maturity, different seasons of would be included in an interchange file the year and different catch areas. as follows (where cmt means comment): Image comparisons between the two the first image itself in databases would help us assess the PCX format beef carcass physical similarities of the different diagram with cut sites species, for example, as size of the identified finfish would be relevant to the the second image itself in penguin's diet. GIF format image of In the same area of wildlife nutrition, cut our supply of nutrient data, coupled with Ŷ Acknowledgments images, will assist others attempting to reproduce the dietary aspect of native Funding for this work has come from the (7) Simon, B. (1993) PC Magazine, New Zealand Department of Health and July, pp. 371–382 Public Health Commission, and the (8) US Department of Agriculture Foundation for Research, Science and (1993) Nutrient Database for Technology. We acknowledge Standard Reference, Release 10, permission to publish the pukeko USDA, Washington, DC photograph from R.B. Morris, (9) James, K.A.C., Waghorn, G.C. Department of Conservation, New Powlesland, R.G., & Lloyd, B.D. Zealand. (1991) Proc. Nutr. Soc. NZ 16, 93– 102 (10) Klensin, J.C. (1991) Trends Food Ŷ References Sci. Technol. 2, 279–282 (11) Klensin, J.C. (1992) INFOODS (1) Pennington, J.A.T., & Butrum, R.R. Food Composition Data (1991) Trends Food Sci. Technol. 2, Interchange Handbook, UNU Press, 285–288 Tokyo (2) Truswell, A.S., Bateson, D.J., Madafiglio, K.C., Pennington, Ŷ Other Key References on Image J.A.T., Rand, W.M., & Klensin, J.C. Compression (1991) J. Food Comp. Anal. 4, 1, Barnsley, M.F. (1993) Fractal Image 18–38 Compression, A.K. Peters, Wellesley (3) Wallace, G.K. (1991) Commun. Storer, J.A. (1988) Data Compression ACM 4, 30–45 Methods and Theory, Computer Science (4) Watson, D.G., Beck, H.W., & Jones, Press, Rockville, MD P.H. (1991) Am. Soc. Agric. Engin. Netravali, A.N. (1988) Digital Pictures: 91, 7017–7024 Representation and Compression, (5) Carlson, W.E. (1991) Comp. Graph. Plenum Publishing Corporation, New 25, 67–75 York, NY (6) Simon, B. (1993) PC Magazine, Russ, J.C. (1992) The Image June 29, pp. 305–313 Processing Handbook, CRC Press, Boca Raton, FL Figure 1. Gold berries (a new cultivar) photographed with a color index. Figure 2. Food packaging for foods recorded in the New Zealand Food Composition database.

Figure 3. Apricots with different shades of orange

Figure 4. New Zealand pumkin (left): in shape, color and size is very unlike its North American counterpart (right). Figure 5. Some (almost) unique New Zealand foods for which descriptors and/or codes would never suffice. From left to right: (top) feijoas, pukeko and (bottom) karaka berries. Information Needs and Computer Systems

Computer Construction of Recipes to Meet Nutritional and Palatability Requirements

Leslie R. Fletcher, Patricia M. Soden

Department of Mathematics and Computer Science, University of Salford, Salford, Lancs M5 4WT, UK

This paper describes a microcomputer package which carries out the inverse process to dietary analysis—that is, given a list of nutrient targets the software modifies a food list so that its nutritional analysis meets those targets. The initial aim of the work was the development of a decision support system to be used by dietitians, nutritionists and other medical personnel when giving dietary advice to patients with chronic diseases such as diabetes and renal failure. This paper contains a detailed example of another application of the same software, namely the formulation of recipes for acceptable versions of traditional dishes which also meet predetermined targets for some key nutrients.

eorge Stigler's solution (1) of the classical diet problem — ensuring adequate nutrition at minimum cost — is a celebrated example in optimization and is Gfrequently mentioned in textbooks. However, it is of limited practical significance in human dietetics — the “optimal” solution contains only five foods — and, rather more importantly, the method used is rather inflexible. In particular, given the objective of minimum (monetary) cost and a range of foods from which to choose, the nutrient targets uniquely determine the solution. Adding non-nutritional constraints, limiting the quantities of particular foods in the optimal diet, for example, will ensure that the computed diet is more varied (2, 3). Nevertheless, there is still only one solution for each collection of targets and there is no convenient way of taking individual preferences into account. We have developed, and that the resulting dish is not only implemented in microcomputer palatable but also has a predetermined software, a different model of the diet nutritional composition. problem (4, 5, 6). This generates, in a Solving the diet problem is the natural way, varied diets which meet the inverse of the familiar, and needs and wishes of individuals as well (mathematically) much simpler, process as nutritional targets. In this paper we of nutritional, or dietary, analysis. The describe an application of this same software implementation of our model is model to the modification of a recipe so an extension to the dietary analysis package Microdiet (7), which is based, modified version of the recipe which in turn, on the authoritative UK food would produce a dish reasonably similar analysis data (8, 9, 10, 11). The new to lasagne verdi but with the modified software selects, from all the possible nutritional composition shown in the combinations of foods with a column labeled “Target” in Table II. predetermined nutritional composition, Although these targets are only one which is as close as possible to the illustrative, they are also intended to wishes of a client or patient. This uses a reflect recent expert advice to UK standard variant of conventional linear citizens (14) regarding desirable dietary programming (12, Chapter 14). Some modifications. algebraic details are given by Fletcher et The other columns in Tables I and II al. (5) and we will report others show the various stages in the elsewhere, particularly those relating to modification of the recipe until, at our expression of the basic optimization version F, an acceptable version was in dimensionless terms. This has proved obtained. The test of “acceptability” was to be an important technical device, the willingness of the first author's family allowing all the targets to be assessed to consider eating the resulting dish. relative to each other when seeking, for Ingredients were exchanged, introduced example, other ingredients to include in into, or removed from the recipe at a recipe, and circumvents a possible various stages in the modification difficulty mentioned in (2, p. 389). process. A blank entry, denoted by “-”, Careful formulation of the algebraic indicates that the particular ingredient model has also ensured that the solution was not considered at that stage in the to the dual problem (12, Chapter 5) optimization. The reference to “olive oil” provides significant nutritional insight. in Table II indicates that a (lower) limit was placed on this quantity of this Ŷ Method ingredient during the final stages; it is A recipe for lasagne verdi was taken convenient to list this with other, from a domestic cookery book (13) and nutrient, targets. There was no target on a nutritional analysis carried out. The the quantity of fat in the recipe and it recipe and some of the corresponding appears in Table II for illustration only. nutrient totals are shown in the indicated Had the fat content of the diet become columns of Tables I and II, respectively. too high (or too low) a further constraint As a demonstration of the capabilities of could have been added to limit this. the model it was decided to seek a Table I. Ingredients and quantities in lasagne recipe Food name Ingredient quantity (g) in recipe number A BCDEF (original) Meat sauce Onions, raw 250 250 250 250 250 250 Butter, salted 40 0 0 - - - Removed after stage C Olive oil 30 0 0 0 15 30 Beef mince, raw 300 115 100 100 100 100 Reduced to 100g after stage B Lentils, boiled - - 100 100 100 100 Introduced after stage B Haricot beans, boiled - - 100 100 100 100 Introduced after stage B Garlic, raw 5 5 5 5 5 5 Mushrooms, raw 100 100 100 100 100 100 Bay leaf, dried 2 2 2 2 2 2 Tomatoes, canned 400 747 449 416 417 417 Sugar, white 10 10 10 10 10 10 Basil, fresh 5 5 5 5 5 5 White sauce Flour, plain white 25 25 25 25 25 25 Butter, salted 25 0 0 25 6 13 Milk, cows, whole 300 300 300 - - - Exchanged after stage C for Milk, cows, semi-skimmed - - - 300 300 - Exchanged after stage E for Milk, cows, skimmed - - - - - 300 Topping Cream, double 40 16 9 - - - Exchanged after stage C for Yoghurt, low fat, natural - - - 40 40 40 Cheese, cheddar type 50 50 50 - - - Exchanged after stage C for Cheese, reduced fat, cheddar-type - - - 50 50 50 Lasagne, boiled 225 477 225 225 225 225

Although these quantities do not Ŷ Results constitute an acceptable recipe, these The steps in obtaining the displayed results and other subsidiary results from results (Table I, Table II) were as the linear programming show that other follows. Recipe A refers to the original ingredients are required to complement recipe. When the targets were set a those already there. The subsidiary software alert pointed out that the fibre result also enabled the pulses contents of garlic, bay leaf and basil introduced thereafter to be selected were recorded as “unknown”. Recipe B from amongst the variety of possible represents the smallest change to the new ingredients. quantities of the ingredients in recipe A which will meet the nutrient targets set. Table II. Nutrient targets and analyses for recipe Nutrient Target Analysis for recipe number: A BCDEF (original) Fiber (g) >20 13 20 25 25 25 25 Energy (kcal) <1500 2475 1502 1342 1383 1377 1381 Sodium (mg) <1100 1444 816 739 1008 843 907 Potassium (mg) >3500 3453 3882 3501 3499 3499 3500 Iron (mg) >16 17 16 17 16 16 16 % energy from fat <35 66 35 35 35 35 35 Quantity of olive oil (g)>15 - - - - 15 15 (in stages E and F) Fat (g) - 182 58 52 54 54 54

Recipe C shows the beneficial effect percentage energy derived from fat and on the changes to the recipe of the new on the quantity of one of its ingredients. ingredients. The subsidiary results from Other targets which the software can this stage show that the limit on the accommodate include the P/S ratio, the percentage of energy from fat is causing amino acid profile of the protein and the most of the changes made by the ratios of the quantities of two program to the ingredient quantities. ingredients. This last target is available Exchanges of existing ingredients for to ensure that, for example, the the lower-fat alternatives were made. quantities of flour and milk in a Recipe D shows the results of making computed recipe were appropriate for a these modifications to the starting recipe white sauce. and recomputing the smallest changes The model also allows targeting of which will enable the nutrient targets to nutrient density rather than nutrient be met. This still resulted in the removal totals though some modification of the of all the olive oil so a lower bound of 15 software would be required to g was imposed on this ingredient. implement this. However, in seeking to Recipe E represents the smallest minimize the overall change to changes to the modified recipe required ingredient quantities in moving to a to meet the nutrient targets with at least nutritionally acceptable recipe, the 15 g of olive oil. The consequent present software tends to maintain the reduction in the remaining quantity of total weight of the recipe approximately butter in the white sauce was judged to constant, leading to a stable relationship be unacceptable so semi-skimmed milk between nutrient totals and nutrient was replaced by skimmed milk. The density. investigation closed with recipe F which Ŷ References was deemed an acceptable version of the original recipe. (1) Stigler, G.J. (1945) J. Farm Econ. 27, 303–314. Ŷ Conclusions (2) Henson, S. (1991) J. Agric. Econ. We have demonstrated the use of a 42, 380–393. linear programming model of food and (3) Smith, V. E. (1963) Electronic nutrition in updating a recipe to reflect Computation of Human Diets, current dietary expert opinion. The Michigan State University Press, eventual recipe in the example East Lansing, MI discussed here had limits placed on the totals of various nutrients, on the (4) Fletcher, L. R., & Soden P. M. Second Supplement to McCance (1991) Diab. Nutr. Metab. 4 (S1), and Widdowson's The Composition 169–174 of Foods, HMSO, London (5) Fletcher, L. R., Soden P. M., & (11) Holland, B., Unwin, I. D. & Buss, D. Zinober, A. S. I. (1994) J. Oper. H. (1991) McCance and Res. Soc. 45, 489–496 Widdowson's The Composition of (6) Soden, P.M., & Fletcher, L.R. Foods, 5th Ed., Royal Society of (1992) Br. J. Nutr. 68, 565–572 Chemistry, Cambridge. (7) Bassham, S., Fletcher, L. R., & (12) Chvatal, V. (1983) Linear Stanton, R.H.J. (1984) J. Microc. Programming, W. H. Freeman and App. 7, 279–289. Company, New York, NY (8) Paul, A. A. & Southgate, D. A. T. (13) Allison, S. (1977) The Dairy Book of (1978) McCance and Widdowson's Home Cookery, Milk Marketing The Composition of Foods, 4th Ed., Board of England and Wales, HMSO, London Thames Ditton (9) Paul, A. A., Southgate, D. A. T., & (14) Committee on Medical Aspects of Russell, J. (1980) First Supplement Food Policy (1991) Dietary to McCance and Widdowson's The Reference Values for Food Energy Composition of Foods, HMSO, and Nutrients for the United London Kingdom, HMSO, Londo (10) Tan, S. P., Wenlock, R. W., & Buss, D. H. (1985) Immigrant Foods: Information Needs and Computer Systems

Requirements for Applications Software for Computerized Databases in Research Projects

Dorothy Mackerras

Department of Public Health, University of Sydney, NSW 2006, Australia

umerous programs have been written to access nutrient databases using words rather than numeric codes. In general, they have been directed towards the Nneeds of clinical dietitians but many of their features, such as graphs of individual dietary intakes, are irrelevant to the needs of nutrition researchers. A recent dietary survey conducted on a Pacific island highlighted some of the data entry needs of researchers. Survey participants described food intakes using standard volumes and measures (fluid oz, oz, g, mL), household units (bowl, can, slice, tablespoon), small, medium, large (glasses, coconuts, pandanus, donuts, papaya) and locally developed measures (mountain table/teaspoon, small and large tuna steaks, cm of reef fish). Neither the teaspoon nor the tablespoon matched the metric or US standards.

The abilities of two programs, A and require double coding on at least a B, from two different countries to meet proportion of forms to examine the error researchers' needs are described. Both rate. If the software allows household or these programs, or their earlier versions, common measures as food descriptors, have been available for a number of the weight of household measures of years and are widely used in their each food only needs to be “coded” respective countries. Both programs once into the program and individual were used on a Compaq Deskpro diets will not require conversion into 486/33M computer with a math grams prior to entry. Both programs coprocessor and 8 MB RAM including allow standard measures (cup etc.) to 558 KB of available conventional be used. In addition, Program A allows memory. Program A was used for three household measures to be defined surveys involving food frequency per food and the user can choose from questionnaires and diet records and 51 different words to describe the Program B in a survey gathering 24- serving. Program B allows only one hour recall data (its food frequency household measure, called either capabilities have not been used). serving, item, slice or piece, to be Facilitating data entry is important. A defined per food. Abbreviations used for major goal is to reduce the amount of data entry should be standard or coding required prior to entry. Every intuitively obvious. Program A uses SI step that has to be coded will also units (“g” for gram etc.) and abbreviations such as “oz” for ounce. By the diet files within the database and a contrast, Program B uses “a” for gram, separate database of food composition “b” for ounce etc., and this increases the information can be made for each study. likelihood of data entry errors. Both This means that some care is needed to programs allow the household prevent the file becoming too large to measures to be altered which is backup. Program B saves each diet file important in cross-cultural studies. as a separate file. This makes backup However, Program B requires the easier, but means that alterations to the operator to change each of the nutrient database (e.g. deletions) may make the values in the database if the gram files invalid. weight of the measure is changed Dietary data are often exported into a whereas this is calculated automatically statistical program. This can be used for in Program A. Neither program appears detecting errors in the data, such as to be capable of converting the assigned outliers etc., and for analysis. Data may volume of cups, pints etc. between the need to be cleaned and exported metric, imperial and US systems or of several times prior to the final analysis allowing new words to describe serving being done. Most surveys involve large sizes (e.g. mountain tablespoon) which numbers of people and many lines of would have been useful in the study. data per respondent and so batch After entry, data need to be checked processing is needed to export a large and cleaned. It is also useful to have a number of diet files into a single file, code for an unknown food so that generally with a rectangular ASCII incomplete records are flagged until the format. Program A has this function relevant coding decisions are made. As which is clearly described in the manual. diet records may contain 25 or more It took approximately ten minutes to different food items per day it is useful if export about 9000 lines of data from 45 the program allows foods to be inserted food frequency files each containing anywhere in the diet list so that the about 200 lines of data. Program B does printout matches the order of the original not have an inbuilt export function form. Program A has this feature but although the company will write a Program B does not. Neither program program on request. It took three hours appears to have a range checking to export 4160 lines of data from 369 facility; this would be particularly useful files and required reconfiguring the for food frequency information when the computer to free all the conventional list of foods can be pre-specified. memory. These functions will take It is also useful to be able to enter longer on slower computers. some other data into the dietary Particular needs of research in program. Both programs allowed long developing countries therefore include: names for the subject and the field will ƒprograms which allow flexibility in take numbers instead of letters. Thus which system of units is used items such as subject number, date of (metric, imperial, US) and which also interview, interviewer code and allow for a mixture of systems and household number could all be coded units and for words which describe and used as the subject's “name”. non-standard units Managing the database needs care, especially if the same program is being ƒprograms which allow new (local) used to analyze data from several serving descriptors to be specified different surveys at the same time. The ƒprograms which can find foods with programs had different approaches to names of only one and two letters the data organization. Program A saves long ƒfoods may need to have multiple participation in all phases of the spellings or entries in countries research, and improve the speed and where spelling is not yet quality of information processing and standardized. data output. Attention to some of these details would allow local people greater Section V

Food Composition Data and Population Studies

his Session was chaired by Dr Ian Darnton-Hill of WHO-WPRO and began with a keynote address by A. Moller entitled Food Monitoring in Denmark. This was Tfollowed by papers on Food Composition Data Requirements for Nutritional Epidemiology of Cancer and Chronic Diseases by N. Slimani, E. Riboli and H. Greenfield; Developing a Food Composition Database for Epidemiologic Studies in the Pacific Islands by J.H. Hankin, L. Le Marchand, L.N. Kolonel, B.E. Henderson, and Beecher, G.R.; The Effects of Australian, US and UK Food Composition Tables on Estimates of Food and Nutrient Availability in Australia by K.M. Cashel and H. Greenfield, and Quality Control in the Use of Food and Nutrient Databases for Epidemiologic Studies by I.M. Buzzard, S.F. Schakel and J. Ditter-Johnson. These papers are all published in full in the following pages, along with a poster entitled Construction of a Database on Inherent Bioactive Compounds in Food Plants, by A.D. Walker, J.A. Plumb, G.R. Fenwick, R. Preece, & R.K. Heaney.

Posters displayed after the Session were:

ƒRelationship Between a Dietary Measure of Antioxidant Intake and Plasma Levels, Baghurst, K.I., & Baghurst, P.A., CSIRO Division of Human Nutrition, Kintore Avenue, Adelaide, SA, Australia. ƒThe UCB Worldfood Dietary Assessment System Utilizing the UCB International Minilist, Calloway, D.H., & Murphy, S.P., Department of Nutritional Sciences, University of California, Berkeley CA, USA. ƒUse of the Extended Table of Nutrient Values to Assess Nutrient Intakes of Restrained and Disinhibited Women, Champagne, C.M., & Williamson, D.A., Pennington Biomedical Research Center, Louisiana State University, Baton Rouge, LA, USA. ƒOrganochlorine Intake of Victorian Infants from Maternal Milk, Donohue, D.C., Quinsey, P.M., & Ahokas, J.T., Key Centre for Applied and Nutritional Toxicology, RMIT University, Melbourne, VIC 3001 and National Food Authority, Canberra, ACT, Australia. ƒAflatoxin M1 in Human Milk Samples for Australia, El-Nazemi, H.S., Ahokas, J.T., Donohue, D.C., & Neal, G.E., Key Centre for Applied and Nutritional Toxicology, RMIT University, Melbourne, Australia and Medical Research Council, Toxicology Unit, Carshalton, UK. ƒGraile: a Database for Australian Grain Legumes, Horton, J.D., Petterson, D.S., & Mackintosh, J.B., Cowirrie Computing, Relbia, TAS; Department of Agriculture, South Perth, WA; The University of Western Australia, Nedlands, WA, Australia. ƒA Short Questionnaire and Qualitative Fat Index for the Assessment of Fat Intakes on the Basis of the FINMONICA 1982 Survey, Kempainnen, T., Rosendahl, A., Nuutinen, O., Ebeling, T., Pietinen, P., & Uusitupa, M., Departments of Clinical Nutrition and Medicine, University of Kuopio, National Public Health Institute, Helsinki, Finland. ƒHeavy Metals in Taiwanese Diets, Lee, Y.S., & Chou, S.S., Department of Food Science, University of the District of Columbia, Washington, DC, and National Laboratories of Foods and Drugs, Taipei, Taiwan. ƒVariability in Macronutrient Contents of Selected Cereal Products Between Production Batches and Analytical Laboratories, Mugford, D.C., Bread Research Institute of Australia, Inc, North Ryde, NSW, Australia. ƒDatabase on Asian Sensory Preferences, Food Markets and Culture, Ng, F., Bell, G., Prescott, J., Waring, J., & Gillmore, R., CSIRO Sensory Research Centre, Division of Food Science and Technology, North Ryde, NSW, Australia. ƒImpact of Reductions in Fat Content of Australian Pork on Fat Available for Consumption in Australia, O'Dea, K., Mann, N.J., Sinclair, A.J., & Barnes, J.A., Department of Human Nutrition, Deakin University, Geelong, VIC3217 and Australian Pork Corporation, 174 Pacific Highway, St Leonard's, NSW, Australia. ƒComparison of the Use of Australian and UK Food Composition Tables for Estimating Nutrient Intake, Record, S.J., & Baghurst, K., CSIRO Division of Human Nutrition, Kintore Avenue, Adelaide, SA, Australia. ƒThe Link Between Defence Food Intake Studies and a Relational Database, Waters, D.R., DSTO, Materials Research Laboratory, Food Science Branch, Scottsdale, TAS, Australia. ƒEffect of Changes in the Swedish Food Database on Nutrient Estimates from a Food Frequency Questionnaire, Wolk, A., Becker, W., Ohlander, E-M., & Bergstrom, L., Cancer Epidemiology Unit, Uppsala University Hospital, and the Nutrition Division, National Food Administration, Uppsala, Sweden. Food Composition Data and Population Studies

Food Monitoring in Denmark

Anders Møller

National Food Agency of Denmark, Informatics and Computer Section, 19, Mørkhøj Bygade, DK-2860 Søborg, Denmark

In 1983 the National Food Agency established a food monitoring system in order to follow the content of nutrients and contaminants in foods in a systematic manner. When the data from this system are combined with the results from the national food consumption survey of 1985 it is possible to calculate the intake by survey participants of both nutrients and contaminants. Also, it is possible to make an estimate of the maximum intake of food additives. Calculations like these are used as a basis for regulating food fortification and the use of food additives as well as establishing safe levels for the content of contaminants in foods.

ood plays a vital role in the Danish economy; the Danish food production amounts to more than US$ 17 billion. Food exports exceed US$ 9 billion, corresponding to Faround 35 per cent of the country's total earnings from export of industrial manufacture.

In the course of a year each person manufactures and to promote healthy in Denmark consumes an average of dietary habits. one ton of food, which makes food a The National Food Agency comprises central aspect of daily life. In addition, two scientific institutes and four awareness among Danes about the administrative divisions. The institutes, food they eat is increasing. the Institute of Toxicology and the In Denmark national food legislation Institute of Food Chemistry and is the responsibility of three different Nutrition, represent the specialist ministries, the Ministry of Health, the scientific knowledge which forms the Ministry of Agriculture and the Ministry basis for the Agency's administration of of Fishing. Responsibility for the the Food Act and the provision of General Food Act of 1973 lies within the nutritional guidance to the general Ministry of Health, and the executive public. functions are carried out by the National Ŷ The Food Monitoring System Food Agency. The Agency's objectives are to protect consumer health, to The contents of both nutrients and protect consumers against misleading contaminants in foods on the Danish information/fraud, to ensure reasonable market have been analyzed by the conditions for retail stores and National Food Agency and associated period. The results of analyses of the laboratories for several decades. food monitoring system are reported Due to the increasing focus on diet continuously. and health issues, as well as a desire to Every five years a major evaluation of ensure that chemical analyses within the the results of the preceding five-year individual working areas of the Agency period takes place. The first report for a were linked together as a whole, the complete five-year period was published Agency established a food monitoring in 1990 covering the years 1983–87 (2). system in 1983. The foundation of the Ŷ Selected Areas of Monitoring food monitoring system was described in a proposal prepared by an internal The food monitoring system covers working group at the Agency (1). nutrients as well as heavy metals/other The basic objectives of the Danish trace elements, nitrate, pesticides and food monitoring system are to: PCB in selected foods (Table I). With the exception of nutrients, the content of ascertain whether, over the long ƒ these compounds/substances in most term, changes occur in the content cases originates from influences from of desired and undesired substances the external environment. Components in Danish foods. to be monitored are carefully selected ƒcombine such changes with changes on the basis of the existing knowledge in eating habits about their nutritional importance or ƒassess whether changes expose the toxicity, their occurrence in foods, and Danes to nutritional or toxicological the actual consumption of these foods. health hazards Microbiological studies and examinations of radionucleides have obtain background material and a ƒ until now not been included in the food basis for decision-making to remedy monitoring system but will be reported in any problem that may have arisen. the future. A proposal for a ƒTherefore, the practical work with microbiological food monitoring system the Danish food monitoring system is in preparation. implies: In the first five-year period (1983– ƒwatching the content of nutrients and 1987) 10,060 samples were analyzed, in contaminants in selected foods the second period (1988–1992) 9, 341 closely samples, and for the third period (1993– 1997) a total of 8,150 samples are to be watching the consumption pattern of ƒ analyzed. the Danish diet closely. Table II gives more details about the The food monitoring system is sampling with regard to nutrients and designed to supply data about the contaminants. changes over time in the contents in The expenses of the food monitoring foods of nutrients and contaminants. It is system amount to about US$ 3.3 million also designed to be linked with the data per five-year period. During each period from the food consumption survey in a total of approximately 65 full-time order that the nutrient and contaminant persons are devoted to the project, i.e. intake of the population can be approximately 13 full-time positions per calculated. year at the four (formerly five) regional As changes in the content of desired laboratories, as well as several full-time and undesired substances as well as positions at the National Food Agency. changes in dietary habits occur slowly, the monitoring system will run for a long Table I. The elements of the food monitoring system Food category Nutrients Trace elements and Pesticides and PCB nitrate Fruit and vegetables Fat, protein, ash, dry As, Cd, Cr, Hg, Ni, Pb, matter, fiber, vitamin C Se (tomatoes: glutamic In vegetables also acid) nitrate Cereal products Fat, Protein, ash, dry As, Cd, Cr, Hg, Ni, Pb, matter, fiber, vitamins Se B1, B6, Ca, Fe, K, Mg, Na, Zn Milk, milk products and Fat, protein, ash, dry In eggs: As, Cd, Cr, Hg, DDT, dieldrin, HCB, Į- eggs matter, fatty acids, Ni, Pb, Se HCH, ȕ- HCH, lindane vitamins A, B1, B2, Ca (Ȗ- HCH), Fe, K, Mg, Ma, Zn, I heptachloroepoxide, PCB Fish Fat, protein, ash, dry As, Cd, Cr, Hg, Ni, Pb, DDT, dieldrin, HCB, Į- matter, fatty acids, Se HCH, ȕ-HCH, lindane vitamin D (Ȗ- HCH), heptachloroepoxide, PCB Meat Fat, protein, ash, dry matter, Beef, chicken, pork: Fe, Mg, Zn, vitamins B1, B2, B6 OffalFe, fat, protein, ash, dry As , Cd, Cr, Hg, Ni, Pb, matter Se Animal fat DDT, dieldrin, HCB, Į- HCH, ȕ- HCH, lindane (Ȗ- HCH), heptachloroepoxide, PCB HCB: hexachlorobenzene; HCH: hexachlorcyclohexane; PCB: polychlorated biphenyls

Table II. Number of samples examined 1983–1987 1988–1992 1993–1997 Nutrients 1300 920 (900) Trace elements 4100 3285 (2500) Nitrate 1200 370 (750) Pesticides* 3200 3265 (2800) Mycotoxins 260 1500 (1200) * The National Food Agency's control programs include approx. 1200 samples/year covering about 120 different pesticides Figure 1. Potatoes, daily intake, Danes, 15–80 yrs

Criteria for selection of samples and system, see Table I, have been selected analysis on the basis of existing knowledge of their toxicity and occurrence in foods Nutrients. The nutrients included in the compared with food consumption. In the system have been selected based on case of nickel, arsenic, chromium and one or more of the following criteria (3): selenium there has also been a desire ƒthe daily intake of the nutrient in to gain more knowledge about their Denmark is lower than or around the occurrence in Danish foods. recommended level, either for the Only vegetables have been selected population as a whole or for to be monitored for nitrate. The specially exposed groups of the concentration of nitrate in fruit and other population foods is so low that they have only minor significance for human intake of ƒthe nutrient is present only in few foods nitrate. Pesticides and PCB. For a number of ƒthe nutrient shows stability problems years combined control and monitoring ƒthe nutrient is added to one or more analyses have been carried out on foods, either compulsorily or organochlorine pesticides and PCB in voluntarily. fish, meat, eggs, and milk and milk The analyses have been given products. Since 1983 these studies priority based on an overall evaluation have been included in the monitoring according to these criteria and resulted system. The analyses comprise in the substances listed in Table I. persistent organochlorine pesticides. Trace Elements and Nitrate. The Among industrial chemicals, at present trace elements which are included in the the system includes only the polychlorinated biphenyls (PCB).

Figure 2. Fish, daily intake of cooked edible portion, Danes, 15–80 yrs

Ŷ The Danish Food Composition Data ƒIn 1985 the National Food Agency of Bank Denmark carried out a nationwide food consumption survey (4, 5). The After careful evaluation of the results of objectives of the survey were: the analyses the relevant food monitoring data are transferred from the ƒTo identify population groups which Agency's laboratory information are at risk from a nutritional point of management system and stored in the view Danish Food Composition Data Bank. ƒTo evaluate the significance of The food monitoring system is a fortifying foods with nutrients substantial data source, due to the To estimate the exposure of different systematic and continuous flow of new ƒ population groups to contaminants data from the monitoring system into the and food additives databank. At present the databank comprises ƒTo identify foods which contribute information for about 2000 foods on the significantly to the nutrient intake in Danish market. In the databank different population groups information is collected on 255 different ƒTo contribute to studies of the compounds. relationships between diet, health Ŷ The National Food Consumption and disease. Survey Figure 3. Percentage of dietary energy from fat, Danes, 15–80 yrs

The survey included 2242 persons, balance sheets, household budget 15–80 years of age. They constituted a surveys, and other food consumption representative sample of the adult surveys. The conclusion is that the Danish population. The participants in average consumption found in the the survey were interviewed about their survey is very close to the real intake food consumption habits using a dietary except for a few foods and beverages, history method, which was developed such as sugar and alcohol. particularly for this survey. The dietary history method used in The dietary history method gives the present survey enables the ranking information about the usual diet of an of individuals according to their intake of individual during an extended period of foods, nutrients, contaminants and other time. There is no doubt that the method known constituents of food. tends to overestimate regularity in the Ŷ The Food Intake eating pattern. The method itself encourages the participants in the Two different types of food intake survey to emphasize usual food distributions were identified, one for consumption, because it is easier to foods eaten by everyone, and the other remember the usual meal pattern than for foods consumed by only some all the unusual events which interfere sectors of the population. As examples, with the habitual food intake. The results the distribution of the intake of potatoes of the survey are, however, in excellent and fish within the adult population is agreement with the results from other shown in Figures 1 and 2. similar data sources, such as food Figure 4. Daily intake of vitamin A, Danes, 15–80 yrs

The shape of the intake curve for Nutrients potatoes is typical of foods that are consumed by practically everybody. Figures 3 to 5 are examples of the These foods are cereals, white bread, results of the calculations of nutrient rye bread, coarse vegetables, meat, levels. Figure 3 illustrates how the fat- poultry, separable fats and eggs. Figure energy-percentage of the diet of Danish 2 illustrates the shape of the intake women and men is distributed. The fat- curve for foods and beverages which energy-percentage seems to be very are consumed by some people only. high when compared to the Other examples are cheese, soft drinks, recommended level of 30 per cent of the beer and tea. dietary energy from fat. In fact almost all Danish adults seem to eat a diet that is Ŷ The Intake of Substances from higher in fat than the recommendations. Foods As a result of this, the National Food The Agency has developed computer Agency has intensified its information software that allows the data from the campaigns on good eating habits with food consumption survey to be short advertisements on Danish combined with the data from the food television, written material for schools composition databank. Thus etc. Recently an extensive campaign for computation of the intake levels of reducing intake of fats, especially butter nutrients and contaminants of the and margarine, was launched. The individuals who participated in the food motto of this campaign is “Scrape your consumption survey gives an estimate bread”. of the distribution of intake within the adult population. Figure 5. Daily intake of iron, Danes, 15–80 yrs

Figure 4 indicates that a very large above the recommended levels. For iron percentage of Danish women and men it was shown that the contribution of consume more vitamin A than the fortification (inorganic iron) taking recommended level of 800 and 1000 bioavailability into consideration was ug/day, respectively. less than 10 per cent of the Figure 5 shows that the same is not recommended intake. the case for iron. The iron intake of most A new food consumption survey is at women in Denmark seems to fall below the planning stage. A pilot study was the recommended level of 12–18 carried out in the autumn of 1993. The mg/day. main food consumption survey will take The results for nutrient intakes have place in 1994. been used to evaluate the nutritional Contaminants importance of the fortification of foods. As a result of this evaluation the Trace Elements. Table III shows the obligatory fortification of flour and calculated intakes of mercury, cadmium, margarines with vitamins and minerals lead and arsenic from foods. For was abolished in 1987, because the mercury, cadmium and lead it appears contribution of fortification to the total that intakes from foods are well below nutrient intake was shown to be either the PTWI (Provisional Tolerable Weekly unnecessary or negligible. The Intake) proposed by Joint FAO/WHO evaluations showed that intakes of the Expert Committee on Food Additives vitamins and minerals in question (6,7). Special interest has been devoted (vitamins, riboflavin and thiamin; to the lead content in foods since the minerals, calcium, phosphorus) were dominant source of lead contamination of foods, especially vegetables and crops, is lead emitted from motorcars running on petrol.

Table III. Intake of trace elements, all values in µg per person Daily intake Weekly intake Trace element Meanp0.50 p0.90 p0.95 Mean p0.95 PTWI Mercury 7 5 12 15 55 160 350 Cadmium 20 18 28 32 137 250 490 Lead 42 40 66 76 297 532 3500 Arsenic 118 87 233 313 PTWI: based on a body weight of 70 kg p0.50, p0.90, p0.95: 50th, 90th and 95th percentiles

During the last decade the lead where high levels are most likely. There content of petrol has been lowered are too many gaps in our present substantially and unleaded petrol knowledge about the content in other introduced on the Danish market. The foods to allow us to make a calculation influence of this is clearly seen in the of the total exposure of the participants decreasing lead content in many foods, in the survey to pesticides from foods. e.g. offal, beverages (wines in Food Additives particular), berries, certain types of fruits, greens, oat, rye, wheat, rye bread The use of food additives in Denmark is and cod liver. In other food groups no regulated by the National Food Agency change was discernible. These groups through the so-called positive-list (8). comprise meat, imported fruit, roots and The list specifies the maximum amount tubers, cabbage, certain vegetables and of a food additive that can legally be fish. used in individual foods. Although the For arsenic the total intake of organic maximum amount of food additives that and inorganic arsenic is shown in Table might be used is known, there is no III. No PTWI has been proposed for complete picture of the actual use by the arsenic, but a value of 140 µg/day has food industry. Therefore, no calculation been established as PMTDI (Provisional can be made of the actual exposure of Maximum Tolerable Daily Intake) by a the population to food additives, but only group of experts under FAO/WHO in an estimate of the maximum exposure, 1983. This value applies to inorganic which would occur if food producers arsenic only. Most of the arsenic in the used all the permitted food additives in Danish diet originates from fish. This is their maximum amounts. The calculated organic and considered non-toxic to exposure will in all cases be higher than humans. Arsenic is therefore not the actual, since most foods are considered to be a problem in Danish manufactured without making full use of foods. all permitted additives. The calculated Pesticides and PCB. So far it has not maximum intake is, however, of been possible to calculate the exposure considerable interest from a regulatory of the population to pesticides and point of view. It allows a check to be organic pollutants from food in the same made as to whether the limitations that manner as for heavy metals. The have been introduced in the use of analyses of pesticide residues in foods additives in individual foods are realistic have been concentrated on those foods in relation to the ADI-values. Figure 6. Maximum erythrosin intake, Danes 15–80 yrs (ADI: 625 µg/kg body weight/day)

Figures 6 and 7 show, as examples, the “positive-list” is applicable. The the intake distribution of the calculated method is the so-called budget-method maximum intakes of erythrosin and (9). It is not a scientific method, but a benzoic acid/benzoates. From the practical administrative tool to predict figures it appears that while the the maximum intake of a food additive. maximum intake of benzoic The main assumption in the budget- acid/benzoates permitted according to method is that the maximum daily the positive-list is well below the ADI- consumption for an adult is 1.5 kg of value, the same is not the case for foods and 6 L of beverages and water. It erythrosin. is also assumed that only half the foods For erythrosin, however, the ADI- are industrially processed and thus value has been reduced from 625 to 50 contain food additives. As far as the g/kg body weight per day recently (7). liquid intake is concerned the The distribution curve in Figure 6 was assumption is made that only 25 per obtained with the erythrosin levels cent of beverages contain food permitted at the time of the survey. additives. The ADI can be divided Therefore, intake levels have to be between solid and liquid foods compared to the higher ADI-value of according to technological needs. If the 625 µg/kg. required level is too high compared to As a whole, the calculations based the ADI available the additive may be on actual food intake have shown that limited to either solid or liquid foods or to the method used to determine the certain groups of foods. amounts of food additives permitted in Figure 7. Maximum benzoic acid/benzoates intake, Danes 15–80 yrs (ADI: 625 µg/kg body weight/day)

The calculations show that each of problems are monitored less often, while these assumptions is reasonable for 90 newly detected problems are taken up per cent of the adult population. The for inclusion in the system. type of calculations illustrated in Figures Ŷ References 6 and 7 confirm that the budget-method is a reliable tool in the administration of (1) National Food Agency (1984) food additives. Establishment of a Food Monitoring System, Statens Ŷ Conclusion Levnedsmiddelinstitut, Soborg The Danish Food Monitoring System (2) National Food Agency (1990) Food has proven to be a valuable tool for Monitoring in Denmark: Nutrients identifying nutritional or toxicological and Contaminants 19831987, areas where action has to be taken, as National Food Agency of Denmark, well as actions in the area of food Soborg administration and regulation. (3) Haraldsdottir, J., Heidemann, F. & Systematic monitoring of foods is Leth, T. (1982) Establishment of a necessary, also in the future, to ensure Monitoring System for Nutrients in the supply of healthy foods to the Foods, Statens Danish population. As dietary habits are Levnedsmiddelinstitut, Soborg constantly changing, it is important to (4) Haraldsdottir, J., Holm, L., Jensen, adjust the monitoring system on a J.H. & Moller, A. (1986) Dietary continuing basis so that areas without Habits in Denmark 1985, 1. Main Results, Publication No. 136, (7) FAO/WHO (1989) Thirty-third report Levnedsmiddelstyrelsen, Soborg of the Joint FAO/WHO Expert (5) Haraldsdottir, J., Holm, L., Jensen, Committee on Food Additives, J.H., & Moller, A. (1987) Dietary WHO Technical Report Series No. Habits in Denmark 1985, 2. Who 776, Geneva Eats What?, Publication No. 154, (8) National Food Agency (1988) Levnedsmiddelstyrelsen, Soborg Fortegnelse Over Godkente (6) FAO/WHO (1972) Sixteenth Report Tils'tningsstoffertil Levnedsmidler, of the Joint FAO/WHO Expert Levnedsmiddelstyrelsen, Soborg, Committee on Food Additives, Publication No. 171 WHO Technical Report Series No. (9) Hansen, S.C. (1979) J. Food 505, Geneva Protect. 5, 429–434 Food Composition Data and Population Studies

Food Composition Data Requirements for Nutritional Epidemiology of Cancer and Chronic Diseases

Nadia Slimani, Elio Riboli

Programme of Nutrition and Cancer, WHO International Agency for Research on Cancer, Lyon, France

Heather Greenfield

Department of Food Science and Technology, University of New South Wales, Kensington NSW 2033, Australia

Nutritional epidemiology is concerned with, among other things, establishing the association of diet and disease. The principles of nutritional epidemiology drive the requirements for nutrient databases for valid measurement of dietary exposure. The potential impact of random and systematic errors in food composition data on computation of nutrient intakes in prospective multi-center studies is discussed and the needs for modeling studies and time-related databases highlighted.

utritional epidemiology is concerned with, among other things, establishing the association between diet, health and disease. Establishing a relationship relies on Nmeasurement of exposure to a dietary factor and estimating the absolute (incidence) or relative risk (odds ratio) of having a given disease associated with a given level of exposure. The categorization in quantiles of a population distribution represent one type of classification of subjects that can be used usually with three to five classes of exposure from lowest to highest. The establishment of a statistical association relies on the absence of bias in all of the observations including the dietary observations. Systematic errors (bias) have to be excluded as they could affect classification of disease cases and control subjects unequally. Random errors, which have an equal chance of occurring in affected and unaffected individuals, thus affect the classification process equally for all groups. Nevertheless even random errors can affect the validity of a study's findings by distorting the estimation of relative risk towards the null value of 1 and increasing the variance of observations, thus blurring true relationships. Procedures exist for minimizing bias, controlling measurement errors, and preventing misclassification. These procedures rely to an important degree on collecting data according to clearly defined, rigorous, standardized protocols for all aspects of the scientific observations (1). Epidemiological investigations of the rejects statistical outliers and may also role of diet in cancer and other chronic include some subjective judgement in diseases to date have revealed in many acceptance of the individual values from cases a weak association (2), but even which the mean and standard deviations such a weak association is potentially of of values are computed. However, great biological significance due to the Beaton's analyses give valuable large numbers of people likely to indications of the low impact of random experience high or low exposures to biological errors on computed nutrient dietary factors since everyone eats. intakes especially in diets composed of There is a strong case, therefore, for large numbers of foods. It would be continuing work to establish the dietary useful if such analyses could be links. extended to investigations of systematic The major method of measuring bias in food composition values, a topic dietary exposure has been the collection which is of great concern in of data on food intakes and converting epidemiology. these data to nutrient intakes by means Ŷ Nutrient Databases for Nutritional of food composition databases. While Epidemiology there has been progress in understanding the errors which can In examining nutrient database options arise in measurement of food intake, the for any study in nutritional epidemiology, role of errors in conversion of these it is important to consider the aims and intakes to nutrient intakes is not so well methods of the discipline particularly in investigated. In fact, much published order to avoid the ad hoc selection of a literature about food composition database portrayed as the “usual databases in epidemiological studies is approach” in the model described by descriptive rather than analytical (3, 4, Buzzard and co-workers (Table I) (8). 5,) and Willett (3) noted that no formal The first step is the definition of the analysis had been done of the impact of nutritional hypothesis which is to be variability (systematic) in nutrient tested in the study. The several different content of food in nutritional possible approaches in descriptive and epidemiology. The only analytical study analytical nutritional epidemiology appears to be that of Beaton (6) who require separate and lengthier investigated the impact of biological consideration both in terms of the variability in the composition of foods (a nutrient composition database and the source of random error) on nutrient dietary methodology. However, this intakes calculated from two different discussion will be restricted to the one-day intakes (sample diets) by use of concerns associated with a prospective the US nutrient composition tables for multicenter study which it is hoped will which measures of dispersion are given shed considerable light on the (7). Some caution is needed in relationship of diet to cancer and other interpreting these findings given the chronic diseases. nature of food table compilation which Table I. Ideal versus usual approach to planning a diet study Ideal Approach Usual Approach Identify nutrients of interest Select data collection method based on cost and ease of administration Determine level of specificity of food Collect dietary data prior to selecting an descriptions required to assess the nutrients of appropriate database for nutrient analysis interest Select a data collection method that will accommodate the desired level of specificity Develop or modify an existing database to: Select an existing nutrient nutrient database without evaluating: Ŷ accommodate the desired level of specificity Ŷ the level of specificity of foods included in the database Ŷ provide complete, accurate, specific, and Ŷ the completeness, the accuracy, the specificity updated values for the nutrients of interest or the currency of the nutrient database Adapted from Buzzard et al. (8)

database used to analyze the food Ŷ Database Requirements for Prospective Studies intake data. These are discussed below. The EPIC study (European Prospective Need for a Tailored Database Investigation into Cancer and Nutrition) Nutritional epidemiology requires the has the advantages of a large study database to be specifically tailored to population living in several geographical the actual foods reported consumed, areas with different dietary patterns and and according to the dietary method cancer incidences (northern and used, to increase the accuracy of southern Europe), with an appropriate nutrient intakes computed; thus the age and socioeconomic distribution, concept is of a user database which which has the power to establish valid may need to draw on several reference relationships between diet and even database in its compilation. The relatively rare cancers (2). important distinction must be drawn A multi-country prospective study between this approach and other requires within-cohort and between- applications in which it is acceptable to cohort analyses, and further has tailor the dietary intake obtained to the implications for the period over which database (i.e. matching a food with the data will be collected (changes in most similar food in the database). In environment and observers) as well as epidemiology it is the reverse, a point the volume of data to be gathered. The not always understood in the field. For between-cohort analysis is particularly example, a comparative study of two important to determining the impact of databases (9) found a new national large variations of diet on disease database deficient since it did not have incidence since many within-cohort values for lean meat unlike the foreign studies are of national populations with database used previously. This was relatively homogeneous dietary intakes. despite the availability of the published Any prospective epidemiological “lean only” data for the local meats in study planning to analyze and compare the literature which the authors cited dietary information for several countries (10) and which could have been used to will need to take into account several tailor the local database in order to considerations for the food composition avoid over-estimation of the fat intakes of study populations to which the decrease in country B produced a 50 database was subsequently applied. per cent decrease in the difference in Needs for Local Data for Local Foods total fat intake between countries. So far as within cohort analyses are Nutrient composition data for the local concerned, when the fat content of meat foods as consumed in the specific was varied by up to 50 per cent in a countries will be necessary for a multi- mathematical simulation by computer country prospective study of nutrition. there was no effect on classification of To use non-indigenous data, particularly subjects (as expected), but also no for staple foods, could suppress the effect on classification when the content effects of an important potential source of fat in milk and milk products was of dietary variability. For example, fat lowered by 30 per cent at the same time from meat is often of interest in surveys that the fat content of meat was of diet and degenerative disease. increased by 30 per cent. This result However, meat is a food which is known was unexpected and was not affected to vary dramatically in its fat content by any correlation between meat over time and between countries. When consumption and consumption of milk new compositional data for meat and milk products. However, different became available in one particular results can be expected in populations country and was found to be up to 50 with different dietary habits. per cent leaner than the previous data Further simulations on other nutrients set (data origins obscure) (10), the total and other food groups could act to fat available from the food supply daily improve estimates of error in nutrient per capita dropped from 145 g to 119 g intakes obtained using databases. It (i.e. 19 per cent), the total fat available could be hypothesized that other foods daily per capita from meats fell by 27 g may be less heterogeneous between (from 52 g to 25 g) and the total fat countries (particularly some fruits and available daily per capita expressed as vegetables) and therefore local data per cent energy available fell from 37 would not be needed but the final per cent to 33 per cent, in other words, it answer cannot be known until some fell below the (then) dietary target (11). analytical data for such foods are The ratio of vegetable to animal fat available and the hypothesis tested. increased from 0.55:1 to 0.74:1. Hence it can be expected that use of meat fat Nutritional Epidemiology Needs data which are “too high” for one Time-Related Data country, and “too low” for another Foods change in composition over time, country (in relation to the “true” values) particularly when there are changes in could either obscure a true difference in breeding of plants and animals, changes fat intakes between the two countries or in feeding regimes, and changes in artificially create a difference where preparation of foods for retail sale (e.g. there is none. butchering). Changes in regulations Some simple tests using pilot dietary affecting foods (e.g. the introduction of data from two countries participating in mandatory fortification with a nutrient the EPIC study varying the fat content of such as thiamin) also have the potential meat showed some effect on the to alter their composition over time. difference between national samples (as Food tables and databases, on the other expected). In fact, a 10 per cent hand, tend to be up to a decade out of increase in fat content of two food date, given the delays experienced groups combined (meats, milk and their between collecting and analyzing the products) in country A and a 10 per cent food, the delays in publishing the data and the delays in compiling data into of laboratory work and food intake re- databases. There is potential therefore analyses could be expected to shed for any database to be out of phase with additional light on dietary relationships the dietary intake data to which the with the cancers and other diseases database is applied (12). This point is of which emerge during the 10–15 year importance for studies of trends over course of the observations of the time within a country. prospective study. Such re-analyses of A recent study (13) examined the biological samples such as blood have sodium and potassium intakes of a already been provided for in the EPIC group of 27 females using urinary study (2). excretion and several dietary methods. Needs for a Complete Database One part of the study involved computing the dietary intake of sodium The nutrient database will need to be and potassium by means of analyzing complete (i.e. include all relevant foods sets of 24 hour food intake data against for a given nutrient) if nutrient intakes three subsequent versions of the are not to be underestimated; this is a Australian nutrient database NUTTAB point which has been well-accepted (9) (NUTTAB89 vs NUTTAB90 vs and some data have been provided, NUTTAB91/92) (14, 15, 16). Note that e.g., by Stockley (17) who reviewed the food intake data sets were studies of error associated with missing constants. The values for sodium and values in databases, citing potassium in foods had changed over underestimates of B vitamin intake this period, but since some foods ranging from 1.5 per cent to 14.3 per increased in sodium or potassium by cent, and recoveries of only 69 per cent 10–100 per cent, while others of total polyunsaturated fatty acids decreased by 10–100 per cent, the net analyzed in duplicate diets as opposed effect was of no change in group means to computed nutrient intakes, improving for total intake from foods of these two to 89 per cent when the missing fatty minerals; the changes in food acid values were inserted. Also composition canceled each other out. mentioned were the new starch values for UK potatoes ranging from 11 g/100 g Needs for Analytical Data for Foods to 23.0 g/100 g, according to cultivar, The lack of compositional data for with the average value (weighted by indigenous foods consumed in some of tonnage) being 17.0 g/100 g compared the southern European countries is a to the value of 20.3 g/100 g given in the soluble problem for EPIC. Many useful food tables then current. Starch intake data are being produced in local from potatoes may therefore be only 60 laboratories, and once they have been per cent of that computed from the assessed and scrutinized can be used tables. It should be relatively easy to to update and amend the databases design simulations to investigate the used to analyze the food intake data. effects of missing data using pilot Re-analyses done in the future will studies of dietary intake carried out for refine the estimates of nutrient intakes the EPIC investigation. calculated (rather than alter them Nutrient values should be, as far as completely). The analyses undertaken possible, analytical values can of course be expanded once data representative of the foods consumed for important but often missing by the study population. It might be components such as carotenoids, expected that this approach would vitamers E and other bioactive minimize bias. The food composition compounds become available. This kind analyses, ideally, should reflect any potential regional differences in foods each country, conversions to the same which are key sources of nutrients. The basis of data expression will be needed extent of regional variation in foods and for nutrients and foods. Specific nutrients is not well established and the software is needed to meet all of the potential biases which may occur by the requirements identified and initial use of non-representative data are a requirements for such software have potential source of misclassification of been described (18). These subjects, as well as obscuring between- requirements for a comprehensive, cohort effects. multi-country user database differ Data Imputation considerably from a national reference database and will have to be specifically Data imputation, which may be incorporated in purpose-built software. necessary for tailoring the database is, by definition, a biased procedure and is Documentation of Nutrient Database likely to affect classification of subjects. The nature of epidemiological work Bias will be minimized by basing makes difficult the question of imputations on analytical data wherever replication of a study as a method of possible from the country concerned. validation. An indispensable part There is some misunderstanding about therefore of the reporting of any imputed data, and whether they are epidemiological research is a better than analytical data or not. The requirement to document the nutrient point is that an imputation, by definition, database used in sufficient detail to is done against a previous set of enable detailed scrutiny when under analytical data, and the assertion of review or when comparisons are made validity cannot be tested without new with other studies. This point has analyses, creating a circular situation. recently been re-emphasized (17). For epidemiological studies a trade-off Ŷ Conclusion between missing data and imputed data has to be made. Again a series of A major barrier to achieving the “ideal simulation tasks could possibly identify approach” is the difficulties the majority the priorities for chemical analysis and of users experience in expressing their indicate where compromises could be needs in a way which permits a custom made. database to be compiled. There are two powerful tools Quality Control potentially applicable to the problems On the practical side, the problems for a posed by large-scale multi-country food composition database are relatively studies. First, so long as great attention simple to understand. Prospective is paid to the collection of the dietary studies have to address (among other intake and other data, these data could problems) the implications of personnel be re-analyzed against date-stamped and methods of measurement being sets of food composition data in the subject to change during the study. As future enabling a clarification of the the database has to be maintained over nutrient exposure. Second, further an extended period, quality control will studies involving computer modeling demand extensive computerized data which examine the potential impact of documentation, including dating of food defined systematic errors in food analyses (12). composition data on dietary intake data, Mode of Data Expression particularly by using sample populations with a wide variety of food habits, and To compare countries, even where with intakes corrected to energy intake, indigenous data sets are available for would undoubtedly be useful in Handbook No. 8 series, USDA, validating differences in dietary Washington, DC exposure. (8) Buzzard, I. M., Price, K.S., & Finally, future prospective studies Feskanich, D. (1991) in The Diet could consider the option of collecting History Method, L. Kohlmeier (Ed.), “food archives” in which sample diets Smith-Gordon and Company Ltd, from study populations are collected and London, pp. 39–51 stored at low temperatures for future (9) Magarey, A., & Boulton, T.J.C. (and replicated) analysis in much the (1991) Aust. J. Nutr. Diet. 48, 128- same way in which plasma or urinary 31 samples for biochemical markers are (10) Greenfield, H. (Ed.) (1987) Food currently collected and stored (2). Technol. Aust. 39, 181–140 (11) Cashel, K.M., & Greenfield, H. Ŷ References (1995) J. Food Comp. Anal. (in (1) Hennekens, C.H., & Buring, J.E. press) (1987) Epidemiology In Medicine, (12) Buzzard, I.M. (1991) in Proceedings Little, Brown and Company, Boston, of the 16th U.S. National Nutrient MA Databank Conference, The CBORD (2) Riboli, E. (1992) Ann. Oncol. 3: Group Inc, Ithaca, pp. 73– 77 783-91 (13) Jia, Y. (1992) MSc thesis, (3) Willett, W. (1990) Nutritional University of New South Wales, Epidemiology, Oxford University Sydney Press, Oxford (14) Commonwealth Department of (4) West, C.E., & van Staveren, W.A. Health and Community Services (1991) in Design Concepts in (1989) NUTTAB89, diskette, AGPS, Nutritional Epidemiology, B.M. Canberra Margetts & M. Nelson, (Eds.), (15) Commonwealth Department of Oxford University Press, Oxford, pp. Health and Community Services 101– 119 (1990) NUTTAB90, diskette, AGPS, (5) Paul, A.A., & Southgate, D.A.T. Canberra (1988) in Manual on Methodology (16) Commonwealth Department of for Food Consumption Studies, Health, Housing and Community M.E. Cameron & W.A. van Services (1991) NUTTAB91/92, Staveren, (Eds.), Oxford University diskette, AGPS, Canberra Press, Oxford, pp. 121–144 (17) Stockley, L. (1988) J. Hum. Nutr. (6) Beaton, G.H. (1987) in Food Diet. 1, 187–195 Composition Data: a User's (18) Greenfield, H., Hémon, B., Slimani, Perspective, W.M. Rand, C.T. N., & Riboli, E. (1991) Windham, B.W. Wyse, & V.R. NUBEL/EURONIMS Meeting, Young, (Eds.), UNU Press, Tokyo, Antwerp pp. 194– 205 (19) Murphy, S.P. (1993) Aust. J. Nutr. (7) US Department of Agriculture Diet. 50, 176 (1976- ) Composition of Foods: Raw, Processed, Prepared, Agric. Food Composition Data and Population Studies

Developing a Food Composition Database for Studies in the Pacific Islands

Jean H. Hankin, Loïc Le Marchand, Laurence N. Kolonel

Epidemiology Program, Cancer Research Center of Hawaii, Honolulu, HI 96813, USA

Brian E. Henderson

Salk Institute, La Jolla, CA 92037, USA

Gary Beecher

Nutrient Composition Laboratory, U.S. Department of Agriculture, Beltsville, MD 20705, USA

As part of collaborative surveys of lifestyle risk factors for cancer and other chronic diseases in several Pacific Islands, diet studies were conducted among samples of semi-urban 45–65 year old men and women living in each island. Local nutritionists, dietitians, and other health workers identified the food items usually consumed, along with the seasonal fruits and vegetables that were major sources of carotenoids. The food composition table used to calculate nutrient intakes was developed during and following the survey, using a variety of procedures, including recipe calculations, laboratory analyses for carotenoids, and sourcing data from national and international food composition tables. The original carotenoid data for the Pacific Islands fruits and vegetables are presented in this paper.

eveloping a food composition database for emerging and somewhat isolated nations, such as the South Pacific Islands, presents an interesting challenge for Dnutrition researchers. There are problems in identifying the various traditional and imported foods, determining the usual food preparation methods, and assigning appropriate nutrient values to rare items not found in published food composition tables. We had an opportunity to meet this challenge in our recent study of diet and other lifestyle risk factors for cancer in several South Pacific Islands. For the dietary component of the study, the objectives were to obtain representative data on the usual diets of the islanders and to characterize the dietary intakes according to particular nutrients and other dietary components, as well as selected food items and food groups. This paper will review the background of the study and the procedures followed for identifying the foods usually eaten, developing the diet history questionnaire, determining the composition of the local foods, and creating the database for each island. In addition, we will discuss some of the problems that may occur in developing a database for an emerging country and offer a few suggestions that may be helpful.

The variation of incidence rates Ŷ Background of the Study within and among the ethnic groups Since 1980, the South Pacific suggested that environmental, and in Commission (SPC), with the assistance particular lifestyle factors, may be of the University of Southern California associated with these variable cancer Comprehensive Cancer Center patterns. To identify particular risk (USCCCC) and the Cancer Research factors, the SPC, CRCH, USCCCC and Center of Hawaii (CRCH), has been the Ministry of Health of each island recording all reported cases of cancer in conducted cross-sectional surveys in the South Pacific region. Analysis of the the Cook Islands, Fiji, French Polynesia, incidence data revealed marked and New Caledonia between 1988 and variation in the rates of several site- 1992. The objectives were to collect specific cancers among the different data on the prevalence of lifestyle ethnic and island populations (1,2). For factors (such as smoking, drinking, diet, example, the Polynesians in Hawaii, reproductive history, physical activity French Polynesia, Cook Islands, and and obesity) among representative New Zealand, tend to have high rates of samples of semirural adults and to several cancers which most likely are correlate these data with the observed related to diet. For instance, stomach cancer incidence patterns. These island cancer rates are generally higher among communities are undergoing rapid Polynesians as compared to the other economic, technological and social islanders. Lung cancer rates are high change, which is having an impact on among all Pacific Islanders except their eating patterns, especially in urban among the Melanesians and Indians in areas. For instance, the use of imported Fiji. Breast cancer and prostate cancer foods has resulted in a modification of are also relatively high among the their traditional food practices. We Polynesians. However, the rates of hoped that the study findings would lead colon cancer among Polynesians living to greater knowledge about the causes in Hawaii and New Zealand are low in of cancer in the South Pacific Region comparison to the Caucasians living in and would be utilized by the Ministries of these respective countries. Among the Health for planning public health Melanesians, New Caledonians have interventions to control cancer and other considerably higher rates of lung cancer chronic diseases. than the Fijians, as well as the Indians in Fiji. This is of particular interest because Ŷ Methods the Fijians have high lifetime rates of The same methodology was followed in smoking. each country to obtain comparable results. The surveys were conducted in the same season (June through August) conferred with other nutrition, health and of the year. Random samples of agricultural personnel, and prepared a approximately 250 semi-rural males and list of food items for the diet history females, 50 to 65 years of age, from questionnaire. The items included both each main ethnic group living on the western and traditional foods that island were included in each of the covered several food groups, such as surveys. The questionnaires included a starches, breads and spreads, meat, diet history, information on cigarette poultry and fish, vegetables, fruits, smoking, alcoholic consumption, snacks, beverages, etc. Information on physical activity, and medical and the usual methods of food preparation, reproductive histories. Additional including the use of particular fats and components included anthropometric oils and coconut cream in food data (weight, height, triceps and skinfold preparation, was also identified. measurements); plasma and serum Developing the Diet History samples which were subsequently Questionnaire analyzed for carotenoids and tocopherols; and urine samples which The diet history listed each item were analyzed for sodium and cotinine individually. The local names of each (an indicator of smoking history). food were included in the questionnaire, which was administered by trained Identifying the Food Items for the nutrition and health education Dietary Assessment personnel. The format included columns To assess the role of diet in the etiology for recording weekly or monthly of diseases, such as cancer and heart frequencies and usual serving sizes. To disease, investigators seek information assist participants in nominating the on the usual dietary intake of quantities consumed, the island's individuals. Generally, a diet history nutritionists or dietitians developed method which provides an estimate of appropriate visual aids, such as root the frequencies and amounts consumed vegetables preserved with a shellac during a specified period of time is coating, Polaroid photographs of recommended (3–5). To estimate the medium and large servings of usual diet of the islanders, we utilized a vegetables, plastic meat models, and diet history that included those food different sizes of familiar bowls and items that were likely to be consumed cups. during a one-month period. This time As another measure of dietary intake, interval seemed appropriate because of we collected a 24-hour recall of foods the similarity in the dietary patterns of consumed the day before the interview. the villagers from month to month. In This was done before the diet history, so addition, seasonal fruits that were major that participants would become familiar sources of carotenoids and ascorbic with recalling what they ate and how to acid were included in the questionnaire. use the visual aids for estimating The selection of the particular food amounts consumed. items for the diet history began several Determining the Composition of months before the survey. The Foods Consumed in Each Island nutritionists, dietitians and epidemiologists from the Ministry of On arrival in each island, we visited the Health in each country identified the various produce markets and village foods usually consumed at least once a stores to observe firsthand the available month, along with the seasonal items. food supply. Unfamiliar foods were They reviewed recent dietary surveys, purchased and identified, and the contents of commercially prepared items medium fat and were used by the were separated and weighed to develop interviewers for coding the reported fish estimated quantities of the ingredients. items. Recipes of various mixed or traditional One of our major objectives was to dishes were also obtained. In addition, obtain estimates of the carotenoid procedures were designed for collecting values of the vegetables and fruits and preparing the fruits and vegetables grown on each island. One of us (GB) for carotenoid analysis. The labels of performed the laboratory analyses of cereals, rice, flour, breads, crackers, these items. In each area, 15 to 20 and similar items were scrutinized to highly consumed foods were selected identify the ingredients and to determine from local markets or home gardens. All if the products were enriched or fortified. foods were prepared as normally It was also necessary to investigate the consumed within each population. A available meat, poultry and fish. Beef representative sample of each food was was generally frozen and imported, and packaged, frozen, and shipped on dry it was difficult to identify the particular ice to Beltsville, MD, for subsequent cut and its fat content in the frozen analysis. Carotenoids in extracts from state. The chief nutritionist of each each food were separated and island suggested the probable cut of quantified by a combination of high meat, percentage of fat, and usual performance liquid chromatography and method of preparation. Lamb from New UV-visible spectroscopy (8). The items Zealand was utilized in the Cook Islands included dark leafy greens, other green, and Fiji. Food composition data on New yellow and red vegetables, and a few Zealand lamb were available from the yellow and orange fruits. The green U.S. Department of Agriculture (USDA) vegetables, in particular, included some (6), whereas data on mutton flaps, unfamiliar items, such as hibiscus consumed by the Cook Islanders, were leaves, amaranthus, wild fern, and found in a report by Platt (7). The Fijians drumstick leaves. We located some of consumed both fresh and canned goat. these items in various food composition Values for fresh roasted goat were tables (6,9,10). If nutrient data were not found in USDA (6), but no data were available, we compared the items to available for canned goat. We compared similar vegetables of the same color and the taste and appearance of the two shape and imputed the food products, and based on their similarity, composition values. Although these decided to use the same values for both procedures are not error-free, they are items. Chicken was similar to the acceptable for comparing the diets of stewing chickens of Hawaii, and we various groups of islanders. estimated the cooked items as about 20 Each group of islanders consumed per cent fat (6). some items unique to their own setting. All of the islanders consumed a large For example, in French Polynesia, two variety of reef and ocean fish and “Chinese” plate lunches were popular shellfish. In general, people described and were listed in the diet history. The them by their size (small, medium or first was a mixture of pork, dried white large) or by their traditional names. The beans, macaroni, green beans, rice and nutritionists recommended that fish be soy sauce, and the second contained classified according to their estimated chicken, cabbage, noodles, sausage, fat content. Fresh fish of high fat carrots and soy sauce. The soy sauce content, such as salmon, were rarely was obviously the Oriental component! available. The names of the local fish We purchased the lunches, separated were then classified according to low or and weighed the ingredients, estimated the amount of soy sauce, and tuber would be comparable among the developed approximate “recipes” for the Pacific Islands, we utilized the USDA database. Similarly, canned products values (6) for each area. A few “new” imported from France, such as root vegetables were consumed, for “cassoulet”, were purchased, and the example, “wild yams” in Fiji. We used kind and amount of each ingredient the same values as regular yams, but weighed to develop a “recipe”. This assigned different code numbers so the procedure was followed for estimating items could be identified. We also used the contents of various mixed dishes or the same values for plantains and green sandwiches that were eaten frequently. bananas which are most likely In addition, the diet histories in each comparable in composition. They differ, area included a number of traditional of course, in their size, but not in the main dishes, desserts and snacks. The way they are consumed. We decided to nutritionists, other staff, and family use the values of barracuda (2.6 per members contributed information, which cent fat) and Spanish mackerel (6.5 per was used to develop a formula for the cent fat) (6) for the low fat and medium composition for each of these mixed fat categories of fish, respectively. With dishes. a few exceptions, values for shellfish Creating the Database were generally available. A “new” item was “bêche-de-mer” or sea slugs, and Because there were no comprehensive we were fortunate to locate it in the East food composition data for the Pacific Asia tables (9). To insure that our values Islands, we utilized reliable sources of for canned fish were appropriate, we published data whenever possible. Our purchased samples to determine the most frequent resources were the USDA percentage of oil and solids and Nutrient Database for Standard modified the USDA nutrient data, if Reference (6), McCance and warranted. Widdowson's Composition of Foods A few rather exceptional food items (10), Food Composition Table for Use in were consumed by some of the East Asia (9), and an article by Mangels islanders, such as “roussettes” or flying et al. (11). In addition to energy and foxes in New Caledonia. The proximate macro- and micronutrients, the data set values were obtained from Cecily includes values for dietary fiber, starch, Dignan of SPC (personal nonstarch polysaccharides, carotenoids communication), However, we could not and tocopherols. All values represent find values for raw and grilled worms foods as commonly consumed including (“vers de bancoule”), which were recipes which were calculated from the occasionally consumed by Melanesians data for cooked ingredients. No further in New Caledonia. This was one of the adjustments were made for potential very few items not included in the losses after food preparation. Energy dietary analysis. and carbohydrate values were not The carotenoid values of the adjusted when data from different analyzed vegetables and fruits were sources were combined. added to the data set. If analytical data Some of the items included in the were not available for a particular data set may be of interest. The root vegetable or fruit, we averaged the vegetables presented little problem, laboratory values for the same item from because the various sweet potatoes, the other islands and utilized the taro, and breadfruit are also popular imputed data. Published values from among the Polynesians in Hawaii. Mangels et al. (11) were selected for Assuming that the values of the same fruits and vegetables that were not analyzed. It is of interest to note the sampling, geographic location, light and variability of the carotenoid contents of soil, and other factors. For the other the same items from the different dietary components of fruits and islands. Table I shows the variation of vegetables, we assigned the same carotenoids in Chinese cabbage, taro values used for the comparable foods in leaves, leaf lettuce, and hibiscus leaves the Hawaii database (unpublished data). in three of the islands, whereas Table II Analysis of the association of dietary presents the difference of carotenoids in risk factors and cancer incidence in the pumpkin, tomatoes, and papayas in the South Pacific Islands is in progress and four geographic areas. The variations will be reported within the near future. among the islands are probably due to

Table I. Carotenoid content of selected green vegetables from Pacific Islands (mg/100 g edible portion)a, b Food Į-carotene ȕ-carotene Lutein Chinese cabbagec Cook Islands 0 2900 1270 Fiji 0 4570 7470 French Polynesia 0 1111 1470 Taro leavesd Cook Islands 62 4580 3630 Fiji 0 4210 8660 French Polynesia 0 7400 9640 Leaf lettuce, raw Cook Islands 0 1810 903 Fiji 0 2150 2040 French Polynesia 0 1230 1560 Hibiscus leavese Cook Islands 187 5660 4300 Fiji 280 5700 8890 a Nutrient Composition Laboratory, USDA, Beltsville, MD b Green vegetables from New Caledonia were not analyzed (see text for method of imputing the values) c Chinese cabbage (bok choy) was steamed 3–5 minutes and drained d Taro leaves were boiled 40 minutes and drained e Hibiscus leaves were steamed 10 minutes and drained

than the French. Second, although Ŷ Problems and Suggestions foods may have the same name in Based on our experience in developing different countries, they may differ in a food composition database for the food composition. For instance, Chinese Pacific Islands, we are aware of the cabbage (“bok choy”) was dark green in potential problems that may occur in one island, medium green with white analyzing dietary data from isolated stems in a second, and light green with populations. First, it is important to know yellow flowers in a third. These the local names used for various foods. differences probably explain the For example, in Fiji, each item had a variation in their carotenoid values. Melanesian name and a Hindi name, Similarly, in some areas, we observed a whereas in French Polynesia, most difference in the color of a vegetable adults used the Tahitian name rather that was locally grown as compared to the same vegetable that was imported. This information, along with laboratory These items were treated as separate analysis of various antioxidants and foods according to the local or ethnic appropriate botanical data, may permit names. It may be helpful to photograph reasonable imputations of values for the unfamiliar vegetables and to match their food composition database. colors with a set of colored markers.

Table II. Carotenoid content of selected yellow and red vegetables and fruits from Pacific Islands (µg/100 g edible portion)a ȕ- Food Į-carotene ȕ-carotene Lycopene Lutein cryptoxanthin Pumpkin, peeledb Cook Islands 236 951 2240 131 0 Fiji 17603040 1900 2210 0 New Caledonia 1290 4000 0 560 0 Tomato, raw, whole Cook Islands 0 515 1620 0 0 Fiji 0160 2550 130 0 French Polynesia 0 620 4730 210 0 New Caledonia 0 570 7540 110 0 Papaya, yellow, raw, flesh only Fiji 0 100 0 0 560 French Polynesia 0 260 0 140 2470 New Caledonia 0 60 0 70 760 Papaya, red, raw, flesh only Cook Islands 0 137 1940 0 6180 French Polynesia 0260 3040 90 960 New Caledonia 0100 3960 50 620 a Nutrient Composition Laboratory, USDA, Beltsville, MD b Pumpkin was peeled, boiled 30 minutes, and drained

Third, processed foods are likely to was sweetened with considerable sugar be imported from various countries. For and diluted with water. Observing the example, canned, frozen and packaged preparation of “juice” is recommended, products from New Zealand, Australia so that the appropriate nutrient values and France were available in different can be assigned. islands. The labels may suggest that the Fifth, recipes are needed for items are similar to those found in the traditional and ethnic mixed dishes, investigator's native country. However, desserts and snacks. Although island this cannot be assumed. Items, such as recipes may be printed in tourist baked beans, canned or frozen mixed publications, it is preferable to ask vegetables, sausages, etc., need to be several local people for their recipes and checked to identify their approximate use this information to develop contents before selecting published prototype recipes for the food values. Fourth, most islanders used the composition database. Finally, term “juice” (or the local name) loosely. knowledgeable nutritionists and For instance, concentrated syrups were dietitians familiar with the eating often diluted with water and called patterns of the population are the keys “juice”; if real juice was used, it generally to achieving a realistic database that is area-specific and meaningful for (5) Block, G., & Hartman, A.M. (1989) analyzing the dietary intakes of the in Nutrition and Cancer Prevention. population. Investigating the Role of Micronutrients, T.E. Moon & M.S. Micozzi (Eds.), M. Dekker, New Ŷ Acknowledgements York, NY, pp. 159–180 (6) US Department of Agriculture We are grateful to the following (1993) Nutrient Database for nutritionists for their generous Standard Reference, Release 10, assistance in our surveys: Taiora USDA, Washington, DC Matenga Smith, Ministry of Health, (7) Platt, B.S. (1980) Tables of Rarotonga, Cook Islands; Mona J. Representative Values of Foods Chand, Ministry of Health, Suva, Fiji; Commonly Used in Tropical Maeva Barral, Ministry of Health, Countries, Special Report Series Papeete, Tahiti; and Dominique Daly, No. 302, Medical Research Council, Noumea, New Caledonia. We also London thank Cecily Dignan, Nutritionist, South (8) Khachik, F., Beecher, G.R., Goli, Pacific Commission, for her generous M.B., & Lusby, W.R. (1991) Pure support. Appl. Chem. 63, 71–80 Ŷ References (9) Leung, W.T.W., Butrum, R.R., & (1) Henderson, B.E., Kolonel, L.N., & Chang F.H. (1972) Food Foster F. (1982) Nat. Cancer Inst. Composition Table for Use in East Monog. 62, 73–78 Asia. Nat. Inst. of Arthritis, (2) Henderson, B.E., Kolonel, L.N., & Metabolism and Digestive Dworsky, R. (1986) Nat. Cancer Diseases, Bethesda, MD Inst. Monog. 69, 73–81 (10) Holland, B., Welch, A.A., Unwin, (3) Willett, W. (1990) Nutritional I.D., Buss. D.H., Paul, A.A., & Epidemiology, Oxford University Southgate, D.A.T. (1991) McCance Press, Oxford and Widdowson's The Composition (4) Hankin, J.H. (1991) in Research: of Foods, 5th Ed., Royal Society of Successful Approaches, E.R. Chemistry, Cambridge Monsen (Ed.), American Dietetic (11) Mangels, A.R., Holden, J.M., Association, Chicago, IL, pp. 173– Beecher, G.R., Forman, M.R., & 194 Lanza, E. (1993) J. Am. Diet. Assoc. 93, 284–296 Food Composition Data and Population Studies

The Effects of Australian, US and UK Food Composition Tables on Estimates of Food and Nutrient Availability in Australia

Karen M. Cashel

School of Human and Biomedical Sciences, University of Canberra, P.O. Box 1, Belconnen, ACT 2616, Australia

Heather Greenfield

Department of Food Science and Technology, The University of New South Wales, Sydney, NSW 2052, Australia

Until the late 1980s, Australia used national food composition tables that were compiled in the late 1960s, predominantly from overseas sources, or foreign tables, particularly those of the UK or the USA. New tables, The Composition of Foods, Australia (COFA), based on an ongoing national analytical program, have been progressively released from 1989. The quantity and adequacy of the foods and nutrients available for consumption in Australia, 1990–91, calculated on the basis of the new Australian tables are compared here with those obtained using the US or UK tables. There are marked differences in the edible weights of foods and the amounts of nutrients available for consumption when the different databases are used. The most marked effect is on the quantity, type and sources of fat in the food supply, assessed as at least 60 per cent higher from meats, and 15–22 per cent higher in total using the data from the US or UK. Iron and zinc are all higher and retinol activity, vitamin C and magnesium lower using the foreign data. Calcium is 35 per cent higher when UK data are used and thiamin 59 per cent higher when US data are used.

ood composition databases are essential components of nutritional monitoring and surveillance, and of much health-related research, yet many countries have Ftraditionally relied on the United States of America (US) or United Kingdom (UK) tables rather than develop their own national tables. Many individual users also rely on non-local data as their source of information. Inappropriate food composition data have the potential to undermine or misdirect the research or nutrition effort, but few studies have been done to provide quantitative evidence of this.

Until the late 1980s, Australia, like range of nutrients, including data on many other countries, relied on a fatty acids, sugars, amino acids and national food composition database organic acids. This database continues which incorporated data from a variety to grow on an annual basis. of sources, including from overseas The new analytical data on Australian tables, scientific publications and food foods provide a unique opportunity to industry information (1). By the late compare the gross and nutrient 1970s, the inadequacies of the composition data of local foods with information provided in the range of food data from overseas sources for items and nutrients, had many major apparently similar foods, and to assess users, particularly researchers, turning the effect of using local data on the to other sources of data. In Australia, in determination of foods and nutrients the main, users were either developing consumed rather than overseas data. In their own databases by supplementing this paper, the foods available for the Australian tables with data from consumption per capita (13) are used to overseas tables, the food industry and demonstrate and compare the US, UK journal publications, or using overseas and Australian tables. computer-based packages as their Data on the per capita food supply principal source. This approach was have provided the only consistent exacerbated by the growing availability measure of trends in foods and nutrients of overseas databases, including in consumed in Australia. The food supply software packages, well in advance of data are used to monitor the nutritional their Australian print only counterpart. adequacy of the food supply, and, in the The most widely introduced and used absence of more specific consumption overseas food composition data in data at household or individual level, Australia were those from UK (2), or the have provided the basis for developing a US (3), available in print or on computer range of public health nutrition policy tape and/or incorporated into software and programs, including the nutrition packages. The use of US or UK data in component of the National Health Goals Australia was usually justified by and Targets (14, 15). arguments that the health problems and The food supply data represent foods food patterns were similar, and the as available, rather than as prepared Australian tables were too limited in their and consumed (i.e. in cooked and/or coverage of foods and nutrients. mixed form). This level of definition of In 1989, revised food composition “food consumed” allows ready data for Australia (4), began to be identification of the scope and source of released. These data were based any differences found specific using entirely on an ongoing national food alternate sources food composition analysis program initiated at the data. Some of these differences may be beginning of the 1980s (5). The previous difficult to identify, or may be overlooked national tables (TCAF) (6, 7) included in foods as consumed due to the effects fewer than 650 food items, and just 16 of different methods of food preparation nutrients, while in 1993, the new tables and combination. (COFA) (4, 8–12) include some 1400 In this paper, the effects of using US food items and a greatly expanded or UK rather than Australian national food composition databases are of the population, the calculated assessed. Specifically, factors nutrients available per capita are influencing the quantity and adequacy of compared with the weighted population the foods and nutrients available for recommended dietary intakes (WPRDI). consumption in Australia will be Prior to this comparison, thiamin and determined and compared. vitamin C are adjusted to make some allowance for losses during food Ŷ Methods processing and cooking and niacin Food Composition Data equivalents are calculated (13). The WPRDI is derived by calculating the The data used are the food composition sum of nutrients needed to provide the tables, or series of tables developed for RDIs (19) for the proportion of the national use in Australia, the US and the population in each age and sex group, UK. These are, in Australia, and the WPRDI is then expressed per Composition of Foods, Australia (4,8– capita. 12); for the US, the USDA series Composition of Foods (3); and from the Analyses undertaken UK the 1978 HMSO edition of McCance Quantity of Food. The effect of and Widdowson's The Composition of differences in edible portion factors Foods and the subsequent supplements (EPF) was assessed using fruit, released in the 1980s (2,16,17). vegetable and meat data from the three The official printed data sources sources. These factors reflect the rather than commercial packages were proportion of the food that is edible and used. As many of the commercial usually eaten by the population. For computer-based packages have example, for a fruit such as the raw modified or extended databases, this orange, the COFA EPF of 0.74 indicates approach was to ensure that only the that 74 per cent of the food item is official data were used. Further, the considered edible flesh, the other 26 per printed versions include detailed cent (in this instance, skin, seeds, pith) information and explanatory notes and is usually discarded. appendices to assist the user to Nutrients Available — Effect of interpret and apply the data. Differences in EPF. Meats and Ŷ Food Consumption Data vegetables were used as the basis for this assessment. The EPF of each of The quantity and type of foods available the three data sources were used to for consumption per capita (AC) in calculate the nutrients available from the 1990– 91 (13) in Australia are used determined edible weights of meats and (AC). The edible weight of foods and vegetables using COFA nutrient associated nutrients available are composition data. calculated using the most appropriate Nutrients Available — Effect of data selected from the three data Differences in Nutrient Composition. For sources. For example, the edible portion this example, the COFA EPF were used factors (EPF) for carcase meats should as the basis for determining the edible allow for losses at the level of both the weight of meats and vegetables. butcher (carcase to retail meats) and the Nutrients available in the food supply consumer (retail to raw edible meat) from these foods were then calculated (18). using each of the three nutrient To assess the adequacy of the food composition data sources. supply to meet the nutrient requirements Table I. Effect of different edible portion factors from different food composition tables on the weight of fruits, vegetables and meats available for consumption (kg per capita per year) Quantity COFA UK US FEWa EPFb EWc EPF EW EPF EW Foods kg kg kg kg Fruits Oranges 30.0 0.74 22.2 0.75 22.5 0.73 21.9 Apples 16.7 0.92 15.4 0.77 12.9 0.85 14.2 Bananas 12.7 0.64 8.1 0.59 7.5 0.65 8.3 Grapesd 9.7 0.98 9.5 0.88 8.5 0.96 9.3 Pineapples 9.1 0.67 6.1 0.53 4.8 0.52 4.7 Melons 7.5 0.60 4.5 0.56 4.2 0.50 3.8 Other citrus 6.3 0.70 4.4 0.51 3.2 0.52 3.3 Pears 6.0 0.90 5.4 0.72 4.3 0.92 5.2 Peaches 3.1 0.90 2.8 0.87 2.7 0.76 2.4 Othere 6.9 0.86g 5.9 0.87g 6.0 0.86g 5.9 Totalf 108.0 84.3 76.6 79.2 Percent COFA NA NA 90.9% 94.0% EWc Weighted EPF 0.78 0.71 0.73 Vegetables Potato 63.5 0.82 52.1 0.86 54.6 0.75 47.6 Tomato 25.8 0.99 25.5 1.00 25.8 0.91 23.5 Onions 10.3 0.88 9.1 0.97 10.0 0.90 9.3 Carrots 8.2 0.90 7.4 0.96 7.9 0.89 7.3 Peas 6.7 0.36 2.4 0.37 2.5 0.38 2.5 Lettuce 5.9 0.87 5.1 0.70 4.1 0.95 5.6 Pumpkin 5.7 0.80 4.6 0.81 4.6 0.70 4.0 Cabbage & other 5.4 0.77 4.2 0.78 4.2 0.73 3.9 green leafy Cauliflower 4.8 0.57 2.7 0.62 3.0 0.39 1.9 Sweet corn 3.9 0.52 2.0 0.66 2.6 0.36 1.4 Celery 3.4 0.79 2.7 0.73 2.5 0.89 3.0 Otherh 12.6 0.86g 10.8 0.73g 9.2 0.83g 10.4 Totalf 156.2 128.6 131.0 120.8 Percent COFA EWc NA 101.9% 93.9% Weighted EPF 0.82 0.84 0.77 Meats Beef 39.2 0.66 25.9 0.83 32.6 0.80 31.4 Vealj 1.5 0.59 0.9 (0.83) (1.2) (0.69)k (1.0) Lamb 21.8 0.63 13.7 0.84 18.3 0.84 (18.3) Pigmcat 18.0 0.65 11.7 0.74 13.3 0.82 14.8 Offal1 3.8 0.98 3.7 0.96 3.6 0.98 3.7 Poultry 25.4 0.62 15.7 0.64 16.3 0.69 17.5 Totalf 109.7 71.7 85.3 85.2 Percent COFA NA 119.1% 118.9% EWc Weighted EPF 0.65 0.78 0.78 a = fresh equivalent weight b = edible portion factor c = edible weight d = includes FEW of grapes to be dried e = apricots, figs, plums, berries, figs, cherries, custard apples, mangoes, pawpaws, strawberries, olives f = rounded from more detailed individual data items g = total edible weight/total FEW h = beetroot, beans, cucumber, eggplant, marrows, mushrooms, sweet potato etc j = data in brackets derived using EP factors for beef k = data in brackets dervied using EP factors for composite boneless meat

Nutrients Available — Effect of Despite differences in the EPF of up Differences in EPF and Nutrient to plus 27 per cent or minus 32 per cent Composition. The effects of differences of the edible weight (EW) of the in edible portion and nutrient individual fruits and vegetables, the composition data in each of the three impact on the total edible weight of data sources were assessed for all food these commodities available for groups, including the meat and consumption is much smaller (minus 10 vegetable groups. The relative per cent to plus 2 per cent). The use of contributions of the macronutrients to the UK and US data gives a total EW of total energy were calculated. vegetables 101.9 per cent and 93.9 per Nutrient Adequacy. The total cent of that obtained using COFA. The nutrients available in the food supply effect on total EW of fruits is greater, were then assessed for adequacy with results 90.9 per cent and 94.0 per against the WPRDI. The proportion of cent using UK and US data compared to energy contributed by the COFA. macronutrients was also determined. The other major food group on which The range of nutrients included was EPFs have a marked effect is the selected on the basis of consistency meats. Table I shows that for beef, lamb across all data sources, and on the and pigmeat there is a consistently basis of those for which there are RDIs higher EPF for carcase meats in the UK for use in Australia (19). and US databases than in COFA. The effect of using these EPFs to calculate Ŷ Results the raw EW of meat available for In this paper, the COFA data are used consumption is an EW of meats and as the basis for all comparisons made. poultry of 119.1 per cent and 118.9 per Quantity of Food Consumed cent when UK or US factors are used, respectively, compared to COFA. The Each of the three food composition basis for the revised EPF for Australian tables provides EPF for foods such as carcass meats is reported elsewhere fruits, vegetables and meats. Table I (18). shows the EPF for a range of raw fruits, vegetables and carcase meats from Nutrients Available for Consumption each of the three data sources. On a Effect of Differences in EPF. Using weight basis, the fruit and vegetable COFA nutrient data as a constant in all items comprise 93 per cent of all fruits, calculations, Table II indicates the effect and 92 per cent of all vegetables of the different EPF from each of the available for consumption in Australia. three sources on the nutrients The remaining items from these food calculated as available for consumption groups are included in the “other” from vegetables and meat. The effects category. For both fruits and vegetables are generally consistent with the there are marked differences in the EPF differences in total EW shown in Table I. for individual foods reported in each The exception to this is retinol activity source. contributed by the meats. The EPF for meats (beef, veal, lamb, pigmeat) are data for boneless meats when available low in COFA compared to all other in the US databases reduces this to 115 sources, but the EPF for offal are per cent of the COFA fat contribution. similar. While small amounts of retinol The energy contributed by vegetables are contributed by muscle meats, as using UK and US nutrient data are shown in brackets in Table II, offal is the similarly higher compared to COFA due major determinant of the quantity of mainly to the considerably higher retinol contributed by meat. This is carbohydrate levels. Calcium, riboflavin responsible for the similar retinol and thiamin contributed by vegetables contributed by all three sources. are higher when US and UK data are Effect of Differences in Nutrient used. Magnesium and retinol activity Composition per 100 g Edible Portion. levels from vegetables are also higher Table III shows the impact of the than COFA when the US or UK data are differing nutrient data from the three used, as is niacin from meats. These data sources on the contribution of results reflect the generally higher levels vegetables and meats to the nutrients of these nutrients reported in these data available in the food supply. The sources. Retinol activity, however, is quantities of EW of food are calculated lower from meats when data sources using the COFA EPF. This table shows other than COFA are used. Differences that the fat contribution from meats was in the composition of offal are primarily 147 per cent and 133 per cent of that of responsible for the variation obtained. COFA using the UK and US data, While offal is the main source of retinol respectively. Combined with the activity contributed by the meats group, associated variations in protein levels, the use of US data suggest a this results in the energy contributed considerably greater contribution from from meats also being 124 per cent and other meats, particularly poultry, and a 121 per cent, respectively, compared to considerably lower contribution when COFA. Using fat-trimmed composite the UK data are used.

Table II. Effect of differences in source of EPF on nutrient contribution from vegetables and meats, quantity per capita per day. COFA used as nutrient composition source Source Protein Fat Carbohydrate EnergyCa Fe Mg Zn RetinolThiaminRiboflavin Niacin Vitamin activity C g g g kJ mgmgmgmg µg mg mg mg mg Vegetables COFA 6.8 0.5 26.0 583 45 2.0 51 1.1 477 0.25 0.16 3.3 73 UK 7.0 0.5 27.3 611 45 2.0 52 1.2 499 0.25 0.15 3.4 74 US 6.3 0.5 23.8 538 43 1.9 47 1.1 460 0.23 0.15 3.1 67 Meats COFA 37.4 32.6 0.2 1850 15 3.5 36 5.0 1839 0.31 0.64 7.8 2 (21)a UK 44.3 39.0 0.2 2198 17 4.0 43 6.2 1813 0.36 0.69 9.2 2 (24) US 44.3 39.0 0.2 2198 17 4.0 43 6.0 1838 0.37 0.69 9.2 2 (25) a = values in parentheses are for retinol activity from non-offal meats

Effect of Differences in EPF and food groups, except that the different Nutrient Composition Data. Table IV source data EPF factors have also been provides a similar comparison for all applied. For vegetables, for example, Table I showed that while the EPF from composition tables are used, being up to the different data sources varied 156 per cent that of COFA for calcium considerably for any particular from vegetables when UK data are vegetable, the differences were small for used. Zinc contributed by vegetables the total weight of vegetables. The using US data, is nearly twice that combination of differences in EPFs and obtained using COFA. The use of UK or nutrient data at the individual vegetable US data also suggest a considerably level, however, result in very different higher total retinol activity from estimates of the nutrients available for vegetables compared to COFA: 129 per consumption from vegetables. The cent and 151 per cent, respectively. This carbohydrate contribution from is also seen with thiamin and riboflavin vegetables compared to COFA is higher (120 per cent and 125 per cent, using the UK (154 per cent) and US respectively, that of COFA when UK (138 per cent) data. The minerals all data are used). In contrast, vitamin C is show variation with data source. around 90 per cent that of COFA when Calcium and magnesium levels are UK or US data are used. higher when the UK and US food

Table III. Effect of differences of nutrient composition on nutrients available from vegetables and meats, quantity per capita per day. COFA used as an EPF source Source Protein Fat CarbohydrateEnergyCa Fe Mg Zn RetinolThiaminRiboflavin Niacin Vitamin activity C g g g kJ mgmgmgmg µg mg mg mg mg Vegetables COFA 6.8 0.5 26.0 583 45 2.0 51 1.1 477 0.25 0.16 3.3 73 UK 5.9 0.4 38.1 729 69 2.0 63 1.0 591 0.29 0.20 3.1 66 US 5.9 0.7 38.8 726 52 2.4 62 1.0 736 0.28 0.18 3.5 70 Meats COFA 37.4 32.6 0.2 1850 15 3.5 36 5.0 1839 0.31 0.64 7.8 2 (23)a UK 31.0 47.9 0.1 2303 16 3.3 354.9 1430 0.320.63 9.4 1 (6) US 33.4 43.4 0.3 2237 18 3.4 35 5.5 683 0.36 0.68 10.5 2 (100) a = values in parentheses are for retinol activity from non-offal meats

For meats, the differences in both (UK) and 38 per cent (US) of the level EPF and nutrient composition in the obtained using COFA. three data sources further exacerbate When the nutrients available for the trends observed in Tables II and III. consumption from all foods, with Fat, energy, and the minerals calcium, thiamin, niacin and vitamin C adjusted iron and zinc are all higher when data as described in the methods section, other than COFA are used. The fat levels are higher when data other than contribution from meats is 178 per cent COFA are used. The exceptions are and 161 per cent of that of COFA when magnesium, retinol activity and vitamin UK or US data are used, while energy C for both UK and US data, and thiamin, levels are 150 per cent and 145 per riboflavin and niacin equivalents when cent, respectively. The retinol activity the UK data are used. The higher levels are all lower using data sources contribution of meat fat to total available other than COFA; being 77 per cent fat suggested by the use of UK and US data has the effect of reducing the obtained using UK and US data is also relative importance of the added fats higher. and oils as a source of fat in the national Alcohol content varies with the data diet. Using COFA data, added fats source, being lower when UK data are contribute 60 per cent more fat in the used (93 per cent) and slightly higher national diet than the meats; using the when US data are used (102 per cent). other data sources suggests that the The retinol activity data vary from 67 per contributions of meats and added fats cent (US) to 93 per cent (UK) of those of are about equivalent. Consequently, the COFA. Vitamin C is also 16 per cent ratio of animal fats to vegetable fats lower using data sources other than COFA.

Table IV. Effect of different sources of EPF and nutrient composition on nutrients available for consumption per capita per day Source Protein Fat CarbohydrateAlcoholEnergy Ca Fe Mg Zn RetinolThiamin RiboflavinNiacinVitamin activity C g g g g kJ mg mg mg mg µg mg mg mg mg COFA Meats 37.4 32.6 0.2 0 1850 15 3.5 36 5.0 1839 0.31 0.64 7.8 2 Seafood 5.1 1.3 0 0 138 23 0.3 8 1.0 6 0.01 0.03 1.0 0 Milk & milk 19.4 21.4 19.8 0 1442 659 0.6 52 2.5 215 0.20 0.76 0.4 5 products Fruits 1.9 0.2 26.2 0 480 39 0.8 22 0.4 42 0.12 0.07 0.6 53 Vegetables 6.80.5 26.0 0 583 45 2.0 51 1.1 477 0.25 0.16 3.3 73 Grains 26.8 3.7 183.0 0 3700 49 5.3113 1.6 0 0.85 0.71 9.6 0 Eggs 2.2 1.7 0.1 0 7 0.3 2 0.2 27 0.01 0.07 0 0 Nuts 1.9 4.6 0.6 0 210 15 0.3 36 0.5 0 0.04 0.08 0.7 0 Oils & fats 0.2 52.6 0.2 0 1952 4 0 0 0 294 0 0.01 0.1 0 Sugars 00 122.5 0 1958 4 0.1 0 0.1 0 0 0 0 0 Alcohol1.0 0 6.9 17.5 648 15 0.1 21 0 0 0 0 1.3 7 UK Meats 36.6 58.1 0.1 0 2775 18 3.9 41 6.0 1410 0.37 0.70 10.9 2 Seafood 4.9 1.8 0 0 150 18 0.3 9 1.1 8 0.03 0.04 1.3 0 Milk & milk 19.9 21.4 20.1 0 1452 668 0.3 53 2.2 293 0.14 0.81 0.4 5 products Fruits 1.4 0.2 23.5 0 408 42 0.829 0.3 53 0.12 0.06 0.6 48 Vegetables 6.1 0.5 40.1 0 763 70 2.0 65 1.0 614 0.30 0.20 3.2 66 Grains 23.9 4.2 195.6 0 3687 300 5.5 85 2.3 0 0.75 0.07 5.0 0 Eggs 2.2 1.9 0 0 105 10 0.3 2 0.2 33 0.02 0.08 0 0 Nuts 1.8 4.1 1.5 0 177 12 0.2 17 0.2 0 0.03 0.04 0.6 0 Oils & fats 0.1 52.9 0 0 1961 2 0.1 0 0 281 0 0 0 0 Sugars 0 0 128.1 0 2051 54 0.1 0 0.9 0 0 0 0 0 Alcohol 0.9 08.0 16.2 613 38 0.6 30 0 0 0 0.12 1.3 0 US Meats 39.5 52.5 0.2 0 2691 21 3.9 33 6.5 693 0.42 0.75 12.2 2 Seafood 5.2 0.6 0.1 0 119 13 0.4 8 0.5 5 0.01 0.05 0.7 0 Milk & milk 20.0 19.8 20.7 0 1416 679 0.4 59 2.4 189 0.16 0.76 0.4 4 products Fruits 1.6 0.6 31.5 0 515 39 0.5 22 0.2 55 0.11 0.09 0.68 49 Vegetables 5.5 0.6 35.8 0 672 51 2.2 57 2.0 720 0.26 0.18 3.3 64 Grains 26.0 2.8 194.1 0 3853 4611.3106 2.5 2 1.82 1.10 14.2 0 Eggs 2.1 1.9 0.2 0 112 10 0.4 2 0.2 27 0.02 0.05 0 0 Nuts 1.8 4.1 1.5 0 196 15 0.3 19 0.2 0 0.03 0.04 0.6 0 Oils & fats 0.3 52.6 0.1 0 1953 7 0 1 0 240 0 0.01 0 0 Sugars 0 0 122.5 0 1958 4 0.1 0 0.9 0 0 0 0 0 Alcohol 0.9 0 10.9 17.9 719 19 0.3 22 0.1 0 0.02 0.09 1.4 0 Summary totals (adjusted)a COFA 103 119 385 17.5 13060 87313.334212.5 2901 1.52 2.51 42.6b 102 UKc 98 145 416 16.2 14140118114.133114.2 2693 1.50 2.12 39.5b 85 USd 103 136 418 17.9 14205 90419.832915.5 1931 2.42 3.10 50.6b 86 a = rounded from more detailed individual data items. Thiamin and vitamin C adjusted for losses with processing and cooking. Niacin equivalents calculated b = niacin equivalents c = without fortification of flour, total calcium = 926 mg; iron = 13.1 mg; thiamin = 1.14 mg and niacin equivalent = 37.5 mg; without fortification of skim milk powder, retinol activity = 2672 mg d = without fortification of flour and rice, iron = 12.1 mg; thiamin = 1.20 mg, riboflavin = 2.18 mg and niacin equivalent = 40.5 mg Table V. Macronutrient contribution to total energy a (per cent) Data source Protein Fat Carbohydrate Alcohol COFA 13.6 34.3 48.1 4.0 UK 11.7 37.9 47.1 3.3 US 12.5 36.0 48.0 3.5 a adjusted to ignore minor contributions to total energy from minor sources such as organic acids

The use of the UK data indicates that Table VI shows that the use of the three grain products are a major source of data sources give values for protein, calcium, providing 25 per cent of the retinol activity (even with contribution of total, as compared with only 5 per cent offal discounted), thiamin, riboflavin, obtained using the other food niacin equivalents, and vitamin C that composition sources, while the use of are at least 50 per cent in excess of the US data suggests that grains recommended intakes. For calcium, contribute twice as much iron and COFA and the US data suggest that thiamin and nearly 50 per cent more there is little excess available in the food niacin and riboflavin. Both results are supply relative to the requirements of due to fortification of wheat flour, with the population, while the use of the UK calcium in the UK, and wheat flour and data suggests there is a comfortable rice with iron, thiamin, niacin and excess of 41 per cent of this nutrient. riboflavin in the US. The fortification of With the exception of COFA, the wheat flour with iron, thiamin and niacin available level of zinc is at least 29 per in the UK is not as apparent in these cent in excess of requirements. The use results. of COFA data also gives lower levels of Table V presents the macronutrient iron in excess of the WPRDI than the data using the three data sources, as other data sources. The adjustment of per cent contribution to total energy. the US and UK databases used to Data sources other than COFA result in “remove” the fortifying nutrients from a higher contribution from fat and a wheat flour or rice reduces these consequent lower contribution from differences, with the excess of WPRDI carbohydrate, alcohol and protein. for iron, thiamin and niacin equivalents Nutritional Adequacy of the Food then being lower (compared to COFA) Supply when US or UK data are used. Ŷ Discussion Quantity of Food Consumed may reflect differences in the parts of EPFs are highly variable and have the the food that are considered edible in potential to markedly affect the the local community. For example, in estimates of nutrient intakes. There are some food cultures, spinach stalks are many possible reasons for the variation discarded, in others they are consumed. reported in the different databases. It Further, in Australia, in response to the may be cultivar related, or due to local demand for lower fat meats, there have preference for a particular unit size or been changes in developing animals stage of maturity; or it may reflect the with different characteristics, in degree of pre-market trimming of preslaughter feeding and handling inedible or unattractive components. For practices, in butchering techniques and example, mature carrots used to be in retail fat trimming practices (20). marketed in Australia with their green These affect both EPF and nutrient leaves. These are now removed prior to composition. sale. Alternatively, differences in EPF

Table VI. Effect of differences in source of EPF and nutrient composition on the assessment of the nutritional adequacy of the food supply (per cent in excess of WPRDIa) Source ProteinEnergy Ca Fe Mg Zn Retinol ThiaminRiboflavin Niacin Vitamin activityb equivts. C g kJmgmgmgmgµgmg mg mgmg WPRDI 45.8 9283 838 9.2 261 11 685 0.89 1.36 15.2 34 COFA 125 41 4 45 31 14 324 7085 180 200 (58) UKc 114 52 41 53 27 29 293 69 56 160 150 (11) (42) (87) (28) (147) USc 125 53 8 115 26 41 182 172 128 233 153 (32) (97) (35) (60) (166) a = WPRDI (ABS, 1993); Mg & Zn calculated for this paper b = values in parentheses are for comparisons without contribution from offal meats c = values in parentheses for calcium, thiamin, riboflavin and niacin equivalents are for comparisons without fortification of wheat flour or rice. Nutrients Available for Consumption These and other factors influence the actual composition of a food, and the Users have a number of sources of food relevance of the food composition composition databases available to database used. There are, however, them. This is particularly apparent in other factors that influence the countries such as Australia where a interpretation of food composition data, national food analysis program is only of and the comparability of data from recent origin. Local food availability, different sources. For example, the food regulations, food preferences and analytical methods used to determine preparation practices all influence the nutrient levels and the mode of actual gross and nutrient composition of expression of these nutrient data may foods. The food composition database vary between food composition tables. selected for use, unless specific to the Methods of Analysis. These can have local food supply, can have a marked a large effect on the reported value of a effect on the outcome of a study, both nutrient in a food. This effect can be so on nutrients, and on foods as sources of striking that data from two different nutrients. tables cannot necessarily be combined and be expected to provide a US tables the term used is vitamin A, meaningful assessment of the dietary and the values expressed as retinol intake for that nutrient. equivalents or International Units. The most obvious example is Energy may be expressed as kilojoules carbohydrate. The carbohydrate data in or kilocalories, and the factors used to COFA and the UK tables represent a calculate energy vary as described direct analysis of the sugars and starch above. Total energy may include other content of foods. The carbohydrate data energy-contributing components such in the US do not represent direct as organic acids, as in COFA. measures, but rather are calculated “by Missing Data. In the printed version difference”, a method which includes of the UK tables (2), for example, the dietary fibre in the carbohydrate data. fact that there are no measures of zinc The US tables add to the confusion by for a variety of foods, particularly fish, is reporting other measures of fibre clearly indicated. In computer based components, namely crude fibre and tables, zeros may be inserted with pectin. obvious problems for the user who is The method of determining unaware that “O” may represent either carbohydrate in foods also affects the no nutrient detected at the level of associated energy calculations. When delectability of the analytical method carbohydrate is determined “by used or, no data available on the level of difference”, the carbohydrate energy this nutrient in this food. This can result conversions factors used are food type in an incorrect perception of a food such specific, and allow for the potential fibre as fish as a food source of zinc, an component. This is not the case when inappropriately low value for the total carbohydrate is determined by analysis. dietary intake of this nutrient, and an Other common examples of the effect incorrect interpretation of the results of different methods of analysis are obtained. This is an obvious example of vitamin C, vitamin A and dietary fibre. the value of checking the data on the The analytical methods used in the computer version against the official COFA and UK tables for vitamin C, for published copy. A related problem for example, include ascorbic acid and users of computer databases is that of dehydroascorbic acid. The US data are national differences in nutrient measures of reduced ascorbic acid only. fortification regulations and practices. Data for total ascorbic acid including the The impact of this on study outcomes dehydroascorbic acid form is given in and appropriate interpretation of data is footnotes where available. clearly shown in this paper. Specific Modes of Expression. Nutrients may information is needed to both identify be expressed differently in different and adjust for these effects and even tables. The term dietary fibre may then the effects may be masked by include different components dependent variable voluntary nutrient additions upon the method used. Carbohydrate (e.g. in breakfast cereals). components (starch, sugars and dietary Nutritional Adequacy of the Food fibre) are expressed as monosaccharide Supply equivalents in the UK tables (2), but as the direct measure in the Australian Assessment of the amounts of a nutrient tables (4,8–12). Vitamin A is expressed available per capita against the WPRDI as retinol and ȕ-carotene equivalents in is the basis for monitoring the trends in the UK; retinol equivalents, retinol and the nutritional adequacy of the ȕ-carotene equivalents in COFA and the Australian food supply. The level of measures are direct weights, but in the nutrient in excess of the WPRDI, is used as an indication of the “safety margin” reflecting the iron fortification of wheat for that nutrient. In recent years, the flour and rice. Differences in breakfast National Health and Medical Research cereal fortification with iron also have an Council has expressed concern about effect, but varies with product as well as several nutrients in the Australian food country. supply — thiamin, calcium and iron Calcium. Calcium intakes are of (21,22). considerable concern in Australia due to Thiamin. The data in Table VI data the prevalence of osteoporosis and data indicate thiamin at 70 per cent in excess indicating a decrease in consumption of of the WPRDI using COFA, whereas if milk and milk products in adolescence the standard US data were the basis of (24). In 1992 a recommendation “to the assessment, the 172 per cent increase the intake of calcium excess would be grounds for containing foods” was added to the complacency. Grains are the major revised dietary guidelines for Australians source of thiamin when all three data (21). The use of COFA or US data to sources are used, the absolute level of assess the adequacy of this nutrient in thiamin contributed by grains when US the food supply, indicate only a small data are used is much greater safety margin. The level of calcium compared to the other two sources. This suggested by the standard UK data, reflects the level of thiamin fortification however, would not raise the same of wheat flour in the US. Conversely, degree of concern because of the “removing” this added thiamin from the calcium fortification of flours in the UK. US or UK wheat flour suggests that Country Specific Food Composition there is a lower excess of 35 per cent of data WPRDI compared to COFA at 70 per cent. The reasons for this are the All the data sources are derived from naturally higher level of thiamin in national food composition databases Australian flour due to higher extraction developed to best represent the local rates, and, at that time, a segment of the food supply. COFA is based primarily on flour supply contained voluntarily added a national food analysis program. The thiamin. UK database is underpinned by a Iron. The use of COFA shows that national analytical program, however, iron levels are 45 per cent in excess of the US database is primarily compiled WPRDI (considerably lower than from analytical data produced by obtained using the previous Australian independent, mainly US based food tables at 93 per cent of the WPRDI researchers. (25). This information coupled with a From the results in this paper, the national survey of schoolchildren in use of UK and US food tables gives 1985 indicating that 9 per cent of 15 nutrient estimates that are closer to year old girls had compromised iron those obtained using TCAF, the “old” status based on biochemical Australian tables, than using COFA (25) assessment (23) led directly to a probably reflecting the previous reliance recommendation to “increase the on the data from these two sources in consumption of iron containing foods” in the compilation of the “Australian” food the national dietary guidelines (21). By composition tables. Even with access to contrast, the use of the UK or US food the more recent US and UK data used composition tables would not have in this paper, the Australian food caused such a degree of concern, at 53 analysis program has shown that there per cent and 115 per cent in excess of are real differences in the composition WPRDI, respectively, the US results of locally produced and consumed Ŷ References foods. (1) English, R. (1981) Food Tech. Aust. Ŷ Conclusion 33, 103–106. The analysis in recent years of foods (2) Paul, A.A., & Southgate, D.A.T. currently available and consumed in (1978) The Composition of Foods, Australia has provided the first HMSO, London opportunity to assess the effect of using (3) US Department of Agriculture data from other countries on perceptions (1976- ) Composition of Foods: of foods and nutrients available in Raw, Processed, Prepared, Agric. Australia. Handbook No. 8 series, USDA, The data presented in this paper Washington, DC show that using overseas data sources (4) Cashel, K., English, R., & Lewis, J. to estimate nutrient availability can (1989) Composition of Foods, produce significant errors in the Australia, Australian Government assessment of the nutrient adequacy of Publishing Service, Canberra the food supply, and of the relative (5) English, R. (1986) Trans. Menzies significance of foods as sources of Found. 11, 25–34 nutrients. The implications for the (6) Thomas, S. & Corden, M. (1970) development of nutrition programs, Tables of Composition of Australian goals and targets are obvious. Foods, Australian Government This paper makes a strong case for Printer, Canberra ongoing support for the local food (7) Thomas, S., & Corden, M. (1977) composition program, and for the use of Metric Tables of Composition of the Australian food composition Australian Foods, Australian database in all Australian nutrition Government Printer, Canberra programs and research studies. (8) English, R., Lewis, J., & Cashel, K. While the value of good food (1990) Composition of Foods, composition data which are relevant to Australia, Volume 2, Cereals and the local food supply has been Cereal Products, Australian demonstrated in this study, high Government Publishing Service, standards in use of the data will not Canberra occur unless users are adequately (9) Lewis, J., & English, R. (1990) trained. Such education should include Composition of Foods, Australia, the need to know about local food Volume 3, Dairy Products, Eggs determinants, as well as how to use and and Fish, Australian Government interpret information such as different Publishing Service, Canberra analytical methods, modes of data (10) English, R., & Lewis, J. (1990) expression, and the rates for sampling, Composition of Foods, Australia, method choice and analytical quality Volume 4, Fats and Oils, Processed assurance (26,27). Meats, Processed Fruit and Vegetables, Australian Government Ŷ Acknowledgments Publishing Service, Canberra The Australian Bureau of Statistics (11) Lewis, J. & English, R. (1990) kindly made available unpublished Composition of Foods, Australia, details of the estimates of foods Volume 5, Nuts and Legumes, available for consumption. We thank Beverages and Miscellaneous Michael de Looper, Australian Institute Foods, Australian Government of Health and Welfare, for assistance Publishing Service, Canberra with the WPRDI calculations. (12) Lewis, J., Holt, R., & English, R. (1992) Composition of Foods Australia, Volume 6, Infant Foods, Government Publishing Service, Australian Government Publishing Canberra Service, Canberra (20) Warren, B., & Channon, H. (c1990) (13) Australian Bureau of Statistics Lamb Cutting Notes 1. More Fat (1993) Apparent Consumption of Means Less Saleable Meat, Foodstuffs and Nutrients, Australia, Rutherglen Research Institute, 1990–91, ABS, Canberra Victoria (14) English, R. (1987) Towards Better (21) National Health and Medical Nutrition for Australians, Australian Research Council (1992) Dietary Government Publishing Service, Guidelines For Australians, Canberra Australian Government Publishing (15) Health Targets and Implementation Service, Canberra Committee (HTIC) (1988) Health for (22) National Health and Medical All Australians, Australian Research Council (1989) Report of Government Publishing Service, the 108th Session of the Council, Canberra Australian Government Publishing (16) Holland, B., Unwin, I.D., & Buss, D. Service, Canberra. (1988) Third Supplement to (23) English, R., & Bennett (1990) Med. McCance and Widdowson's The J. Aust. 152, 582–586. Composition of Foods, 4th edition: (24) English, R., Cashel, K., Lewis, J., Cereals and Cereal Products, Royal Bennett, S., Berzins, J., Waters, A., Society of Chemistry, Nottingham & Magnus, P. (1988) National (17) Holland, B., Unwin, I.D., & Buss, D. Dietary Survey of Schoolchildren (1988) Fourth Supplement to Aged 10–15 Years, No. 1, Foods McCance and Widdowson's The Consumed, Australian Government Composition of Foods, 4th edition: Publishing Service, Canberra Milk Products and Eggs, Royal (25) Cashel, K., & Greenfield, H. (1995) Society of Chemistry, Nottingham J. Food Comp. Anal. (in press) (18) Cashel, K., & Greenfield, H. (1994) (26) Greenfield, H. (1990) Food Aust. Br. J. Nutr. 71, 753–773. 42, S1–S44 (19) National Health and Medical (27) Greenfield, H., & Southgate, D.A.T, Research Council (1992) Food Composition Data: Recommended Dietary Intakes For Production, Management and Use, Use in Australia, Australian Elsevier Applied Science, London Food Composition Data and Population Studies

Quality Control in the Use of Food and Nutrient Databases for Epidemiologic Studies

I. Marilyn Buzzard, Sally F. Schakel, Janet Ditter-Johnson

Nutrition Coordinating Center, Division of Epidemiology, School of Public Health, University of Minnesota, 1300 South Second Street, Minneapolis, MN 55454-1015, USA

This paper describes procedures used by the Nutrition Coordinating Center (NCC) at the University of Minnesota for maintaining food and nutrient databases. NCC's databases are designed to support an automated system for dietary data collection and nutrient calculation for clinical trials and other nutrition research and large population-based studies. The three major databases include the Nutrient Database, the Food Database, and the Brand Database. The Nutrient Database consists primarily of food composition data for “core” (non-recipe) foods. The Food Database drives the system's interactive prompting for detailed food descriptions and specification of amounts consumed. This database also contains all of the non-nutrient data required to link food descriptions with one or more entries in the Nutrient Database and convert amounts consumed to gram weights for nutrient calculation. The Brand Database, which contains food and nutrient information for commercial products, is used to update the other two databases. Each of these databases and the quality control procedures used for maintaining them are described. Because many of the studies using NCC databases are long term projects, time-related database maintenance and quality control procedures are required. These procedures permit routine updating to reflect the changing marketplace and the availability of new or improved data, while also ensuring comparability of dietary data collected at any point in time.

he Nutrition Coordinating Center (NCC) at the University of Minnesota maintains food and nutrient databases to support a system for the collection and nutrient Tcalculation of dietary data. The system is designed primarily for clinical trials and other medical research and epidemiologic studies investigating relationships between diet and health (1–3). The Minnesota nutrition data system has been used for hundreds of research studies over the past two decades. The majority of these studies have been funded by the US National Institutes of Health. A brief overview of the requirements for the system will provide a basis for understanding the functionality of the databases that drive the system. interested in sodium intake may opt not Ŷ Overview of the Minnesota Nutrition Data System to ask questions about salt use in food preparation, and the system will The Minnesota nutrition data system automatically assign the default was designed to meet the needs of its amounts. If one study needs foods usersprimarily large, population-based added to the database for a special nutrition research studies. These needs study population, these foods become include the following: standardized available to all users. procedures for collecting food intake The three major components of the data, especially for multi-centered current version of the Minnesota studies collecting data at many different nutrition data system are shown centers by many different individuals; a schematically in Figure 1. They include high level of specificity for describing interactive data collection, automated foods, including methods of food coding, and nutrient calculation. Over preparation and brand identification; the years the system has evolved to food and nutrient databases that are take advantage of technological frequently updated and which contain no advances and more accurate and missing values of nutritional efficient procedures for collecting and significance; a food composition processing the data (4). When the database that continues to expand to system was first developed about 20 include the nutrients of current research years ago, the first two steps were not interest; rapid data processing; and automated (1,2). Food intake data were database maintenance procedures that collected on paper, and the quality of permit accurate comparison of food and the data depended largely on the skills nutrient intakes over time. of the interviewer or on how well To avoid the need to maintain subjects were trained to keep food different versions of the system for records. The data collection process has different users, the most stringent now been automated so that the requirement demanded by any one user computer provides all of the prompts is provided to all users. For example, if required to describe foods at the only a few studies need specificity for appropriate level of detail (5). sodium, this level of specificity is provided to everyone; those who are not

Figure 1. Major components of the Minnesota nutrition data system Figure 2. The three databases that drive the Minnesota nutrition data system Similarly, coding was initially done on time and effort, since coding has paper by trained food coders, and the traditionally been the most labor coded data were then entered into the intensive part of processing dietary data. computer by data entry operators. The three major components of the Despite the use of duplicate data entry automated system have now been and subsequent computerized edit incorporated into a software package checks, there was potential for that is currently being used at transcription and other errors. About ten approximately 150 research institutions years ago, an on-line coding system in the US and Canada (5,7). A was developed which allowed coders to customized version of the system was enter the data directly into the computer developed for collecting 24-hour dietary (6). Edit checks could then be invoked recalls for the Third National Health and at the point of data entry, which greatly Nutrition Examination Survey (NHANES enhanced the accuracy and efficiency of III) which is now in its fifth year of data the coding process. collection (8). The final step of completely The three databases required to automating the coding process was not maintain the automated nutrition data possible until we had completed the system are shown schematically in development of the interactive data Figure 2. They include the Nutrient collection component. Only when all of Database, the Food Database, and the the detail required for coding is captured Brand Database. Each of these by the computer can the coding be databases will be described in greater totally automated. This enhancement, detail below. which was implemented about five years Ŷ The Nutrient Database ago, resulted in substantial improvements in accuracy and standardization, as well as savings in The NCC Nutrient Database is the and/or the method used to impute the smallest of the three databases. It data, if applicable (9). The database consists of the following data fields: also includes fields for three different ƒfood codes for approximately 1700 food grouping schemes to facilitate “core” (i.e., non-recipe) foods analysis by food groups and to accommodate different research ƒfood descriptions for each food code objectives. ƒamounts per 100 g for each of 94 Sources of data for the NCC Nutrient food components for each food Database are described in detail ƒreference codes for each nutrient elsewhere (9). The primary sources are value USDA data tapes and publications, information from manufacturers of brand ƒserving sizes designated by the US name products, and the scientific Food and Drug Administration and literature. The USDA Nutrient Data Base the US Department of Agriculture for Standard Reference is the major (USDA) for labeling purposes USDA data set used by NCC (10). The ƒfood group designations for several USDA Survey Database (11) provides different food grouping classification many imputed values that are not schemes. included in the Data Base for Standard Our philosophy is to keep the number Reference. USDA Handbook No. 8 (12) of foods in the Nutrient Database as provides additional information not small as possible to minimize included in the Standard Reference data maintenance efforts and facilitate rapid sets such as specific factors for updating (6). Foods are included only in calculating energy values, standards for their “as eaten” state; for example, foods enrichment of grain products, and that are never eaten raw are not values for the amounts of separable included in this database. The majority lean and fat of retail beef cuts. Other of the foods in the Nutrient Database USDA data sets used by NCC include are single ingredient foods, but there are various provisional tables and bulletins also a number of commonly consumed (9). multi-ingredient processed foods such Nutrient data for brand name as cheese, bread, and sausage. Each products are becoming increasingly food entry is described in detail in a text important as the consumption of field; the Latin or scientific name is also processed foods continues to increase included if applicable. in the US. Values for commercial The 94 food components in the products are obtained from the NCC current NCC Nutrient Database include: Brand Database (described below). The energy; the proximate nutrients (protein, scientific literature is another important fat, carbohydrate, and alcohol, plus source of nutrient data, especially for water and ash); animal and vegetable those nutrients included in the NCC protein plus 18 amino acids; 23 database that are not currently provided individual fatty acids; cholesterol; starch; by USDA. Food composition tables from six simple sugars; total dietary fiber and other countries are occasionally used to three fiber fractions; nine minerals; 17 obtain values for foreign foods not vitamins, including two vitamin A included in the USDA data sets. fractions and four fractions of vitamin E; Ŷ The Food Database plus caffeine, aspartame, and saccharin. Every nutrient value in the database is The NCC Food Database exceeds the associated with a reference code size of the Nutrient Database by documenting the source of the data approximately ten-fold. It includes the hierarchy of food descriptions that drives large piece; slice; or package), raw to the interactive prompting for detailed cooked yields, and edible portion food identification (5, 7). The hierarchy conversions. Also included is a consists of about 17,000 food maximum serving size for each food in descriptions for foods consumed in the database to serve as a quality North America. This includes brand control check for unusually large name descriptions as well as generic amounts. descriptions, in addition to a large Sources of data for the Food number of ethnic and regional foods, Database are documented by reference dietary supplements, and medications codes. The primary sources are the containing caffeine and sodium. The coding manual section of the USDA hierarchy is organized in a manner that survey database (11) and information reflects the way people think about from manufacturers. Several other foods, rather than according to any USDA publications related to food scientific classification. The hierarchical weights, yields, and portion sizes are organization facilitates the prompting for also used for the Food Database. These food description detail by presenting a publications are referenced by Schakel series of menu selections that become et al. (9). progressively more detailed until the NCC currently maintains two food is adequately described. separate versions of the Food For each food in the hierarchy of food Database. Food descriptions in the one descriptions, the Food Database version are linked to the NCC Nutrient provides all of the data required to link Database, whereas in the other version, the food with one or more entries in the the foods are linked to the USDA Survey Nutrient Database and to convert Database (11). The latter version, which amounts expressed in various common has been customized for collecting 24- units to gram weights for nutrient hour dietary recalls for NHANES III (8), calculation (5,7). Examples of other data includes some additional modifications fields in the Food Database include to enhance comparability of nutrient codes that designate the type of food calculations with calculations from the preparation method; ingredient listings USDA surveys. and amounts for recipes and Ŷ The Brand Database formulations; designation of ingredients that require further description (such as The NCC Brand Database contains food the type of fat used in a recipe); default and nutrient information for selected assignments, based on nationally categories of commercial products. This representative market research or food database continues to grow as the food consumption data, which designate the marketplace expands and changes. In most common of the available options the US there are approximately 1000 when a subject cannot provide the level new products introduced every month of detail requested; any geometric (13). So trying to keep up with even the shapes (e.g., cube, sphere, wedge, or most popular foods is a never ending cylinder) in which the food might be process. We currently maintain data for described; one or more density about 7000 products in the Brand conversion factors, depending on the Database. Information from this various forms in which a food can be database is used routinely to update measured (e.g., solid, chopped or both the Nutrient Database and the grated); and other amount conversion Food Database. factors, such as the weights of food- Brand name information is used for specific portions (e.g., small, medium, or several different purposes. In some cases it is needed to adequately identify brand facilitates identification of the the food that is consumed. For example, appropriate caffeine level. Another very subjects may describe a food by its important use of brand information is to brand name, such as “Coke,” rather help determine the amount of the food than by a more generic description, consumed. For example, a subject may such as “car bonated cola beverage.” report consuming a small container of Brand name data may also be needed low fat yogurt. Since low fat yogurt is to differentiate between similar products available in several different “small” that differ significantly in composition. sizes, knowledge of the brand will often For example, if different brands of permit accurate determination of the carbonated cola beverages differ with amount consumed. respect to caffeine content, knowing the

Figure 3. Example of a screen from the Food Database Maintenance System used for the NHANES III survey Although our goal is to update each rapidly might not be completely updated food category at least annually, changes for several years, other than obtaining in the marketplace often determine manufacturers' information on a few new priorities for updating. For example, we products that appear on food intake might update frozen entrees and ready- records obtained from research to-eat cereal several times during the subjects. year due to the influx of many new Not all brand name food categories products and product reformulations in are included in the Brand Database. If these categories, whereas other there are no significant differences categories that are not changing so among brands within a food category with respect to nutrient content or available; package size; serving size; serving amount, that category is not servings per package; ingredient listing; included in the database. For example, and preparation instructions. Nutrient brand name canned vegetables are not values provided by the manufacturer, included in the database. The current including label data as well as analytical version of the Brand Database includes or calculated data, are included in the 25 food categories. database. Although analytical data are Examples of the types of data fields preferred for nutrient calculation, studies in the Brand Database include: the sometimes prefer label values for product code, an arbitrary number developing educational materials. assigned by NCC; the product name; a Nutrient values obtained from other detailed description of the product; the sources, such as from the literature, are name of the manufacturer and the also included. Sources of all information product's Universal Product Code are indicated by reference codes, and (UPC); the product category the dates of receipt of the data at NCC designation; density information, if are noted.

Table I. A partial list of NCC edit limits for entering nutrient values Nutrient Food groups Limit/100 g Alcohol Alcoholic beverages 16 g Other 0 g ȕ-carotene equivalents Fruits/vegetables 7000 µg Margarine 1100 µg Other 800µg Calcium Cheeses 1000 mg Other dairy, soups, sauces, candy 300 mg Cold cuts, seafood 120 mg Other 100mg Cholesterol Eggs 1700 mg Cold cuts, organ meats, shellfish 500 mg Animal fat, shortening 230 mg Dairy products 140 mg Meat, poultry, fish 100 mg Salad dressing, gravy 80 mg Bread, crackers 75 mg Other 20 mg

databases described above. These Ŷ Quality Control Procedures procedures include the following: data Quality control procedures are critical for evaluation based on established criteria; maintaining food and nutrient databases automated data entry whenever because the potential for error in dealing possible; comparison of new with pre- with hundreds of thousands of data existing data; wellorganized data entry elements is very great. Quality control screens; edit checks at the point of data procedures are designed both to reduce entry; review of all manually entered the potential for error and to increase data; checks for consistency among the likelihood of identifying errors that related data fields; and review of data occur inadvertently. The quality control fields within food groups for consistency procedures used by NCC for database with expected ranges of values. maintenance are similar for the three When data are available from new default code number was globally multiple sources, criteria are used to inserted into all recipes that included the select the most appropriate values (9). default margarine. Analytical data are generally preferred Well-designed data entry screens over calculated data; however, the can reduce the potential for error. Such quality of the analytic procedures used, screens may be formatted in a manner the extent to which the data represent that simulates the format of the input nationwide sampling and eating habits, documents, thus decreasing the amount and the currency of the data are also of eye movement required by the data important considerations. USDA data entry operator. An example of a well are generally preferred over other data designed data entry screen is shown in sources, and refereed publications are Figure 3. Note that the different types of preferred to other publications, such as data are separated by labeled boxes, so meeting proceedings or text books. If no specific data are easy to locate. The published data are available, we screen is not too crowded, which makes occasionally use unpublished data, such it easy to view. as those provided by a reputable To further reduce the potential for laboratory. Because missing nutrient error, data are entered into the database values are calculated as zeros, they can in the format in which they are received. result in significant underestimations of For example, nutrient values provided nutrient intakes (14, 15). Therefore, by the manufacturer may be entered when values are not available from any into the Brand Database as amount per sources, we calculate or impute them serving, amount per Reference Daily using established procedures (3, 15– Intake (RDI), or amount per 100 g. Edit 17). A great deal of nutritionist effort is checks at the point of data entry permit involved in imputing data to ensure that immediate correction of keying errors. there are no missing values of nutritional These checks flag data that are out of a significance in the database. Imputed given range or do not conform to other values are replaced with analytical data field specific restrictions. A partial list of when they become available. edit limits used for entering nutrient Whenever possible, data entry is values is presented in Table I. Since automated to enhance efficiency and nutrient composition may vary reduce the potential for data entry substantially among food groups, the errors. Most USDA data are now limits are usually food group specific. available in electronic form, and we are Another type of edit check is the able to link many of our Nutrient flagging of incomplete data sets; for Database entries with USDA food example, some data fields require that descriptions via the reference codes. the reference codes be designated We hope that manufacturers will before the computer will accept the eventually provide product information data. electronically, which would greatly All data entered manually into the reduce the potential for error. When database are cross-checked by a database changes are required for an second database nutritionist. Data that entire classification of foods, rather than are manually entered into the Nutrient for a single food item, a computer Database are stored in a temporary file. program is written to implement these Each database nutritionist is assigned a global changes. For example, when the color that identifies the values entered default for “margarine, regular stick, by that individual; if changes are brand unknown” was changed to reflect recommended by the second more recent market research data, the nutritionist, these changes must be verified by the nutritionist who originally compare new data from manufacturers entered the data. Only after verification with any pre-existing data for the same of any changes are the new data product. If differences are noted, the accepted for posting to the permanent manufacturer is contacted to determine data file. if the differences are due to a Reports are routinely generated reformulation of the product or to comparing new data with previous data. improved composition data. This For example, when a new version of the information allows us to determine USDA Nutrient Database for Standard whether a new entry needs to be Reference is installed, a report of created or an existing entry updated. differences between the new and the For this reason, we routinely request pre-existing values is generated. Any information for all food products unusually large differences are verified marketed by a manufacturer, not just for through communication with USDA staff. the new products. Similarly, reports are generated to

Table II. Examples of relational edits for the Nutrient Database Compare: With: Acceptable difference: Sum of proximatesa 100 g ±5 g 4(pro)+4(carbo)+9(fat)+7(alc)b Total energy ±12% Sum of amino acids Total protein -20% Sum of fatty acids Total fat -5 to -20%c Soluble fiber + insoluble fiber Total dietary fiber ±10% a proximates include protein, carbohydrate, fat, alcohol, water and ash b pro=protein; carbo=carbohydrate; alc=alcohol c acceptable difference depends on type of food

New versions of the Food and relational edits for the Food Database Nutrient Databases are generally include verification of rules such as: released concurrently. During the four- every recipe ingredient reported by week period prior to the release of the volume must have a density; and every new versions, all modifications to the food that can be described in terms of a databases cease while the efforts of the geometric shape must have a solid database nutritionists are devoted to the density. There are currently 26 of these various quality control procedures. relational edits for the Food Database. Three types of quality control reports are Although many of these edits are generated prior to the release of the invoked at the point of data entry, others new versions. One type, the relational must be verified before a new version of edits, are reports that examine the Food Database is released. consistency among different fields in the Another type of quality control check database (3). For example, appropriate conducted before the release of a new relationships among nutrients in the version of one of the databases is a Nutrient Database are verified by review of computer listings of selected comparison of calculated values with fields by food category. For example, expected values. The calculated values vitamin A values for all processed must fall within an acceptable range of cheeses in the Brand Database are the expected result. Table II lists a few scanned for any outliers which must examples of the 28 relational edits then be verified from the original data currently generated for verification of the source. This type of review is conducted Nutrient Database. Examples of for all nutrients in the Nutrient Database that are not included in the relational procedures ensures comparability within edit type of consistency checks shown long-term studies and among studies by in Table II. Many of the non-nutrient eliminating the confounding due to using data fields are also subjected to this different databases and coding practices type of review. For example, the options at different time periods. for various food shapes in the Food Database changes that reflect real Database are compared within food changes in the foods people are eating groups, and any differences must be or in food preparation methods, must be justified. differentiated from those changes that A final quality control report that is represent new or improved data for generated before the release of a new foods that have not changed. Time- version of the Food and Nutrient related changes include most Databases is the calculation of nutrients marketplace changes, such as new for a test set of menus specifically products, new serving sizes of existing designed to include a wide variety of products, product reformulations, and foods, as well as all of the functionalities discontinued products. Changes in food of the calculation software. These preparation methods, such as increased calculations are compared with trimming of fat from meats or use of less calculations from the previous versions salt and fat in recipes, may also of the databases; any differences must represent time-related changes. Non- be verified as the result of intended time-related database changes include modifications to the databases or to the such changes as new or improved calculation software. analytic data, or better data for The three NCC databases are calculation of nutrient retentions in currently maintained separately, but a cooked or processed foods. project is underway to integrate them Procedures for time-related database into a single database management maintenance require quality controls to system. This will enhance our quality ensure consistency among subsequent control by permitting us to automate database versions. For every change more of the data entry than is now made to the Food Database, the possible. For example, selected nutritionist must indicate whether or not information from the Brand Database the change is retroactive to previous will be able to be automatically versions of the database. The computer transferred to the Food Database rather will not accept a change unless this than having to be manually entered. information is entered. For example, improved data, such as more accurate Ŷ Time-Related Database Maintenance and Quality Control values for raw to cooked conversion Procedures factors, are always retroactive to all previous versions of the Food Database. The need for comparability of dietary Edit checks at the point of data entry data over time requires use of time- prevent making changes that would related procedures for maintaining the compromise consistency with previous Food and Nutrient Databases (18). versions of the database. For example, Time-related database maintenance deletion of a food in the Food Database procedures permit recalculation of cannot be retroactive to previous previously collected food intake data at versions because studies must be able any subsequent time to take advantage to edit food intake records collected on of new or improved data, including previous versions. updates to both the Food Database and New versions of the Food and the Nutrient Database. Use of these Nutrient Database are released approximately every six months. Time- example, NHANES III will have used 12 related changes are handled differently different versions of the Food Database for maintaining the Food Database, for collecting and coding dietary data which is used for data collection and over the six years of the survey. coding, than they are for maintaining the Because all editing of dietary recalls Nutrient Database, which is used for must be done using the version on nutrient calculation. The current version which the data were collected, the of the Food Database must always editing would be very cumbersome reflect the current marketplace. without the functionality of the Multi- Products no longer on the market must Version Food Database. be deleted from the database to make All of the quality control procedures sure that they are not selected when previously mentioned have been they are no longer available. For the incorporated into the maintenance Nutrient Database, however, foods must software for the Multi-Version Food never be deleted, since a new version of Database. Careful adherence to time- this database may be used for related procedures for maintaining calculating nutrients for dietary data databases allows investigators to collected at any time in the past. recalculate their dietary data Discontinued foods continue to be automatically at any future time to take updated to reflect improved nutrient data advantage not only of new nutrients and and the addition of new nutrients to the other food components that have been database. added to the Nutrient Database, but also Thus, for a long term study, many of improved nutrient and non-nutrient versions of the Food Database must be data that have become available. These used for collecting and coding the data, procedures make it possible to monitor whereas a single version of the Nutrient trends in foods and nutrient intakes over Database is used for calculating time, which is especially important for nutrients for the entire study. Whenever meeting the objectives of ongoing a new version of the Nutrient Database dietary surveys and long term research is released, dietary intake data for the studies. entire study may be recalculated on that Ŷ Acknowledgments version. This will ensure that the most current data are used for nutrient Funding to support this work has come calculations and that the calculations primarily from the National Heart, Lung, are comparable throughout the study. and Blood Institute and the National To facilitate ongoing editing of dietary Cancer Institute of the National data for long-term studies, NCC has Institutes of Health, Bethesda, MD, and developed a Multi-Version Food from the National Center for Health Database that collapses all existing Statistics, Centers for Disease Control versions of the Food Database into a and Prevention, Hyattsville, MD. single database. Each subsequent Specific grants and contracts include the release of the Multi-Version Food following: NIH/N01HV-12903, NIH/RO1- Database incorporates only those data HL-42165, NIH/R01-CA-36522, and which have changed since the previous CDC/200-89-7014. release of the database. This eliminates Ŷ References the redundancy that exists among individual versions of the database. Use (1) Dennis, B., Ernst, N., Hjortland, M., of the Multi-Version Food Database Tillotson, J., & Grambsch, V. (1980) permits the editing of food intake data J. Am. Diet. Assoc. 77, 641– 647 collected at any point in time. For (2) Tillotson, J.L., Gorder, D.D., & (11) US Department of Agriculture Kassim, N. (1981) J. Am. Diet. (1993) Nutrient Data Base for Assoc. 78, 235–240 Individual Food Intake Surveys, (3) Sievert, Y.A., Schakel, S.F., & Release 6, National Technical Buzzard, I.M. (1989) Contr. Clin. Information Service, Springfield, VA Trials 10, 416–425 (12) US Department of Agriculture (4) Buzzard, I.M., Price, K.S., & (1976- ) Composition of Foods: Feskanich, D. (1991) in The Diet Raw, Processed, Prepared, USDA History Method, L. Kohlmeier (Ed.), Agric. Handbook No. 8 series, Smith-Gordon and Company, USDA, Washington, DC London, pp. 39–51 (13) Gorman, B. (1990) Prep. Foods (5) Feskanich, D., Buzzard, I.M., New Prod. Ann. 159, 16–18, 47–52 Welch, B.T., Asp, E.H., Dieleman, (14) Buzzard, I.M., Price, K.S., & L.C., Chon, K.R., & Bartsch, G.E. Warren, R.A. (1991) Am. J. Clin. (1988) J. Am. Diet. Assoc. 88, Nutr. 54, 7–9 1263–1267 (15) Posati, L. (1985) in Proceedings of (6) Buzzard, I.M. (1989) in Nutritional the Tenth National Nutrient Status Assessment of the Databank Conference, National Individual, The Food and Nutrition Technical Information Services, Press, Inc., Trumbell, pp. 87–98 Springfield, VA, pp. 124–133 (7) Feskanich, D., Sielaff, B.H., Chon, (16) Schakel, S.F., Warren, R.A., & K., & Buzzard, I.M. (1989) Comp. Buzzard, I.M. (1990) in Proceedings Meth. Prog. Biomed. 30, 47–57 of the 14th National Nutrient (8) McDowell, M., Briefel, R.R., Databank Conference, The CBORD Warren, R.A., Buzzard, I.M., Group, Inc., Ithaca, NY, pp. 155– Feskanich, D., & Gardner, S.H. 165 (1990) in Proceedings of the 14th (17) Westrich, B.J., Buzzard, I.M., National Nutrient Databank Schakel, S.F., & McGovern, P.G. Conference, The CBORD Group, (1993) in Proceedings of 18th Inc., Ithaca, NY, pp. 125–131 National Nutrient Databank (9) Schakel, S.F., Sievert, Y.A., & Conference, Washington, ILSI Buzzard, I.M. (1988) J. Am. Diet. Press, p.276 Assoc. 88, 1268–1271. (18) Buzzard, I.M. (1991) in Proceedings (10) US Department of Agriculture of the 16th National Nutrient (1993) Nutrient Database for Databank Conference, The CBORD Standard Reference, Release 10, Group, Inc., Ithaca, NY, pp. 73–77 USDA, Washington, DC Food Composition Data and Population Studies

Construction of a Database of Inherent Bioactive Compounds in Food Plants

Andrew D. Walker, Roger Preece

Institute of Food Research, Computing Group, Norwich Research Park, Colney Lane, Norwich, NR4 7UA, UK

Jenny A. Plumb, Roger Fenwick, Bob K. Heaney

Institute of Food Research, Food and Molecular Biochemistry Department, Norwich Research Park, Colney Lane, Norwich, NR4 7UA, UK

umerous incidences have been recorded where naturally-occurring dietary components have contributed to chronic and acute illness and occasionally to Nhuman fatalities. The causative agents, termed “natural (or inherent) toxicants”, are commonly present in food plants in order to provide protection against fungal, insect and herbivore attack. Perhaps the best known example of the effects of such a toxicant is the severe gastrointestinal and neurological disturbances observed following consumption of damaged, green or sprouted potatoes containing high levels of glycoalkaloids. Other classes of compounds with well-defined physiological effects include glucosinolates (in brassica vegetables and condiments), lectins (in legumes), isoflavones (in soya), cyanogenic glycosides (in cassava and legumes) and psoralens (in parsnip and celery). In order to study the varying accessible/critically evaluated data on biological effects of these natural the occurrence and levels of inherent toxicants (or of naturally occurring non- biologically active compounds in foods nutritional compounds offering is not yet available. Researchers at the protection against heart disease or Institute of Food Research (IFR) have cancers), it is essential that the available designed and are currently compiling a information on the content of these database to include information on bioactive compounds in foods is readily occurrence, levels and factors affecting accessible to workers in the plant levels of natural toxicants, anti-nutrients science, food science, nutrition and and protective factors in foods. clinical areas. Although databases exist The database, constructed using the which contain data on the nutrient ORACLE Relational Database compositions of foods, readily Management System (RDBMS) (1), provides information on the levels of hundred and twenty foods have been specific compounds in food plants, on coded giving a total number of records the country of origin, the plant part of about 1100. analyzed, any preparation methods Together with the records from used; information on varieties and individual references, further fields have sample numbers is included together been entered including a thesaurus of with full citations for all references used. alternative compound names, a The relational data model used can be thesaurus of food names, and a textual used to construct the database using “comments” screen containing other RDBMS. The system was information on the way in which levels originally developed on a DEC VAX may change as a result of processing, minicomputer and has been storage, cooking, agronomic and successfully installed on an IBM PC 486 environmental conditions etc. All compatible. references relevant to this field have Primary literature searches on been entered. Output screens have specific natural toxicant occurrence and been developed linking food name, food factors affecting levels have been part, preparation method and compound carried out using CD ROM databases name. The mean levels of each (e.g. Agricola and Food Science and compound together with maximum and Technology Abstracts) and are routinely minimum levels are calculated from all updated using Current Contents on disk. the data in the database assigned a Collected references are read and any satisfactory quality code, and are relevant references cited therein are presented with supporting information additionally obtained for use as a on the number of records and secondary source of data. Reprints references used to obtain these data. describing occurrence data for a number Planned developments include of naturally-occurring toxicants are expansion of the number of foods and critically assessed for data quality compounds within the database, according to defined criteria. These inclusion of data relating to toxicological criteria, developed at IFR after wide effects, and the setting up of a consultation, contain guidelines on European database on non-nutrient acceptability of analytical method, in composition of foods. sampling, unequivocally-identified plant Ŷ Acknowledgment species, etc. Only data which satisfy these criteria are entered into the This project has been funded by the database. Ministry of Agriculture, Fisheries and Currently, eight compound classes Food. (covering 65 compounds) including Ŷ Reference glycoalkaloids, glucosinolates, psoralens (furocoumarins), (1) ORACLE 6 The ORACLE alkenylbenzenes, saponins and Corporation UK Ltd, Bracknell, UK hydrazines have been entered. One Section VI

Copyright, Food Industry and Food Safety Considerations

his last Session of the Conference was chaired by Professor Geoff M. Wilson of the University of New South Wales, and commenced with a keynote address Tentitled International and Australian Copyright Considerations in Data and Data Compilations by S. Ricketson. This was followed by papers on Non-nutrient Databases for Foods by K. Louekari (presented by V. Piironen), Food Composition Databases in the Food Industry by O. de Rham, The Databases of the Australian National Food Authority, by J. Lewis and S. Brooke-Taylor, Data Considerations for Nutritional Labeling in the United States, by J. Tanner, and Functional Foods for Specific Health Use - the Needs for Compositional Data, by K. Shinohara. These papers are all published in this Section. Copyright, Food Industry and Food Safety Considerations

International and Australian Copyright Considerations in Data and Data Compilations

Sam Ricketson

Faculty of Law, Monash University, Wellington Road, Clayton, Vic 3168, Australia

Copyright protection of data and data compilations both in Australia and internationally is both qualified and incomplete. The paper reviews the basic principles of copyright, and then considers their specific application to data, tables and compilations. The principal domestic law examined is the Copyright Act 1968 (Cth) and the case law arising under this Act and in jurisdictions with similar common law backgrounds, such as the USA and UK. The principal international instrument considered is the Berne Convention for the Protection of Literary and Artistic Works which is currently in the process of revision. The paper examines the scope of protection for data and tables and compilations of data, with particular reference to the requirements for protection, the exclusive rights obtained and the question of entitlements. Brief consideration is also given to non-copyright protection which may be available.

nlike other statutory intellectual property rights, such as patents, designs and trade marks, Australian copyright law, which is contained in the Copyright Act U1968, is not dependent upon a system of registration or compliance with any other kind of formality, such as the giving of notice or the deposit of copies. Protection arises once a work is “made”, that is, once it is reduced to some kind of “material form”. This form may be visible or invisible (1). Apart from this, all that needs to be established is that the author of the work is qualified for protection in Australia by reason of her nationality or residence status (2) as extended by the international conventions to which Australia is party (3), or by reason of first publication in Australia (4) or in another country which is party to the same international conventions as Australia (5). Protection is available to a wider range of productions under the general classifications of “works” and “subject matter other than works”. The first covers “literary, dramatic, musical and artistic works”, which in turn embrace a disparate collection of subcategories such as books, plays, paintings, tables and compilations, computer programs, photographs, buildings, choreographic works and sculptures (6). These are protected so long as they are “original” (7) (see further below). The second covers material of a more “industrial” character, namely sound recordings, films, broadcasts and published editions of works (8). Copyright protection confers quite extensive “exclusive” rights on the copyright owner. Of these, the most important are the reproduction, adaptation, public performance and broadcasting rights (2). Duration of protection is also extensive, the basic rule being that, in the case of works, it applies for the life of the author of the work plus 50 years (9). In the case of subject matter other than works, the basic term is shorter (50 years from first publication or making). Copyright in Relation to Databases in patent or design law. All that is required is that the alleged author has In essence, database are simply a contributed skill, time and effort to the particular species of the broader genus creation of the alleged work and that the of tables and compilations which have latter is not the result of copying from long been the subject of protection elsewhere (13). This approach has a under Anglo-Australian copyright law. liberating effect in many instances, in so Thus, the definition of “literary work” in far as it frees courts from the invidious subsection 10 (1) of the Copyright Act task of comparative aesthetic judgment 1968 includes in considering the eligibility for “… (a) a table, or compilation, protection of works such as the pulp expressed in words, figures or novel, the hackneyed dramatic script symbols (whether or not in a visible and the amateur painter's daub. It has form); …” also provided protection to a vast array The words in brackets seem to be an of subject- matter that would otherwise express legislative indication that tables have little claim to be “literary” or and compilations expressed in “artistic” but which nonetheless embody electronic form are comprehended the results of the application of high within the meaning of literary work and levels of intellectual and/or physical this is confirmed in the Explanatory effort on the part of their creators. Early Memorandum to the Copyright instances of this included railway Amendment Bill 1984 which states that timetables (14), anthologies of poetry it was intended to include computerized (15), catalogues of merchandise (16) data banks which might not be and betting coupons (17); more recently, expressed in any visible form of notation protection has been accorded to (10). subject-matter as diverse as the forms However, it is not the case that all of a card index accounting system (18), tables and compilations (however a table of scores and winning symbols expressed) are automatically protected for a poker machine (19), computer under the Copyright Act 1968. As with programs in source code (20) and all other works, it is still necessary that engineering and design drawings (21). the table or compilation in question be Some of the above would properly be an “original” literary work in order to classified as tables or com- pilations qualify for protection (11). Anglo- and, while the protection accorded to Australian copyright law has never them as “original literary works” can be placed a particularly high premium on readily seen as an appropriate the requirement of originality in safeguard against third parties who comparison with other jurisdictions would “reap where they have not sown”, where some level of personal such protection nonetheless can lead to intellectual creation is often required difficulties. (12). There is no requirement of novelty The essence of a table or compilation or inventiveness, such as are necessary is that it comprises a mass of raw data or information, on which the tabulator or “The law of copyright does not compiler has then imposed a particular operate to give any person an order or arrangement. So far as the exclusive right to state or describe individual items of data are concerned, if particular facts” (24). these enjoy copyright protection this is In other words, the element of entirely separate from the protection selection and compilation involved in which may subsist in the table or this case was so small that to grant compilation (see further below): what is protection would have been equivalent protected in the case of the table or to protecting the raw data itself. There compilation is simply the element of are a number of other Australian and arrangement, selection or ordering that English decisions to similar effect (25), has gone into its construction. This is although this is an area where there is certainly an extension of the everyday room for judicial disagreement (26), and meaning of “authorship” and, for this there are a number of decisions to be reason, the courts have not been found in common law jurisdictions where prepared to confer protection on all only minimal quantities of selection and tables or compilations. Even if the unfair arrangement have been held capable of competition rationale of copyright attracting copyright protection (27). In protection is admitted, the courts have this regard, it is worth nothing that the still required that a particular level of skill courts have generally been reluctant to in compilation, selection and separate the mental skill and effort arrangement be displayed (22). involved in ascertaining or calculating a Where the skill of selection and particular item of information, such as a arrangement requires considerable wager on the outcome of a football literary knowledge and taste, as in the match, from the skill and effort which is case of an anthology of poetry, the applied to the actual presentation and judgment that the resultant compilation ordering of that item in the final table or is an original literary work is not a compilation for which protection is difficult one to make. Nonetheless, it is sought — chronological list of football clear that the element of compilation fixtures (28); weekly fixed-odds betting must be clearly identifiable, if not coupon (29); score table used on poker substantial, for protection to be machine (30). The effect of this is to accorded under this heading. Thus, protect the mental and physical effort Anglo-Australian courts have often been involved in collecting data just as much reluctant to extend protection to “mere as the effort devoted to ordering and lists”, where the skill applied has been presenting it. Thus, if a party wishes to simply that of gathering and presenting use that information for his own items of information in a fairly purposes, he must gather it himself, mechanical way. For instance, in the rather than take advantage of the famous High Court decision of Victoria plaintiff's efforts in compiling it. In the Park Racing & Recreation Grounds Co colorful words of a nineteenth century Ltd v Taylor (23), it was held that judge speaking in relation to a roads copyright did not subsist in information directory, the defendant must “count the posted by proprietors of a race course milestones” himself (31). On the other inside the course as to the names and hand, it should be stressed that this numbers of the starting horses, the protection of the data in a plaintiff's table horses scratched, the numbers of the or compilation is parasitic, in the sense winners and so on. In the words of that there must still be some element of Latham CI, the reason for this was that: tabulation or compilation which has been imposed upon that data by the plaintiff: there can be no protection just collation, Australian copyright law for raw unprocessed data and the effort should protect the particular format in that has gone into their collection. which the database compiler has In this regard, there may be now be a chosen to present the data. While the distinction between US and Anglo- minimum of skill required may not be Australian copyright law. In a recent high, it should nonetheless be decision, the US Supreme Court has remembered that the “authorial” quality made it clear that it will not give that is protected is that of tabulation or protection to “simple” compilations of compilation and the protection thereby data, where the elements of compilation given to the data is to the data so lack the necessary degree of “personal tabulated or compiled, not to the data intellectual creation” on the part of the themselves. compiler. On this basis, an alphabetical The qualification referred to above list of telephone subscribers was held concerns data compilations and tables incapable of attracting copyright stored in electronic form in, or for use in, protection (32). a computer or computer network. It is a By contrast, Australian and English requirement of the Copyright Act 1968 courts still appear satisfied by a lower that each protected work should have level of intellectual input by the compiler an “author” and, further, that this author (33). should be a human being (34). This is Copyright as Applied to Food by contrast with other subject matter Databases protected under the Act (sound recordings, cinematographic films, How are these principles to be applied broadcasts and published editions) to food databases or compilations? The where protection is given to the maker, immediate, and simple, answer is that broadcaster or publisher (as the case the above principles apply to these may be) and where these persons may kinds of compilations in exactly the be bodies corporate (35). same way as they apply to compilations The requirement of human in general. Thus, it will be necessary to authorship for compilations may lead to show that (a) effort has gone into the difficulty in the case of those stored in collection of the food data contained in electronic form. It is possible, perhaps the table or compilation, and (b) that probable, that in such cases the some identifiable skill and effort has necessary element of tabulation or been applied to the organization and compilation will not have been supplied presentation of that data, that is, that it by a human operator but by a separate is not simply an undigested mass of computer program that has classified “raw information”. Given the general and organized the data in accordance character of food data compilations and with the directions of the human tables, this necessary element of operator. While the computer program arrangement will usually be present. In itself may have a clearly identifiable this regard, and subject to one human author, it may be more difficult in qualification noted below, it should not such a case to identify the necessary matter that the database is in electronic element of human authorship in the form, as required for computer use and compilation which has been constructed storage, or whether it is in the form of with the use of that program. In the “hard copy”, whether printed or in recent British Copyright, Designs and microfiche or microfilm. So long as the Patents Act 1988, this potential difficulty data are presented with some minimum has been overcome by a specific degree of tabulation, ordering or provision dealing with “computer- contributed to the making of a database: generated” works: there will usually be a team of many “In the case of a literary, dramatic, persons who have participated in the musical or artistic work which is project. While Australian copyright law computer- generated, the author readily recognizes the concept of joint shall be taken to be the person by authorship (38), there may simply be too whom the arrangements necessary many persons to make this workable, for the creation of the work are particularly when questions of undertaken”(36). ownership and duration of protection While this provision is of general arise (39). Furthermore, many application to all works created by databases are continuing productions computer, for example, computer-aided that are added to on a regular basis, for design drawings, in the case of an example, by updating or revising entries, electronic database it would mean that incorporating new data, and providing the person who made the arrangements different kinds of methods of presenting for the collection of the data, their and analyzing those data. These storage in the computer and the use of additions to the database may come any compilatory program for the from one source, or may come from organization of the data would be many, for example, where the database regarded as the author (and therefore is networked and may be added to from first owner of copyright). any station on the network. In these In the absence of such a provision, circumstances, the copyright status of a the Australian copyright law is not so person who makes one or two of these certain, but I would suggest that a ongoing entries into the database is far similar result would probably be reached from clear. Problems of this kind by an Australian court faced with the indicate that the protection of electronic question. Thus, it can be argued that databases as original literary works there will still be the need for does not sit easily with the traditional considerable human input into the concepts of authorship and originality construction of an electronic database, that apply under the present Copyright even where a computer program is used Act. A provision of the kind contained in for this purpose. Decisions as to the the British Act given above would kind of data to be stored, the method of therefore assist in assigning authorship organization to be adopted and the ownership to one party alone, namely mode of retrieval will still need to be the person who makes the made by a human operator and the arrangements for the construction of the appropriate instructions given to the database. computer to achieve these results. By Ownership Issues. The determination way of rough analogy, the use of the of authorship questions in relation to a program can be seen as a tool or aid in database is of crucial significance in this process, in much the same way as determining the important issue of who a typewriter, word processor or camera owns the copyright in the database. Not can be seen as aids in the creation of only is there a requirement of human other kinds of literary and artistic works authorship for the subsistence of (37). copyright under the Australian Act, but Greater difficulties, however, may there is a general rule that the author is arise where there are a number of also the first owner of that copyright. persons involved in the creation of the This situation is modified in the case of compilation. Rarely will it ever be just employee authors, where the ownership one or two persons who have vests in the employer if the database has been made in the course of their the words “Opportunity Knocks” as employment (40). In other cases, it will the title of a television program (44) be necessary for the would-be copyright and a British court refused protection owner to receive an assignment (or for the word “Exxon” (45). The transfer) in writing of that copyright from unspoken reason for these decisions the author or authors (41). This will be of seems to be an application of the de vital significance where the compilation minimis principle, namely that such is prepared by a third party under some items are too insubstantial to be contractual arrangement. If the data are protected as literary works in their provided by A, and B then prepares the own right (46). An alternative view is compilation, in the absence of an that they are lacking in the express assignment B will own the necessary degree of originality (47), copyright in the resultant compilation. although this may be another way of Where there are multiple authors who saying the same thing. On the other are not A's employees, it will be hand, originality, in the sense of skill necessary to have separate and research, was not lacking in the assignments from each so far as their choice of the word “Exxon” in the contributions to the compilation are Exxon case, and the reason given concerned. by the English Court of Appeal for Protection of Data Used in Data denying it protection as a literary Compilations work was that, in itself, it conveyed neither “information and instruction, The above discussion has been [n] or pleasure, in the form of literary concerned only with the question of enjoyment” (48). Accordingly, the copyright protection for the data storing of a title or name in a compilation itself. What of the data that database will not generally raise any are used in the compilation? The copyright issue, and the same would answer to this depends on the nature be true with respect to single items and quality of the data in question. of data in relation to food and food However, the following general composition. propositions can be stated. ƒWhile single items of data may not ƒThere can be no copyright protection constitute original literary works in for “simple” facts on their own, for their own right, it is possible that a example, a statistic, chemical combination or assembly of data formula, short verbal description, or relating to a particular matter could the like. It is possible that data in any attract protection as a table or of these forms could form part of a compilation on the same principles food database. However, copyright discussed above. An example might law has always refused to protect be a table which analyzes the facts or items of an insubstantial nutritional content of a particular nature. Thus, there is a series of old food (49). If this were so, it would be English cases in which it was held necessary for the compiler of a that copyright did not subsist in the database who wished to include that titles of books (42) and, in one table within his own compilation to celebrated later case, the Privy seek permission from its authors. Council denied protection to the title From some of the examples I have of a song, “The Man who Broke the seen, it is possible that this has not Bank at Monte Cristo” (43). More happened in many situations recently, a New Zealand court has involving the compilation of food held that there was no copyright in composition databases, and strictly this would involve an infringement of essentially gives protection against the copyright in any separate tables copying or derivation. The latter need or compilations of data that were not be literal or “word-for-word”: it is used without permission (50). sufficient if the copy is substantially ƒIt is also possible that some similar or only “colorably” different from individual items of data will be the original. It is also unnecessary for capable of attracting protection in the copy to be in the same medium as their own right if they are more the original: a printed work may be “substantial” in size than those copied in any “material” form (51), and referred to in the first paragraph. It is this would include storage in the hard impossible to indicate what the disk of a computer, on a CD ROM, on quantum required here will be, but, film or microfiche, or magnetic tape. Nor on occasion, the courts have been is it necessary for the whole of the prepared to protect works such as protected work to be taken: under short poems, newspaper subsection 14(1) of the Copyright Act summaries, a series of numbers, 1968, it is enough if a “substantial” part and abridgments or abstracts of is appropriated. “Substantiality” in this longer works, on the basis that they context is not simply a reference to displayed sufficient original effort on quantity: Anglo-Australian courts have the part of the alleged author. It is always stressed that the quality of what not clear how relevant this head of is taken is just as important in protection might be to the compilers determining whether a substantial part of food databases, but possibilities of a work has been taken (52). Thus, would include written analyses and courts will look at the significance or summaries of the nutritional importance of the part taken in relation components and value of particular to the work as a whole, and it may not foods, abstracts of longer papers or matter that a relatively short section is reports, items of explanatory taken. material, and so on. Once these In the case of databases, the works have reached a certain length requirement of substantiality can give and go beyond the simple factual rise to problems. As noted above, the items referred to in the first protected aspect of such works is the paragraph, the possibility of element of selection and arrangement protection will arise and the that has gone into their construction, permission of the owner of that and does not extend to the individual copyright will need to be sought in items that make up the compilation. The respect of any use made of it in a latter may enjoy copyright protection in database. their own right, or not at all, as the case may be, but this is beside the point. If The Scope of Copyright Protection in the allegation is that X has reproduced Relation to Data and Data part of the database of Y, this claim will Compilations only be good if that act has been done As noted above, the exclusive rights in relation to a substantial part of that conferred on owners of copyright in database. Taking individual items of literary works under the Copyright Act data, or even a number of items, will 1968 are quite extensive. It is worth usually not be enough: liability will only saying something further about the more arise where a substantial part of the important of these rights. arrangement or compilation of data is Reproduction. This is the most basic taken. This may provide a severe right of the copyright owner, and limitation on the exploitation of their rights by the owners of copyright in the compilation in deciding what to databases, as is well illustrated by the include. If this is the case, then the English case of Warwick Films v taking of several items that would not Eisinger (53). The compilation here otherwise form a substantial part of the comprised edited transcripts of the trials compilation may well constitute an of Oscar Wilde, but, although the infringement of the compilation as defendant had taken considerable extended to this element of initial extracts from the transcripts (in which selection (57). In such a case, the the plaintiff had no rights), these defendant has helped himself to the portions did not contain any of the efforts and skill of the plaintiff, rather plaintiff's editorial changes or than going back to the data himself and accompanying commentary. making his own selection. In this Accordingly, as they now lacked the respect, however, the technical and element of selection and arrangement storage capabilities of computer which the plaintiff had supplied in his databases may be a disadvantage so compilation, they did not constitute a far as the availability of protection is substantial part of it (54). The same concerned. It is now possible for these point was also made in Ladbroke to be comprehensive within their (Football) Ltd v William Hill (Football) particular field and this, indeed, may be Ltd (55), a case involving a claim for the essence of their commercial appeal. copyright in the layout of a football Thus, if a database contains every item coupon. On the question of whether the of information relevant to a particular appellants had reproduced a substantial field, there can be no element of part thereof, Lord Pearce said: selection that is infringed by the taking “Whether a part is substantial must of any number of these items. It will only be decided by its quality rather than be if the arrangement or collocation of its quantity. The reproduction of a these items within the database is also part which by itself has no originality taken that any question of the taking of will not normally be a substantial a substantial part of the overall part of the copyright and therefore compilation will arise. In the case of food will not be protected. For that which data, unless the individual items are would not attract copyright except by capable of copyright protection, there reason of its collocation will, when may be no infringing act on the part of a robbed of that collocation, not be a person who simply takes the data and substantial part of the copyright and stores it in her own database where it is therefore the courts will not hold its used with that person's data reproduction to be an infringement” management system. (56). Adaptation. A further exclusive right In crude terms, the principle here can of a copyright owner is the right to adapt be expressed as follows with respect to the work. “Adaptation” in this context databases: it will only be the taking of covers a wide range of transformations the “processed” data that will attract of a work, such as dramatizations of copyright liability, in the absence of any nondramatic works, fictionalizations of other copyright that might subsist dramatic works, arrangements of separately in the data itself. musical works and translations from one Nevertheless, this principle may require language to another (58). Following modification in certain circumstances. amendments made in 1984, As noted above, an important part of a “adaptation” now includes a different compilation may lie in the skill and version of a computer program (whether judgement exercised by the maker of or not in the language, code or notation in which the program was originally There are three basic ways in which this expressed). None of these can occur in the case of electronic transformations seem apt to cover what databases: through the supply of might be done with data or a database external storage devices to the user, by a third party, for example, where the such as CD-ROMs; transmission by latter takes one or both and then puts television signals; and transmission over them in his own database where they wire, such as cable or telephone lines. are organized and used in accordance The first of these, sometimes referred to with an entirely different data as “off-line distribution”, clearly involves management program. Although it is the making and distribution of “copies” possible that the items of data and/or of the database and of any works stored the database may constitute original in the database. This is on the basis that literary works in their own right (see the storage devices can be regarded as above), the adaptation right is simply not “reproductions in a material form” of any relevant in the present context unless protected works that are stored in this one or the other of these works can also fashion. be regarded as a computer program. On As to the second, television the other hand, the reproduction right transmission as a means of may still be relevant here, and the initial disseminating information in computer act of storage may constitute an databases is apparently not common, infringing act, irrespective of what is although “teletext” or “videotext” done after this with the data or data systems using television signals have compilation. clear potential for expansion. From a Public Performance, Broadcasting, copyright point of view, however, Cable Diffusion. The exclusive rights of communication of data in this way will the copyright owner also extend to clearly be an infringement of the authorizing the public performance or exclusive broadcasting rights in any display of the work, broadcasting it, and, protected works comprised in the in limited circumstances, transmitting it transmission (60). to subscribers to a diffusion or wire To date, the third method of service (59). All these acts, together transmission by wire or cable remains with the reproduction right discussed the most common means of “on-line” above, are relevant to the way in which distribution of data stored in computer data may be retrieved and distributed. In systems. In so far as this involves the this regard, there is a marked contrast dissemination of protected works, the between electronic and “traditional” or provisions of the Copyright Act are hard copy databases, where the reasonably clear. One of the exclusive common means of retrieval is for the rights of copyright owners of original user to turn the pages or shuffle the works is the right to transmit those cards — acts that clearly do not involve works to “subscribers to a diffusion an infringement of the copyright owner's service” (61). This is a distinct activity to rights. With an electronic database or that of broadcasting, which means data stored in electronic form, however, transmission by means of wireless the processes of transmitting or telegraphy (62). By contrast, communicating data to the user may in transmission to subscribers to a themselves involve infringements of diffusion service entails the distribution copyright, while the actual retrieval of of “broadcast or other matter (whether this data by the user may involve other provided by the person operating the kinds of infringing acts. service or other persons) over wires, or Distribution of Data to Users over other paths provided by a material substance, to the premises of “Where visual images or sounds are subscribers to the service” (63). The displayed or emitted by any latter definition appears apt to cover the receiving apparatus to which they circumstances under which the are conveyed by the transmission of information from many computer electromagnetic signals (whether databases is distributed to users. The over paths provided by a material latter may well be “subscribers to a substance or not), the operation of service” within the terms of the any apparatus by which the signals legislation, namely through a service are transmitted, directly or indirectly agreement with the owner of the to the receiving apparatus shall be database, and will be receiving the data deemed not to constitute via visual display units or printers in their performance … but, in so far as the own premises. This would probably not display or emission of images or extend to networks within particular sounds constitutes a performance, institutions or networks provided on … the performance … shall be some kind of co-operative basis. deemed to be effected by the Retrieval of Data by Users operation of the receiving apparatus” (66). Irrespective of whether data are It therefore seems reasonable to transmitted to users by off-line or on-line regard the display of protected works on communication, the question still arises the visual display unit of a computer as to whether the retrieval and use of terminal as the “performance” of those those data by the user constitute an works by means of a receiving infringing act in relation to any copyright apparatus (where the work is work embodied in that material. communicated by cable or some other Basically, there are two ways in which material path) or “by any other means” retrieval may occur: through the display (where the work is retrieved offline). of data on a visual display unit or The next question is whether the act through the production of printed or of visual display occurs “in public”. No other “hard” copies. definition of this term is to be found in Visual Display. There are two ways in the Act (save for section 28 which which this activity may be analyzed excludes performances given under the Copyright Act 1968. The first exclusively in the course of educational is that the visual display is a public instruction and section 46 which performance of the work displayed (64). provides an exclusion for the visual The second is that this is a reproduction display of broadcast data where this is in a material form of that work (65). The done in premises where persons reside two are by no means mutually exclusive. or sleep), and it therefore remains a Public Performance. This does not matter for judicial interpretation. require performance by a human actor Essentially, the courts have adopted a or performer. Under subsection 27(1), a restrictive approach, confining non- reference to “performance” includes public performances to those in the “any mode of visual or aural domestic or quasi-domestic sphere (for presentation, whether the presentation an extreme interpretation of “quasi- is by the operation of wireless domestic” in the case of nurses and telegraphy apparatus, by the exhibition doctors living within the confines of of a cinematographic film, by the use of Guy's Hospital, see Duck v Bates (67). a record or by any other means; …” It is Thus, the fact that the “audience” is further provided that: limited in some way, for example, to those who pay an admission fee or who are members of a club or association, lasts so long as the human user desires will normally be irrelevant (68). The to have the work displayed on the prime consideration in determining screen. In this regard, reference can be whether a particular performance is “in made to the use of the word “storage” in public” is the character of the audience: the definition of “material form” in whether this is the type of audience subsection 10(1) of the Copyright Act which can be described as the copyright 1968. “Storage” implies some degree of owner's audience in the sense that the permanency and it seems a misuse of owner would look to deriving financial language to describe the transitory benefit from authorizing a performance display of a work on a visual display unit of his work before it. In the present as a form of storage. On the other hand, context, it is relevant to note a the definition of “material form” is significant New South Wales decision in inclusive only and it is therefore which this principle has been applied to arguable that “material form” can the case of free in-house movies that embrace other less permanent forms of were offered as part of the services to fixation. Thus, in an early case under guests at a motel (69). There was the Copyright Act 1911, it was held that clearly a market there for the display of a tableau vivant representing a cartoon such films and the persons who viewed from Punch Magazine was a them were members of the copyright reproduction in a material form of the owners' public, even if they viewed the cartoon (71). films in the privacy of their individual There is no reason under the present rooms. Another decision of the Federal Act why visual display should not Court of Australia has held that there simultaneously constitute both the act of was a public performance where video performance in public and the act of films were played to a small group in a reproduction in a material form. As a bank as part of an in-house training matter of principle, however, it is program (70). By the same token, where obviously undesirable that the one act users of a computer database retrieve should give rise to two distinct grounds information on a visual display unit, they of liability and it would therefore be are displaying or performing in public advantageous for the Act to indicate any protected work which appears on which one should apply to the exclusion the screen and it should not matter of the other. In this regard, it is of whether the unit is situated in a publicly interest to note the recent accessible place such as a library or in recommendation of the Commonwealth the office of an individual user. Copyright Law Reform Committee that Reproduction in a Material Form. The the Act should be amended to make it other alternative is that visual display is clear that screen displays do not a reproduction in material form of the constitute either a reproduction in work displayed. However, it is more material form or a public performance of difficult to reach a firm conclusion here. works stored in computer memory (72) There can be no doubt that, unlike Other Forms of Storage. This storage of a work in a computer hard expression is used simply by way of disk or CD-ROM, in this instance the contrast to visual display and is intended work is perfectly visible and to cover all other forms in which data comprehensible to the human eye. may be retrieved, including hard-copy Furthermore, it is often in a page format printout, facsimile reproduction or by that is similar to that of any printed way of transfer into other devices for version of the work. On the other hand, internal or external storage. The the form of fixation is transitory and only relevant right here is the reproduction right, and as a general proposition it Clause 13, where the reproduction or would appear that any acts of this kind adaptation of Australian food will involve an infringement of this right. composition data from the government However, a question remaining to be database is expressly permitted for resolved in each instance will be certain purposes, such as for display on whether a substantial part of the data or food labels for advertising and database has been reproduced, and the information purposes. From the matters that are relevant to this have copyright perspective, this kind of already been discussed above. permitted use should give rise to no Some Practical Issues. In the light of problems, so long as any specified the above, it will be clear that many limitations or conditions on the use are uses of the data contained in a observed (In the event that they are not, protected database will be controlled by this will be a breach of the license given the owner of the copyright in that and therefore an infringement of database and that permission will copyright). therefore need to be sought for the use Ŷ International Aspects of Copyright or uses that is sought to be made of this Protection data. As there is no registration system for copyright in Australia, it may not Several matters need to be noted here: always be easy to identify the copyright first, the protection of Australian data owner for the purposes of obtaining and databases abroad; secondly, the permission. Nonetheless, this does not protection of foreign data and databases excuse the intending user from seeking in Australia; and thirdly, some brief out the copyright owner, and it will be comparative treatment of developments dangerous to proceed in the absence of that are occurring in other countries with permission. As the use of data in respect to increased protection of these protected databases is not covered by kinds of subject matter. any compulsory license scheme under Protection Abroad and in Australia the Copyright Act 1968, there is no clearly established benchmark for the Australia is a party to the Berne level of fees or royalties that will be Convention for the Protection of Literary payable for the intended use. This will and Artistic Works which now binds over be a matter for negotiation between the 105 countries. These include almost all copyright owner and user, although in of the important developed countries practice there may be clearly accepted and a great majority of developing and levels of royalty that exist in particular former socialist states (the separate fields. In each instance, it will be states of the Commonwealth of necessary to comply with any other Independent Nations are significant conditions that are laid down by the exceptions, but are probably still bound copyright owner, for example, limitations to a lesser extent through their on the amount that may be used, the membership of another international need for acknowledgment of source, convention, the Universal Copyright and the like. Convention). At the same time, it should be noted The effect of the Berne Convention is that some database copyright owners to accord protection to Australian works do not object to appropriate use being in all other countries which are members made of their data where this is clearly of the Convention, and to extend similar in the public interest. This would seem protection here to works that originate to be the case under the Australian from those countries. So far as Australia Food Standards Code, Section A, is concerned, this is done through regulations made under the Copyright some variation as to the levels of Act 1968, the effect of which is to extend originality or intellectual creation protection in Australia to works with required. Relatively little attention, authors who are nationals of these other however, has been paid to the particular countries or to works which have been problems that may be raised by first published in one of those countries electronic databases. Thus, article 2(5) (73). In broad terms, this means that a of the Berne Convention provides that: British or French database will be “Collections of literary or artistic entitled to protection in Australia, in the works such as encyclopedias and same way as an Australian database, anthologies which, by reason of the and this will also be the case for selection and arrangement of their individual items of data that are capable contents, constitute intellectual of being treated as separate copyright creations shall be protected as such, works (see above). A similar position will without prejudice to the copyright in also apply so far as the protection of each of the works forming part of Australian data and databases in these such collections.” other countries is concerned. This This provision refers only to collections occurs without the need for any of copyright works, but as its emphasis formalities, such as registration, is upon the elements of selection and although in the USA registration will arrangement this does not seem to provide the foreign copyright owner with restrict member countries extending certain procedural advantages that may protection to collections of non-copyright make it easier for her to enforce her material that display the same qualities rights in that country. Some differences (74). This is, for example, the position in in treatment may arise from country to the Federal Republic of Germany, country, depending on the level of where “collections of works or of other originality or intellectual creation that is contributions” are protected as works required for protection, but, in any event, where, “by virtue of the selection or it is a fundamental requirement of the arrangement thereof, [they] constitute Convention that foreign and local works personal intellectual creations” (75). are to be treated in exactly the same Likewise, the US Copyright Act 1976 way (the principle of “national protects “compilations” of data as works, treatment”: Berne Convention, article defining “compilation” as: 5(1)). In other words, if German law, for “… a work formed by the collection example, requires a higher level of and assembling of pre-existing personal intellectual creation for an item materials or of data that are of data or a database, the same selected, co-ordinated or arranged in standard will apply to the foreign such a way that the resulting work claimant for protection as to the local as a whole constitutes an original claimant. work of authorship.” (76) International Comparisons with Similar provisions apply in Japan Respect to Databases (77), where an amendment made in 1986 now makes specific reference to It is only possible here to make brief “data base works”. “Data base” is reference to the substantive level of defined as “an aggregate of information protection for databases both such as articles, numerals or diagrams, internationally and in other jurisdictions. which is systematically constructed so As a general matter, compilations are that such information can be searched protected elsewhere on a similar basis for with the aid of a computer.” (78) It is to that under Australian law, but with then provided that these are protected as independent works, “where, by countries, there is a common provision reason of the selection or systematic that provides as follows: construction of information contained “Catalogs, tables and similar therein, [they] constitute intellectual productions in which a great number creations” (79). of items of information have been In general, it seems that many compiled, as well as programs, may national copyright laws provide not be reproduced without the protection for databases, irrespective of consent of the producer until 10 whether they contain protected works or years have elapsed from the year in not (80). However, it also appears that, which the production was published” as in Australia, these laws will generally (83). require the database to possess some A proposal for a similar kind of minimum level of originality or protection for electronic databases was intellectual creation by reason of its discussed at a WIPO/UNESCO selection and arrangement. These Committee of Governmental Experts on levels may differ between countries, and the Printed Word in Geneva in late the result is to produce uncertainty for 1987. It was submitted that this special database producers who wish to obtain protection should comprise the copyright protection on as broad a basis following: the exclusive right of database as possible. (For example, in the USA producers to authorize the reproduction, the number of protected databases may in any manner or form, of their now be considerably less following the databases; similar limitations to this right decision of the Supreme Court in Feist as are applicable in respect of literary Publications Inc v Rural Telephone and artistic works included the Service Co Inc (81). It may also be the database; and a minimum term of case that a number of databases that protection of 10 years from the end of would otherwise qualify for protection in the year in which the database is made that country will be in the public domain available to the public (84). These because they are “works of the United proposals were embodied in a series of States Government” (82). This draft principles. A final principle (85) uncertainty has led to international provided that: proposals that databases lacking the “The specific protection granted to necessary quantum of originality should data base producers according to nevertheless be entitled to special Principles PW17 to PW19 should protection so as to safeguard the leave intact and should in no way considerable investment that is affect the protection of copyright in represented therein. Such protection literary and artistic works included in would be analogous to that granted to electronic data bases.” subject-matter such as sound This received varying degrees of recordings and broadcasters under support from the 31 nations represented neighboring rights legislation in many on the Committee. Some delegations countries or under Part IV of the expressed their reservations concerning Australian Act, but would not be to the sui generis protection of databases, on prejudice of any fuller copyright the ground that this would fall outside protection that might be available for the scope of the international copyright truly original databases. A precedent for conventions and the principle of national this form of protection already exists in treatment (as occurs in the case of the some countries in respect of Nordic countries). It was also argued compilations lacking the necessary that such protection would serve to degree of originality. Thus, in the Nordic dilute copyright protection where it might otherwise be applicable and, on the food data and databases, brief other hand, might result in the protection reference should be made, for the sake of “fairly meager collections which were of completeness, to two other forms of not worthy of protection” (86). protection that may be highly relevant. Reservations were also expressed These are: (a) the equitable action of concerning the shortness of the breach of confidence, and (b) proposed period of protection and the contractual restrictions. difficulty that any requirement of Breach of Confidence publication might pose in particular countries. Nonetheless, although there This is a purely judge-made form of was no clear consensus among protection that safeguards information delegates as to the details of protection, that has been prepared and utilized in it does seem that there was a general circumstances of confidence (90). If this recognition that some protection was information is communicated in appropriate to safeguard the efforts and confidence to another party and the investment involved in the compilation of latter misuses or discloses the electronic databases (87). In a Green information to another without Paper on Copyright published in 1989, permission, the law will often grant a the Commission of the European remedy (either an injuction restraining Communities proposed that there the activity or some form of monetary should be a separate sui generis form of award) against the offending party. This protection for those databases ineligible is a completely informal form of for copyright protection (88). A protection that will depend very much subsequent Draft Directive (1991) upon the plaintiff establishing the prepared by the Commission and still confidentiality or secrecy of his under discussion creates a new right of information and some kind of improper “unfair extraction” in both non-original misuse of that information by the person databases and those protected by to whom it has been confided. It can copyright. even extend to third parties who come At the broader international level, the into possession of the information. The World Intellectual Property Organization action of breach of confidence is of (1992) has proposed modest great importance in the commercial and amendments to the Berne Convention, industrial area where it is used to protect but these would not represent any trade secrets, confidential data and advance on the current Australian law know-how. It therefore has obvious and do not go as far as the EC implications so far as the compilers of proposals. Finally, so far as Australia is databases are concerned, particularly concerned, the Copyright Law Review where these involve information that is Committee has taken the view that, with out of the public domain. some minor exceptions, there is no Contractual Restrictions need for changes in our domestic law in relation to databases (89). It is clear, As a practical matter, these may be the therefore, that it will be some time most effective form of protection as before there is international agreement between data providers and users. This on whether special treatment for will not affect third parties, that is, databases is warranted. persons who are not parties to the contract, but it can be very effective as Ŷ Other Forms of Protection between the contracting parties. Thus, While the bulk of this paper has been they could agree on the precise terms of concerned with the copyright status of use of data, the question of ownership of copyright in data that are created by (16) Purefoy Engineering Co Ltd v one or the other, the level of Sykes Boxall & Co Ltd (1955) 72 remuneration for use, e.g. display, RPC 89 retrieval, recompilation, and so on. (17) Ladbroke v William Hill (Ref 13) Obviously, much more could be said on (18) Kalamazoo (Australia) Pty Ltd v this topic, but, as a general matter, the Compact Business Systems Pty Ltd courts will seek to enforce what the (1985) 5 IPR 213 parties agree and, in specific cases, this (19) Ainsworth Nominees Pty Ltd and may be more effective than relying upon Ainsworth Holdings Pty Ltd v Anclar such heads of protection as copyright Pty Ltd (1989) 12 IPR 551 and breach of confidence. (20) Computer Edge Pty Ltd v Apple Computer Inc (1986) 60 ALJR 313 Ŷ References (21) S W Hart & Co Pty Ltd v Edwards (1) Copyright Act 1968, Section 10(1) Hot Systems (1985) 159 CLR 466 (2) Copyright Act 1968, Section 31 (1) (22) MacMillan & Co Ltd v K & J Cooper (3) Copyright (International Protection) (1924) 93 LJPC 113, 118 Regulations 1969 (23) (1937) 58 CLR 479 (4) Copyright Act, Section 31(2) (24) Ibid, 498 (5) Copyright (International Protection) (25) Chilton v Progress Printing and Regulations 1969 Publishing Co [1895] 2 Ch 28; (6) Copyright Act 1968, Part III Odham's Press Ltd v London and (7) Copyright Act, Section 31 (1)(2) Provincial Sporting News Agency (8) Copyright Act, Part IV [1935] Ch 672; Smith's Newspapers (9) Copyright Act, Section 34(2) Ltd v The Labour Daily (1925) 25 (10) Lahore, J.C. (1977) Intellectual SR (NSW) 593 Property Law in Australia: (26) G A Cramp & Sons Ltd v Smythson Copyright, 1st Ed., Butterworths, Ltd [1944] AC 329; ITP Pty Ltd v Sydney, pp. 280–282 United Capital Pty Ltd (1985) 5 IPR (11) Copyright Act 1968, Subsections 315 32(1) for unpublished works and (2) (27) Canterbury Park Race Course Ltd v for other works Hopkins (1932) 49 WN (NSW) 27; (12) West German Copyright Act of John Fairfax & Sons v Australian 1965 (Act Dealing with Copyright Consolidated Press [1960] SR and Related Rights of 9 September (NSW) 413. 1965), Article 2(2); Italian Law for (28) Football League Ltd v Littlewoods the Protection of Copyright and Pools Ltd [1959] 1 Ch 637 Other Rights Connected with the (29) Ladbroke (Football) Ltd v William Exercise Thereof (No 633 of 22 Hill (Football) Ltd [1964] 1 All ER April 1941, as amended), Article 1 465 (13) Water v Lane [1900] AC 539, (30) Ainsworth Nominees Pty Ltd and MacMillan & Co Ltd v K & J Cooper Ainsworth Holdings Pty Ltd v Anclor (1925) 93 LJPC 113; Sands & Pty Ltd (1989) 12 IPR 551 McDougall Pty Ltd v Robinson (31) Kelly v Morris (1866) 1 LR Eq 697 (1917) 23 CLR 49; Ladbroke (32) Feist Publications Inc v Rural (Football) Ltd v William Hill Telephone Service Co Inc (1991) (Football) Ltd [1964] 1 All ER 465 22 IPR 129 (14) H Blacklock & Co Ltd v C Arthur (33) Waterlow Publishers Ltd v Rose Pearson Ltd [1915] ILRC 2 Ch 376 (1990) 17 IPR 493; Kalamazoo (15) MacMillan v Suresh Chunder Deb (Australia) Pty Ltd v Compact (1890) 17 ILRC 951 Business Systems Pty Ltd (1985) 5 (49) Kalamazoo (Australia) Pty Ltd v IPR 215 Compact Business Systems Pty Ltd (34) Copyright Act 1968, Section 32 (1985) 5 IPR 213 (author must be a “qualified (50) Brown, R.L. (1985) Rutgers Comp. person”, meaning an Australian Technol. Law J. 11, 17–49 citizen, an Australian protected (51) Copyright Act 1968, Subsections person or a person resident in 10(1) and 22(1) Australia) (52) Hawkes & Sons (London) Ltd v (35) Copyright Act 1968, Section 84 Paramount Film Productions Ltd (36) Copyright, Designs and Patents Act [1934] Ch 593; LB (Plastics) Ltd v 1988, Section 9(3) Swish Products Ltd [1979] FSR 145 (37) Roland Corporation v Lorenzo and (53) [1969] Ch 508 Sons Pty Ltd (1992) 22 IPR 245, (54) Ibid, 385 252–253, per Pincus J (55) [1964] 1 All ER 465 (38) Copyright Act 1968, Subsection (56) Ibid, 481 10(1) (definition of “work of joint (57) Jarrold v Houlston (1859) 3 K & J authorship”) and Division 9 Part III 708, 69 ER 1294; Harman Pictures (39) Copyright Act 1968, Section 80 NV v Osborne [1967] 2 All ER 324; (40) Copyright Act 1968, Subsection Elanco Products Ltd v Mandops 35(6) (Agrochemical Specialists) Ltd (41) Copyright Act 1968, Subsection 197 [1979] FSR 46 (1). (58) Copyright Act 1968, Subsection (42) Maxwell v Hogg (1867) 2 Ch App 10(1) 307; Kelly v Hutton (1868) 3 Ch App (59) Copyright Act, Section 31(1) 203; Mack v Peter (1872) LR 114 (60) Copyright Act 1968, Sections Eq 431; Schove v Schminke (1886) 31(1)(a)(iv) (literary, dramatic and 33 Ch D 546; Licensed Victuallers' musical works) and 31(1)(b)(iii) Newspapers Co v Bingham (1888) (artistic works) 38 Ch D 139 (61) Copyright Act 1968, subparagraphs (43) Francis, Day & Hunter Ltd v 31(1)(a)(iv) (literary, dramatic and Twentieth Century Fox Corporation musical works) and 31(1)(b)(iii) [1940] AC 112. (artistic works); para 86(d) (44) Green v Broadcasting Corporation (cinematograph films); there is no of New Zealand (1983) 2 IPR 19 diffusion right in the case of sound (45) Exxon Corporation v Exxon recordings). Insurance Consultants International (62) Copyright Act 1968, Subsection Ltd [1982] RPC 69. 10(1)) (46) Kalamazoo (Australia) Pty Ltd v (63) Copyright Act 1968, Subsection Compact Business Systems Pty Ltd 26(1)) (1985) 5 IPR 213 at 232, per (64) Copyright Act 1968, subpara Thomas J 31(1)(a)(iii) (literary, dramatic and (47) Francis Day & Hunter Ltd v 20th musical works), paras 85(b) (sound Century Fox Corporation Ltd [1940] recordings) and 86(b) AC 112, 123, per Lord Wright for (cinematograph films); there is no the Privy Council performance right in respect of (48) Ibid, 88 per Stephenson LJ quoting artistic works the observations of Davey LJ in (65) Copyright Act 1968, subpara Hollinrake v Truswell (1894) 3 Ch D 31(1)(a)(i) (literary, dramatic and 420 musical works); subpara 31(1)(b)(i) (artistic works); para 85(a) (sound recordings); para 86(a) (66) Copyright Act 1968, Subsection (cinematograph films) 27(4) (67) (1884) 13 QBD 843 (68) See, for example, Jennings v (76) US Copyright Act 1976, Section 101 Stephens [1936] Ch 469; Ernest (77) Copyright Act 1970 (Law No 48 of Turner Electrical Instruments Ltd v 1970, article 12 Performing Right Society Ltd [1943] (78) See now article 2(1)(xter) Ch 167; Australian Performing Right (79) Ibid, article 12bis(1) Association v Canterbury- (80) See WIPO Preparatory Document, Bankstown Leagues Club Ltd Ref. 74, 81 [1964–1965] NSWLR 138; (81) (1991) 22 IPR 129 Performing Right Society Ltd v (82) Copyright Act 1976, Section 105 Rangers Football Club Supporters (83) Swedish Law No 729 of 30 Club [1975] RPC 626. December 1960 on Copyright in (69) Rank Film Productions Ltd v Dodds Literary and Artistic Works, Article (1983) 2 IPR 113 49; Danish Act No 158 on Copyright (70) Australasian Performing Right in Literary and Artistic Works, of 31 Association Ltd v Commonwealth May 1961, Article 49; Norwegian Bank of Australia (1993) 25 IPR 157 Act relating to Property Rights in (71) Bradbury Agnew & Co v Day (1916) Literary, Scientific or Artistic Works, 32 TLR 349 No 2 of 12 May 1961, Article 43; (72) Copyright Law Review Committee, Finnish Law No 404 relating to Draft Report on Computer Software Copyright in Literary and Artistic Protection, June 1993, p. 17 Works of 8 July 1961, Article 49 (73) Copyright (International Protection) (84) (1988) Copyright 42 at 82 Regulations 1969, Reg 4 (Principles PW 17(2), 18 and 19)) (74) WIPO Preparatory Document for (85) Ibid, Principle PW20 the meeting on “The Printed Word”, (86) Ibid, 83 7–11 December 1987, Geneva, in (87) Ibid, 84 [1988] Copyright 42 at 81 (88) Commission of the European Ricketson, S. (1987) The Berne Communities, (1988) Green Paper Convention for the Protection of on Copyright and the Challenge of Literary and Artistic Works: 1886– Technology—Copyright Issues 1986, Centre for Commercial Law requiring Immediate Action, Studies, London, p. 298 Communication from the (75) Copyright Law of 9 September Commission, Brussels, p. 216 1965, article 4. Also Dietz, A. in M. (89) Draft Report on Computer Software Nimmer and P. Geller (Eds.), Protection, June 1993, pp. 16– 17 International Copyright Law and (90) Ricketson, S. (1984) The Law of Practice, Matthew Bender, New Intellectual Property, Law Book Co., York, FRG-22 Sydney, Ch. 42–46 Copyright, Food Industry and Food Safety Considerations

Non-Nutrient Databases for Foods

Kimmo Louekari

Institute of Occupational Health, Topeliuksenkatu 41 a A, SF-0250 Helsinki, Finland

Non-nutrient databases for foods contain analytical data on toxic elements, pesticides, additives, mycotoxins, allergens or natural components which have no nutritional function. Non-nutrient databases can be used to monitor the trend of potentially harmful components in food; to estimate dietary intakes, provided that another file containing food consumption data can be linked with non-nutrient concentratins in foods; to clarify possible interactions between nutrients and non-nutrients; and to assess and measure exposure and risk in population groups. Examples of non-nutrient databases are discussed.

he reason for including a certain food component in a non-nutrient database is (i) that it can cause allergic or other adverse reaction; (ii) that it is potentially toxic to Tthe consumer in excessive amounts or in long-term exposure (e.g. heavy metals or pesticide residues); or (iii) that it may interact with nutrients (e.g. decrease of calcium absorption caused by phytates) (Table I). In practice, it seems reasonable to establish and maintain a non-nutrient database only, when the necessary research capacity for updating the database contents is available and when there is the scientific and/or regulatory interest to use the data. Some categories of non-nutrients presented in Table I are not necessary included in any existing non-nutrient database (e.g. natural toxic substances and pyrolysis products), since data on their concentrations in the relevant food items are not yet available. Table I. Categories of non-nutrients in foods Category (Examples) Nutritional or harmful effects User needs Toxic elements and Accumulation in the body, Intake monitoring, survey of contaminants (Pb, Cd, Hg, Al, chronic effects trends PCBs, dioxins) Pesticides (organochlorine and As above As above organophosphate pesticides) Food additives (nitrates, nitrites, Toxic effects, hypersensitivity Diet planning and the above BHA, BHT, colours) reactions needs Allergens Intolerance reactions, allergies Labeling, usage restrictions, consumer information Nutrient inhibitors (phytates, Decreased bioavailability of Analysis of interactions and tannins) nutrients effects on nutrient status Naturally occurring toxicants Acute and chronic toxicity Risk and intake assessment (patulin, aflatoxin, flavonoids) Pyrolysis products (quinoline Suspected carcinogenicity Exposure estimation, process and indole compounds) optimisation, consumer information

A non-nutrient database should non-nutrient databases to enable the contain average concentrations, ranges assessment of reliability and and tolerances to enable the calculation comparability of data. The detection limit of dietary intakes and comparison with of the analytical method may be much regulations. The range of the observed higher in older studies and this affects non-nutrient concentration is useful the comparability of old and recent when estimating maximum intakes. Data concentrations and estimated intakes. on preparation of the sample are also The code of the food item can be useful necessary, since different types of when a non-nutrient database is handling and preparation (washing, combined with other databases on food peeling, cooking, canning, frying etc.) constituents of food consumption. The can decrease concentration of non- relation code links a record in the nutrients, especially those attached to calculation database (enabling the surface of vegetables and fruits numerical analysis) to the reference (toxic metals and pesticides) and those database which is necessary as a which are decomposed by heat documentation of data sources and (mycotoxins). Description of the food evaluation of data. Often one record in sampling should be included to allow the reference database can be linked evaluation of how well the results with several records in the calculation represent different products, agricultural database, which is convenient in terms areas and seasons. Analytical method, of database maintenance (Table II). year of analysis, and quality control of Unlike nutrients, local conditions, the analysis should also be presented in especially sources of food contaminants, often cause variation in nutrient intake are most likely found in non-nutrient content of foods. This has certain contaminated areas, food to be taken into account in the sampling surveillance and dietary intake studies protocols for food surveillance. In many are often directed accordingly. cases, the maximum non-nutrient Examples are areas where fish is contents and the maximum exposure contaminated by the pulp and paper are of special interest. In many cases, industry (Louekari, Verta, & Mukherjee, blood levels or other biomarkers of unpublished data) and areas where exposure and health consequences are crops receive the fallout from mines or monitored in contaminated areas. Since refineries containing cadmium or lead. health consequences of high non-

Table II. Contents of two related records in a non-nutrient database The calculation database Contents Example Name of the food item Liver, cattle Code of the food item 0102 Preparation status Raw Non-nutrient Cadmium Average concentration 70 Unit µg/kg, wet weight Number of samples 60 The range of 95 percentile of the analytical results 350 The maximum permitted concentrations Not given The relation code to the reference database 0502 The reference database (record 0502) Contents Example Publication Salmi A, Hirn J. The cadmium and selenium contents of muscle, liver and kidney from cattle and swine, Fleischwirtschaft 64:1984:464–465 Analytical method AAS, graphite furnace The analytical quality control NBS ref. material 1577 analyzed Description of sampling 10 slaughterhouses representing the whole country, variation caused by the age of animals was considered Detection of limit of analysis Not given, sensitivity was 0.005µg/20µl Year of sampling 1981 Year of analysis Not given

The food nomenclature used in a Ŷ Examples of Non-nutrient non-nutrient database should be Databases designed to cover and differentiate those foods which contain significant Non-nutrient databases are maintained amounts of the non-nutrients of interest. for example in USA, Netherlands, Vegetables and fruits may contain Denmark and Finland (1, 2, 3, 4, 5). pesticide residues; processed foods These databases contain data on toxic may contain food additives; mushrooms, metals, selected pesticides and in some liver and predator fish may contain high countries, also on mycotoxins, concentrations of toxic metals. radionucleides or organochlorine compounds. Data are often generated by a governmental agency, e.g. the FDA in the United States, and the National relevance of items of information for the Food Agency in Denmark as part of a non-nutrient in question. food surveillance program. Studies in Ŷ Experiences of the Estimation of academic institutes can be incorporated Dietary Intake and in the Analysis of into the database. A description of some Trends existing non-nutrient databases and surveillance/monitoring programs has Dietary Intake of Cadmium — Are the been presented elsewhere (6). Available Data Comparable? In Finland, the steering group of the recently established non-nutrient It is often important to know the trend of database consists of representatives non-nutrient concentration and/or from relevant authorities, research dietary intake. However, the institutes and the food industry. The interpretation of the available data is Finnish database includes summaries of difficult if changes in analytical the origin, toxicity and regulatory status techniques and methods of dietary of non-nutrients, in addition to intake estimation are not taken into concentration of non-nutrients. The first account. For example, it has been version of the database contains data suggested that a decreasing trend in on mercury, cadmium, lead, dioxins, cadmium intake in European countries PCBs and poly- aromatic hydrocarbons, could be seen (7). However, the residues of antibiotics in milk products estimates of intake used by van Assche and hormones in tissue samples of are not comparable since studies have animals. Also microbial contamination been made using different methods and by Salmonella, Listeria and Yersinia is since up to the late 1970s, analytical covered in the database. techniques for determination of With this information, the maintainer cadmium were non-sensitive and not of the database, the National Food accurate at the level observed in foods. Administration, can advise consumers, Furthermore the total diet method respond to the media about food safety utilized earlier tends to result in questions and in future, produce overestimates (8). Using non- certificates for food processors. It is comparable data, no conclusions can be planned that in the future the system drawn about trends of non-nutrient would also enable estimation of dietary concentration. A non-nutrient database intakes using the collected analytical with contents presented in Table II data. enables a decision about whether there The presentation of the data in a are enough data to observe the trends record of the Finnish non-nutrient of non-nutrient concentrations or dietary database is similar to that in Table II, intake. with a few exceptions. Trend of Total Mercury Concentration Preparation/processing status of the in Fish and Intake in Different Socio- samples is not presented in the economic Groups database, maximum permitted In a recent study, we have analyzed the concentrations are in a separate file and intake of mercury using the analytical basic data on sampling is given in the data on these contaminants and the calculation database but not the Finnish Household Survey for food reference database. It was found in the consumption data (of Louekari, Verta, & updating of the database that a general Mukherjee, unpublished data). Some way of presentation could not be strictly results of that and another study (9) are followed but had to be adapted presented to illustrate how a non- according to the availability and nutrient database can be used for estimation of non-nutrient intakes in analyzed, and had to be estimated. We different groups of populations or in used the average concentration of those particular risk groups, and for analysis of fish species, which were not covered in different food groups to the total intake. the food consumption study by name In Finland, most of the intake of but had been analyzed for Hg mercury is caused by eating fish, which concentration. These calculations were accumulate methylmercury of the food recorded in the same file as other data chain of the lakes. In the study on on Hg concentration. Second, in the mercury intake, changes in Hg intake of food consumption file some fish (e.g. people of different socio-economic herring) is presented as smoked, frozen, groups during the period of 1967–1990 fresh, salted. Because the Hg were observed in the polluted areas, concentration is not affected by which include 10–15 per cent lakes and processing, consumption of these coastal waters in Finland. In the study different types of herring were added up areas, Hg is discharged mainly by the and multiplied by the Hg concentration pulp and paper industry and by of herring to obtain the total Hg intake chloroalkali plants. Our results show that from herring. the average pike Hg concentration The Average and Maximum Intake of decreased from 1.52 mg per kg (in PCBs 1967– 68) to 0.60 mg per kg (in 1990– 91). A 1993 study of the intake of The dietary intake of Hg among polychlorinated biphenyls (PCB) from farmers and white-collar workers living the Finnish diet was based on in the study areas was estimated by measurement of seven PCB indicator combining the data on Hg congeners in 99 food samples and the concentrations in food with the data on Household Survey data. The average food consumption during the period of PCB intake from food was 15 mg/day or 1967–1990. It was observed that total 0.25 mg/kg of body weight per day, and dietary Hg intake for farmers was 22 showed no change as compared with and 15 mg per day in the years 1967 estimates from late 1980s. and 1990, respectively. On the other Approximately half of the intake came hand, the total dietary Hg intake for from fish. Cheese, fats and oils white-collar workers was 13 and 8 mg contributed significantly to the intake, per day in the years 1967 and 1990, since their consumption is relatively respectively. The study suggests that high, although the concentration is not although the fish consumption of the at the level observed in fish (up to 2100 Finnish population (except for farmers) mg/kg in pike) (9). has increased slightly, the intake of Hg We also estimated the theoretical has decreased remarkably (by 39 per maximum intake of PCB of an average cent on average). This is due to the consumer assuming that the consumer rapid decline of aquatic Hg discharge eats food containing the maximum especially from the pulp and paper observed concentration of PCB. This industry. The other reason is that theoretical maximum was 41 mg/day consumers prefer fish species, e.g. which is almost three times the average rainbow trout, which contain much less intake. This theoretical maximum is Hg than pike and perch. probably not too far from the actual In compilation of the mercury maximum intake, since in this database, the following problems were calculation the food consumption was at faced: First the concentration of “other the average level. Furthermore, the fresh fish” and “processed fish” were not maximum value in the database is dependent on the number of samples (n Danish Diet, Statens was between 3 and 20 in this case). The Levnedsmiddelinstitut, Copenhagen maximum observed concentration would (5) Southgate, D.A.T., & Walker, A.D. be higher if the number of samples were (1992) Report of FLAIR Eurofoods- greater. People having a high fish Enfant Project 2nd Annual Meeting, consumption and living in contaminated Wageningen Agricultural University, areas, consume the actual maximum Wageningen, pp. 40–43 dietary intake of PCB (9). (6) Louekari, K., & Salminen, S. (1991) Trends Food Sci. Technol 2, 289– Ŷ References 292 (1) Pennington, J. A. T. (1983) J. Am. (7) Van Assche, F., & Ciarletta, P. Diet. Assoc. 82, 166–173 (1992) Cadmium '92, Cadmium (2) State Supervisory Public Health Association, London, pp. 51–54 Service (1983) Surveillance (8) Louekari, K., Jolkkonen. L., & Varo, Program “Man and Nutrition”. P. (1988) Food Add. Contam. 5, Report No. 10, Hague 111–117 (3) State Supervisory Public Health (9) Himberg, K., Hallikainen, A., & Service (1987) Surveillance Louekari, K. (1993) Zeit. Program “Man and Nutrition”, Lebensmitteluntersuch. -Forsch. Hague 196, 1–5 (4) Andersen, A. (1981) Lead, Cadmium, Copper and Zinc in the Copyright, Food Industry and Food Safety Considerations

Food Composition Databases in the Food Industry

Olivier de Rham

Nestlé Research Centre, Manufacturing, Milk and Nutrition Business Unit, Avenue Nestlé 55, 1800 — Vevey, Switzerland

This paper covers specific aspects of the use and production of food composition data in the food industry. Two types of information are stored by manufacturers: quantitative information on an extensive list of nutrient and non-nutrient components (including contaminants, stored separately), and qualitative information about suitability for diets to reduce or eliminate components. The aim of collecting such information is usually to check product compliance with regulations or internal specifications and norms, with a special emphasis on nutritional labeling. Control of costs, processing, storage stability, taste and texture constraints are important goals along with provision of product information to consumers. The origin of the information is usually analytical, calculated or borrowed from published food composition tables, depending on availability, regulatory requirements, product type, and cost of analysis. The meaning of the information is different depending on whether natural products are processed without intentional modification of composition, or whether international alteration of composition occurs, thus engaging the responsibility of the processor to guarantee the declared level. However, considering the natural variability in food composition, the 20 per cent tolerance margin usually accepted tends to discourage the declaration of the actual average value when liability is engaged. A clear understanding of these constraints is important to ensure a fruitful collaboration between industry and other interested parties in the use of food composition data.

he problem of an unequivocal description of a food in a table is a familiar one. The translation of food names needs not only a very good familiarity with the relevant Tlanguages, but also an intimate experience of the cultures and their local variants. subclasses, and internally the use of Ŷ Foods and Food Classes codes. Food Names and Descriptions Food Classification The situation is also difficult with The interest to widen the classical list of commercial product brand names, as food classes (and subclasses if needed) they often have no meaning in is multiple. They help to select an themselves. To complicate the situation ingredient for a given aim from a further, the same brand name can be suitably selected screen listing of the used for different products in various corresponding group(s) of ingredients. parts of the world, or conversely the They allow the origin of a nutrient in a same food product may have different product or in a diet to be traced. They brand names. A detailed description of help in checking the coherence and the food in these cases is necessary, as validity of food composition data. well as the use of the food classes and

Table I. List of food ingredient classes Animal Plant Refined products Meat Cereals Oils, fats, shortenings Fish and seafood Vegetables (leaves, stalks, Proteins, hydrolysates, amino flowers, fruits, bulbs, sprouts) acids Egg Roots Starches Milk and milk products Mushrooms Sugars and sweeteners Algae Dietary fiber concentrates Oilseeds Enzymes Herbs and spices Microbiological starters Tubers (potatoes) Acids, alkalis, salts Pulses (dry beans, grain legumes)Vitamins Fruits Thickeners, emulsifiers, antioxidants, preservatives Berries Processing aids Nuts Colours, aromas and flavours Cocoa Water Coffee and surrogates Tea and herbal teas Wine, spirits, vinegar Yeast and yeast products

Whether a food can be attributed in a could easily be positive in computerized table to more than one class of products databases with the necessary is an important question. The answer is programming precautions. almost always negative in the printed Ŷ Ingredients and Ingredient Classes version of food composition tables as this would mean a waste of space, but Besides the traditional food ingredient classes found in published food composition tables, food manufacturers sauces, desserts, infant foods and use additional raw materials for product clinical nutrition products (see Table II). manufacture: Ŷ Quantitative Information ƒindustrial “raw” materials, often already processed to some extent, Components Contents e.g. peeled, cut, pureed, deboned, Quantitative data on the contents of a cooked, dehydrated, freeze-dried great number of components including ƒfood ingredients not consumed as both nutrients and non-nutrients can be food in themselves and not usually required by technologists or marketing accessible to consumers, specialists. The interest for each constituting additional raw material individual component is very variable classes. These include refined food according to the type of product components (e.g. fiber concentrates, concerned and the aims of the demand. starches, proteins, hydrolysates, Most requests do not ask for more than amino acids, medium chain the proximate composition (“big 4”: triglycerides, hydrogenated fats, energy, protein, fat, carbohydrates) or trans-esterified fats), pure chemicals slightly more (“big 8”: the “big 4” plus (acids, alkalis, salts), enzymes, dietary fiber, sodium, cholesterol, microbiological starters, vitamins, saturated fat), and possibly one or two additives (antioxidants, emulsifiers, vitamins and minerals, but there may be thickeners, preservatives), colors, additional requests. aromas and flavors (see Table I). Dairy product specialists insist on milk Out of these materials, a large number fat and non-fat milk solids. Chocolate of intermediate products, subrecipes or manufacturers have special premixes are produced, that can be requirements related to cocoa solids either transient in processing, stored for and fruit content. Yoghurt producers are later use, or purchased as such. These interested in the transformation of subrecipes or premixes also constitute lactose to lactic acid. Broth and stock an “ingredient” class. producers ask for glutamate figures, meat and sausage manufacturers need Ŷ Products and Product Classes carnitine values, coffee and cocoa Manufactured products may be beverage processors want data about classified according to consumption methyl xanthines. In addition, there are characteristics that are familiar to the national differences due to local consumer, rather than on the basis of legislation. the major ingredient which is sometimes Dietetic product specialists are most very difficult to define. Minimally demanding, including not only a list of processed foods (with only sugar, salt, vitamins and minerals, but also fat or water added or removed) can be occasionally figures for fat and sugar classified within the class of their major substitutes with reduced energy value. ingredient. However, manufactured Breast milk substitutes and clinical product classes go much further than nutrition products require a long list of that: beverages, breakfast items, vitamins, minerals and trace elements, starters, snacks, main dishes, and some amino acids and fatty acids. garnishes, fast foods, side dishes, The figures for fatty acid classes are being requested more and more often. Table II. List of food product classes Simple productsa Dairy Evaporated milk, milk powder, cheese, yoghurt Meat Preserved meats, sausages Fish Frozen fish Cereals Bread, pasta, bakery products Vegetables Frozen prepared vegetables Tubers Potato flakes, frozen croquettes, chips, crisps Beverages Alcoholic Wine, beer, cider, drinks, punch, cocktail, spirits, liquors Soft Fruit juices, vegetable juices, dairy beverages, milk modifiers Cereal drinks Coffees, teas, water, mineral water, sodas Breakfast foods Butter, margarine, jam, spreads, cheese Bacon and sausages, eggs, milk, yoghurt, fruits, fruit juices Starters Soups, broth, stock, “hors d'oeuvres” Snacks Savory, salted Appetizers, cocktail foods, chips, salted peanuts, crepes Sweet Ice-creams, chocolates, cereal bars Pastry, confectionery, sweets, biscuits Yoghurts, fruits Main dishes With meat or fish Without meat or fish Garnishes Potatoes, cereals, pasta, bread, doughs Vegetables, legumes Fast foods Hot Hamburgers, pizzas, quiches Cold Sandwiches, savoury pies Side dishes Salads, cheese, bread, spreads Sauces Salad dressing, mayonnaise Meat sauces Seasonings, condiments, spices Desserts Fruits, dairy desserts, custard Cakes, tarts, sweet pies, fried desserts Infant foods Starter milks, follow-up milks Infant cereals, baby foods Clinical nutrition products Enteral Parenteral a Simply processed, with only salt, fat, sugar or water added

At Nestlé, the INFOODS tags (1) as with the correct safety margins, whereas such are not used, but instead a similar contamination levels vary greatly from system of conventional abbreviations batch to batch and cannot predict the and standard units, to avoid confusion values for future crops. The data should and errors arising during data transfer. be stored with batch identification in a Contaminants separate database that does not allow the usual averaging done in food Data on contaminants contents is composition data. This is so, for quantitative, and it is tempting to put it in example, for aluminum levels in the the same database as the other ingredients for infant formulas. components. But food composition data should have a predictive value, i.e. be Ŷ Qualitative Information valid in the future as well as in the past Elimination Diet Suitability Elimination diets are sought for various free lists” can therefore only be obtained reasons, either medical, religious or from the recipes database, and this philosophical. For example, gliadin, milk information does not belong to the food protein, soy, must be eliminated from composition databse. the diet of people allergic to these foods, A frequent confusion ignores the fact meat from the diet of vegetarians, that some additives are at the same animal products from that of vegans, time natural molecules. The absence of pork from kosher or halal products. This such an additive in the recipe does not information is qualitative (present or imply the absence of that molecule in absent) rather than quantitative the product. Such a molecule is then a (numeric). normal food component to be handled Such qualitative criteria are delicate to quantitatively. For example, claiming handle. An error (which could have very “no-phosphate added evaporated milk” serious consequences making or “no glutamate-added tomato ketchup” somebody very ill) is easily made, suggests that these products are, whereas an occasional low figure for, respectively, phosphate-free and say, protein on the label of the same glutamate- free, which is incorrect. product is harmless. Furthermore, at the same time such The decision whether an ingredient or a claims reinforce and perpetuate product is acceptable or not in a given erroneous beliefs, ideas and fears about elimination diet depends on a number of additives among consumers. complex rules, especially for religious Ŷ Quality of the Information diets. In addition, in order to reach a decision on the suitability of a product all Aims of Information its ingredients must be correctly identified for the given criterion, the Composition data in the food industry recipe must be strictly followed, and the are usually produced to satisfy quality processing (contamination, recycling, assurance aims: compliance with incompatibilities, processing equipment regulatory or scientific composition and conditions, processing-induced standards, and with internal norms and components) must be properly specifications. In recent years, a great addressed. deal of this activity has been geared to nutritional labeling as new regulations Qualitative information is not suitable for appeared in the US (2) and in the EEC components like sodium, cholesterol, (3) with compliance dates in 1994, and lactose, sugar, phosphate, caffeine. In this will continue in the near future. these cases, the aim is a reduction diet In addition, it is often necessary to with a maximum acceptable daily intake control one or the other parameter of and not an elimination diet. Such food composition to ensure optimal parameters are quantitative factors and process control, taste, texture, flavor, must be handled as such. convenience, price, shelf life. Only in a few cases does the publication of food Ingredients and Additives Usage composition information brochures and The recipes database gives quantitative the transmission of data to dietitians or information on the ingredients used in a database compilers receive priority. product. This information is usually Origin of the Information passed to the public in a simplified form of the qualitative list of ingredients. The Food composition data are usually additives are in this respect no different analytical (hopefully on multiple from the other ingredients. “Additives- representative samples), calculated from the recipe, or borrowed from an external database. Whether one or the less standardized products. The other source is to be preferred or legislation usually allow for a 20 per cent rejected depends on the local tolerance. This is by far not large constraints which may include, e.g. enough for minerals and vitamins, and regulations, availability, facilities, costs, the declared values are therefore often and needs for accuracy. below the real average content, to allow Variability a safety margin. This is unfortunate for dietitians, but largely unavoidable. The biological nature of the materials used (e.g. ripeness, variety, soils, Meaning of the Information fertilizer use, rearing and feeding Dietetic products are sold on the basis practices, post-harvest practices) of medical advice, and their nutritional introduces a large variability in the value is of prime importance. The composition of ingredients. In addition, technologist is required to produce a there is a variability in recipes, including food that matches a predefined intentional small variations for various composition, and this is assured by technological, taste, price or regulatory standardization of every single batch of reasons and non-intentional ones for production. Action is taken to guarantee practical reasons. the required levels that are declared on The cumulated variability of food the label, and the producer is composition is mastered to a certain responsible for ensuring that the extent by technologists, but can only declared figures are adhered to under increase as international comparisons normal circumstances. This is also the are made. This raises the question of case when a component level is when two foods are close enough to be controlled (addition or reduction), but considered one and the same in the needs “overages” or safety margins in database. When the need for accuracy the declared values. is low, such cumulations are easily Normal, everyday food products on the accepted, whereas every batch must be other hand do not require such accuracy recorded separately when high accuracy of composition data. In this case, the is needed. It is up to the user to decide processor discloses what nature has the degree of precision that is needed produced, as influenced by processing. and to handle the database accordingly. A yearly production average is in this The food industry composition data case the most significant nutritional compiler enjoys some advantages over figure. One half of the production will other database compilers. Purchase have a content lower than the average specifications are used for a defined list and the other half a higher one. In order of raw materials. Fixed, precise and to ensure that the vast majority of controlled recipes and processes are packages correspond to the declaration used. Manufactured products are (including possible losses during controlled for various reasons (e.g. storage), the declaration might be below legal, standards of identity, cost, storage the actual average value for those stability, taste and texture), which perceived as healthy nutrients, while for reduce the variability of the final those nutrients negatively perceived, composition. such as sodium, the declaration might In full cream milk powders we measured be higher than the actual average value. a relative standard deviation of 5 to 10 Ŷ Conclusions per cent for minerals, and 15 to 25 per cent for vitamins. It is probably greater Food composition information that exists for trace elements as well as in other in the food industry is not primarily aimed at dietitians, nutritionists or this context that a collaboration is database managers, and some possible and fruitful. frustrations are predictable. The cost of Ŷ References producing information for external use as well as the cost of its dissemination (1) Klensin, J.C., Feskanich, D., Lin, V., are high and may be restrictive. The risk Truswell, A.S., & Southgate, D.A.T. of misunderstanding and misuse of this (1989) Identification of Food information is important to highlight, Components for INFOODS Data especially if its currency is not assured Interchange, UNU Press, Tokyo. by regular updates. At the same time, (2) Mermelstein, M.H. (1994) Food the protection of that part of the Technol. 48, 62–71 information which is proprietary is a (3) EEC (1990) Council Directive of necessity. 24th Sept. 1990 or Nutritional Understanding the constraints and the Labeling of Foodstuffs 90/496/EEC. reasons for producing such information Official Journal of the European will foster a common understanding of Community L276/40–44 of its meaning and limits, and will facilitate 6/10/1990. the dialogue and improve its use. It is in Copyright, Food Industry and Food Safety Considerations

The Databases of the Australian National Food Authority

Janine Lewis, Simon Brooke-Taylor, Fay Stenhouse

National Food Authority, PO Box 7186, Canberra, ACT 2610, Australia

The National Food Authority was established in August 1991 as a reform to the food standards setting system. Although development and variation of food standards is its primary role, the Authority is also responsible for the national references on nutrient composition and for food safety surveillance. Three database providing the supporting information systems for these activities are described. The food composition work of the Authority (and the federal Department of Health before it), has focused on revision of the national reference on nutrient composition and the release of the data in a number offormats. Nutrient data are produced mainly from an ongoing food analytical program and managed by means of a computer system, the Australian Nutrient Data Bank. These data enable the Authority to estimate the levels of nutrients in food, and the probable nutritional impact on foods of changes to the compositional and technological aspects of food standards. The published food composition tables also provide a valuable data source for industry when formulating nutrition information for labeling of foods. The Authority also has responsibility for the Market Basket Survey which identifies whether pesticides and contaminants are at levels which pose health risks to consumers. The last completed survey in 1990 concluded that Australian intakes of pesticides and contaminants were well below international limits recommended by the World Health Organization. To facilitate a uniform interpretation, implementation and enforcement of the Food Standards Code, the Authority is developing a national food safety information database, which will link all agencies involved and provide an information network to enable effective use of resources for a rapid national response to public health emergencies related to food.

he National Food Authority (NFA) was established in August 1991 with the proclamation of the Commonwealth National Food Authority Act 1991 (the Act), as Tan independent and expert body with responsibility for the development, variation and review of Australia's food standards. revisions to the food composition tables, Ŷ Functions of the National Food Authority which were compilations of data from a number of sources mainly from The Act specifies a list of 13 statutory overseas but including some Australian functions that the NFA is required to data. During the 1970s, there was a perform. More than half of these growing awareness and call for Australia functions do not relate directly to the to have a more comprehensive and Authority's primary role of developing contemporary set of food composition and revising food standards, but focus tables that reflected the Australian food on food safety research and education, supply (1). and coordination of food recalls and The first step towards this goal was food safety information. In summary, the achieved in 1978 when it was decided to functions are: revise Australia's nutrient composition ƒdevelopment and review of food data completely by establishing a standards program to analyze the Australian food supply progressively. At that time, a co-ordination of food surveillance ƒ Working Party of Australia's peak health ƒresearch and surveys advisory body, the National Health and ƒfood safety education Medical Research Council, was formed ƒco-ordination of food recalls to devise a plan for the collection of food composition data including the priorities ƒdevelopment of assessment policies for analysis of foods and nutrients, and on imported food to review the format for publication. ƒadvice to the federal minister Their recommendations included ƒdevelopment of codes of practice for extending the range of nutrients and industry foods, particularly for take-away and ethnic foods, and for publication of the ƒincidental functions. tables to be in a loose-leaf format to Three of the functions have or will enable easy incorporation of updates result in ongoing activities that are (2). dependent on the databases of the In establishing the program, it was Authority. They are: decided that within the available ƒresearching the nutrient composition funding, the data should be produced to of the Australian food supply serve nutrition and health goals before ƒmonitoring the pesticide and aiming to meet agricultural contaminant levels in the food requirements. Thus, the overall direction supply of the revision program was generally to analyze one composite sample of food ƒestablishing a system of food safety in preference to multiple samples, to surveillance. collect data for primary produce prior to Ŷ Nutrient Composition Database manufactured food products and to sample at the retail level on a regional The responsibility for government- rather than a national basis. After 13 sponsored nutrition composition years of the laboratory program, this activities was transferred to the National approach has enabled Australian Food Authority upon its formation. nutrient data to be produced on a However, these activities have been broader range of foods than would conducted in Australia since the 1930s otherwise have been possible given the under the auspices of the federal funding constraints and the demand for Department of Health. Up to the 1970s, the Department issued three major comprehensive data in as short a time as possible.

Table I. Present range of constituents analyzed Proximate constituents Minerals Vitamins Moisture Sodium Retinola Total nitrogen Potassium Į-carotenea Total fat Calcium ȕ-carotenea Individual monosaccharides Magnesium ȕ-cryptoxanthina Individual disaccharides Iron Thiamin Starch Zinc Riboflavin Ash Phosphorus Niacin Dietary fibera Copper Vitamin Ca Cholesterola Manganese Į-tocopherol a Fatty acids Chloride Vitamin B6 a a Organic acids Fluoride Vitamin B12 Selenium Biotin Sulphur Pantothenic acid a Analysis of these nutrients depends on the type of food

The initial laboratory work was For the first 11 years, the program undertaken at universities in Sydney. focused on obtaining a broad nutrient The majority of the work during the first profile of foods commonly available in five years of the laboratory program was the Australian food supply. By the end of conducted by the Wills and Greenfield this period, over 2500 foods had been team at the University of New South analyzed for their nutrient composition. Wales. Since 1985 however, the As opportunities arose and expertise analyses have been conducted by the became available, the range of analyzed South Australian division of the nutrients gradually expanded from six to Australian Government Analytical ten vitamins, and from six to 13 Laboratories in Adelaide. minerals. The range of nutrients which From the commencement of the now comprises the routinely program, routine analysis of foods commissioned constituents is given in included: Table I. ƒmoisture; total fat; total nitrogen; Because of the expansionary carbohydrate components — approach described above, there were monosaccharides, disaccharides some gaps in the vitamin and mineral and starch; ash; data for foods analyzed during the early part of the program, principally for ƒsodium; potassium; calcium; primary produce and take-away foods. magnesium; iron; zinc In 1992, the laboratory program ƒretinol, Į- and ȕ-carotenes, ȕ- began to resample foods to analyze cryptoxanthin; thiamin; riboflavin; those nutrients which were not included niacin; vitamin C. in the first round of analyses. The focus Other components were analyzed of the program from now on will be to appropriate to the foods, such as build up a more comprehensive profile cholesterol, fatty acids and organic on all foods previously analyzed. The acids. Special programs were also sampling scheme for these analyses is conducted for amino acids and dietary being maintained as close as possible to fiber analyzed by the Englyst method the original specifications. The “catchup” (3). program also provides opportunities for other nutrients with previously doubtful factors for recipe foods. With the values to be reanalyzed. In addition to transfer of responsibility for the food the “catchup” analyses, there will be an composition program to the Authority, ongoing component which will analyze the databank has been relocated from a new foods entering the market. mainframe to the Authority's smaller With the continuation of the food computer system. It is anticipated that analysis program, there is now the there will be future opportunities for opportunity for more detailed information developments to be made to this to be collected. This is considered to be system. From a compiler's viewpoint, particularly useful for staple foods. One the opportunity to improve the system example of this approach is a program regularly is most important. It enables currently in progress to collect and refinements to be made and new analyze composite samples of three of features to be added as well as taking the most commonly consumed types of advantage of improvements in software bread from each of the eight Australian and hardware technology to enhance States and Territories. These results will efficiency and functionality. further extend the information base by The results of the laboratory program enabling calculation of descriptive are now published in a number of statistics such as standard error of the formats. Australia's national nutrition mean. This project is being conducted in reference is a series of loose-leaf collaboration with the Bread Research volumes entitled Composition of Foods, Institute of Australia. Australia (4). There are now six volumes Over the life of the program, the food in the series providing data on some industry and other organizations have 1430 foods. Volume 7 is presently in participated with government in building preparation and will add a further 100 up the store of nutrient composition data foods, mainly restaurant dishes by contributing their own data, or originating from other countries. funding specific programs that analyze Additional formats are available to their products according to agreed meet the needs of a range of data specifications. Successful cooperation users. A summary database NUTTAB has occurred with industry organizations for use in software applications has such as the Australian Meat and been produced for public use since Livestock Corporation, the Australian 1989, although a predecessor of this Dairy Corporation and the Bread database was used to analyze the Research Institute of Australia. national dietary surveys conducted in An integral component of the revision the 1980s. The currently available process was the development of a version NUTTAB91–92 (5) is the third supporting computer system for data publicly available update to the storage, processing and reporting. The database and provides data on 28 Australian Nutrient Data Bank (ANDB) nutrients and energy for some 1580 was a mainframe application developed foods. A revised version is planned for from a pilot system in 1987. A series of release after publication of Volume 7 of enhancements were made to the data the food tables series. The additional bank in 1990. One of these was to foods in this database are those from provide the capability to calculate foods British sources which have continued to from recipes using factors as estimates be included from the original dietary of the weight change on cooking and survey database. Two condensed nutrient retention. There is some scope versions of the tables also have been now for the Authority to undertake the released. Nutritional Values of supporting research into weight loss Australian Foods (6) was produced to meet the needs of educators and expression between the data tables and students in food, biology and health the Code requirements. For example, courses at upper secondary and tertiary the Code specifies that total levels while Food for Health (7) is a carbohydrate content should be simplified guide for the general public determined by difference rather than by which also includes general dietary analysis. The Australian tables report advice. The two latter publications have carbohydrate data by direct analysis. A proved to be among the most popular carbohydrate by difference value can be books released by the government readily calculated from those data tables publishing service. because a value for ash is provided The food composition tables are also routinely, together with the other used as a reference by the food industry proximate data that are required. for food labeling purposes. Part A of the The relevant food standard which Australian Food Standards Code (8) controls the claims for the vitamin and sets out the labeling regulations, mineral content of foods is currently including those for nutrition labeling, being revised. The term “average with which manufacturers must comply. quantity” previously defined in the When a label carries a nutrition claim, a general labeling provisions is proposed nutrition information panel must be to be used as the basis for determining displayed containing, as a minimum, the amount of claimed vitamin or information on the energy, protein, fat, mineral. The Australian database is and total carbohydrate, total sugars, sodium will continue to be a valuable source of and potassium content of that food. information for this purpose. Other nutrients which may be claimed Although provision of nutrient and listed in the panel include amino composition data is not the National acids, starch, cholesterol, fatty acids, Food Authority's primary role, there is a and dietary fiber. The information given profound appreciation of the important in the panel should represent the contribution made by these data to the average quantity of the listed improved knowledge of food and its components allowing for seasonal relationship to health and disease in variability and other known factors which Australia. The Authority is committed to could cause actual values to vary. the ongoing program of laboratory These average quantities can be analysis of Australian foods and the determined from three sources of publication of contemporary and information: comprehensive nutrient composition ƒthe manufacturer's analysis of the data for the benefit of health food professionals, educators, consumers, the food industry and the development ƒcalculation from the actual or of public health policy and programs. average quantity of nutrients in the ingredients used Ŷ Market Basket Survey Database ƒcalculation from generally accepted The Australian Market Basket Survey data. (AMBS) is conducted in Australia every The term' generally accepted data' is two years to estimate the levels of a very broad one, but now that there are pesticides and contaminants including Australian data on a significant range of heavy metals and natural toxins in the foods and ingredients, it is being Australian food supply. Responsibility increasingly used as a reference for for the AMBS was transferred to the labeling purposes. Manufacturers need Authority upon its establishment, from to be aware of the different modes of the federal Health Department. The survey comprises the collection biphenyls. Analyses were conducted and preparation to table-ready state of a also on specific compounds often found specified list of foods for analysis of in particular types of foods. For particular chemical residues. These example, seafood was analyzed for foods are selected to represent all major mercury, peanut paste for aflatoxins and groups of foods in the Australian food potatoes and potato crisps for solanine supply. Foods are collected over one content. The database is sufficiently year according to one of three standard flexible so that it can store such data protocols and prepared according to a when they are commissioned from time set of instructions. Core foods are to time. sampled once every season and in These results are used to estimate every capital city; regional foods are the total dietary intake of these sampled in every capital city but only in substances by six age/sex categories one season; and national foods, such as covering adults and children, and thus to corn flakes, which are not expected to assess the safety of the Australian diet vary between regions, are sampled in with respect to these substances. The only three capital cities in one season. intake data are derived by reference to The results are expressed on a per the consumption patterns given in the kilogram wet weight basis and entered most recent national dietary surveys. into the AMBS database. The number of These values in grams/day are entered foods sampled has been increasing on to a spreadsheet and an estimate of gradually over the years in which the the total dietary intake of each surveys were conducted. In the 1990 substance is then calculated for each survey, 53 foods were sampled, while age/sex category. Dietary intakes at the 62 foods were sampled in the 1992 95th percentile of energy are survey. determined. For the 1992 survey, The laboratory results for each intakes at the mean intake of energy sample are identified on the database were also determined. These results by the food name, its location and the and the summary data on the season of the year in which it was distribution of levels of pesticides and collected. The information from each contaminants found in the sampled survey is maintained in separate files on foods are provided in regularly the database. Currently, data from the published reports of each survey (9). 1985, 1986, 1987, 1990 and 1992 The estimated dietary intakes are also surveys are stored on the database. The compared to the internationally survey was run annually until 1988 accepted safe limits which are the when a review was held which Provisional Tolerable Weekly Intakes recommended that the survey be (PWTIs) for contaminants and conducted biennially. The maximum Acceptable Daily Intakes (ADIs) for range of the residue data for one food pesticides. from the 1992 survey is given by the The results of surveys over the last following example. Twenty-four samples 23 years have indicated where of grilled lamb chops were analyzed for additional vigilance is required in the residues of 57 compounds. They were food production chain as well as three heavy metals, 13 organochlorines estimating the overall levels of these plus three metabolites of substances. This important monitoring organochlorines, 22 organophosphates, function has shown to date that six synthetic pyrethroids, two fungicides Australia's food supply has generally and one class of fungicides, six safe levels of presticides and herbicides and polychlorinated contaminants and these levels have agencies. This will assist agencies been improving since monitoring began. plan activities and allocate resources, and facilitate rapid Ŷ Food Safety Information Database response to food safety concerns. It The coordination of food surveillance will also promote uniform information is a statutory function of the enforcement of standards by Authority. Currently there is no national permitting agencies immediate system to collate these types of data. access to up-to-date information on Most of the States and Territories investigations in other jurisdications however, are collecting and maintaining availability of printed information on information on the results of food ƒ food safety issues, plans, trends, inspection and monitoring activities and contacts within their jurisdictions. Unfortunately, the data are variable in scope and are ƒpotential savings in information stored in a number of formats, some of technology development and which are more technologically support. These are greatest for advanced than others. The types of those States supporting existing information currently collected include systems and for States and surveys of the microbiological and Territories which intend to develop contaminant levels in food, food their own food safety related complaints, recalls and prosecutions. computer systems. Soon after the formation of the It is expected that the development of Authority, the NFA Advisory Committee a national food safety information (NFAAC), comprising representatives of system, which by providing a unique all State and Territory Health information base of food recall, survey Departments, the federal Health and sampling data, will offer significant Department and the Australian benefits to regulatory agencies. Quarantine Inspection Service, was The Food Safety section of the formed. At its first meeting, the Authority is currently collating Committee expressed general support information on food surveillance for the concept of a national food procedures of the States and Territories surveillance database and agreed that a to enable the identification of relevant consultant should be employed to fields for the database and levels of undertake a feasibility study of the security required. It is anticipated that a project. trial will be undertaken shortly between The feasibility report points to the Authority and a State or Territory, to significant overall savings in resources gain experience in the transfer of food at the State and Territory level, as well surveillance data, prior to any large as the broader benefits of a shared and scale implementation. online system. It is proposed that a fully Ŷ References automated national system will include: (1) Greenfield, H., & Wills, R.B.H. ƒan information base to facilitate the (1981) Food Technol. Aust. 33, allocation of resources efficiently 101–130 and effectively by reducing (2) English, R. (1990) Food Aust. 42, unnecessary duplication in sampling S5–S7 activities and assisting to redirect (3) Englyst, H., Wiggins, H.S., & effort to areas best serving public Cummings, J.H. (1982) Analyst safety objectives. 107, 307–318 ƒon-line access to current national (4) Commonwealth Department of information from a range of Community Services and Health/National Food Authority (7) National Food Authority (1991) (1989-) Composition of Foods, Food for Health, Australian Australia, Australian Government Government Publishing Service, Publishing Service, Canberra Canberra (5) Commonwealth Department of (8) Australian Food Standards Code, Community Services and Health Australian Government Publishing (1991) NUTTAB 91–92, AGPS, Service, Canberra Canberra, ACT (9) National Health and Medical (6) Commonwealth Department of Research Council/National Food Community Services and Health Authority (1991) The 1990 (1991) Nutritional Values of Australian Market Basket Survey, Australian Foods, Australian Australian Government Publishing Government Publishing Service, Service, Canberra Canberra Copyright, Food Industry and Food Safety Considerations

Use of Databases for Nutrition Labeling in the United States

James T. Tanner

Office of Special Nutritionals, HFS-451, Center for Food Safety and Applied Nutrition, US Food and Drug Administration, Washington, DC 20204, USA

The Nutrition Labeling and Education Act of 1990 requires that all foods sold in the United States be nutritionally labeled. In response to the requirements of that Act, the Food and Drug Administration (FDA) published final regulations in January, 1993, on enforcement of nutrition labeling. In the regulations, provisions were made for the use of databases for labeling, and a manual was prepared to assist companies and organizations in this task. This manual gives generic information on how to develop and calculate from databases label values which will meet regulations that FDA is required to enforce. FDA does not prescribe how an individual company is to determine nutrient content for labeling purposes but does offer to review a database and to work with the manufacturer to resolve problems before taking any regulatory action. The compliance policy of FDA remains based on analysis of composite samples, performed using methods of the Association of Official Analytical Chemists (AOAC), or other validated methods if no AOAC method exists. FDA will then compare the label values with the results from laboratory analyses. The use of databases for labeling, the regulations of FDA for foods sold in the United States, the compliance policy of FDA, and the future uses of databases are discussed.

n 1973 the US Food and Drug Administration (FDA) promulgated regulations (1) that required nutrition labeling in certain circumstances. The agency took this action Ilargely in response to recommendations of the 1969 White House Conference on Food, Nutrition, and Health (2). manufacturers, however, to provide Ŷ Overview of Database Use nutrition labeling voluntarily on a wider The 1973 regulations required nutrition variety of food products, including the labeling only for certain foods, those exempt foods. with added nutrients or for which a Industry-wide databases were nutrition claim was made in either suggested as a possible means of labeling or advertising. Some foods reducing the cost of developing nutrition such as fresh produce were specifically labeling for individual companies. In exempted. FDA encouraged 1979 FDA, the US Department of Agriculture (USDA) and the Federal product. The firm or organization needs Trade Commission (FTC) encouraged to be judicious in this selection, this concept in a notice (3) published in however, to ensure that the product the Federal Register, describing the labeling is in compliance with the agencies' policies and intentions with regulations for that product. FDA has respect to numerous food labeling developed a manual (5) which will be of issues. In that notice, FDA, while not assistance in identifying data that are of agreeing to approve databases, stated sufficient quality to provide an adequate that it would work with industry to basis for nutrition labeling. In addition, resolve any compliance problems that guidance has been given on when might arise for food labeled on the basis average values may be used and when of a database that the agency had calculated values using the equations accepted. Specifically it stated, “If given in the manual should be used. products bearing nutrition labeling in The agency understands that most accordance with properly evaluated companies will not have sufficient [FDA evaluated] nutrient databases and information to meet the suggested manufactured in accordance with food criteria listed in the manual; however, manufacturing practices are found not to we view this as the “gold standard” and be in compliance with applicable hope that by making a diligent effort nutrition labeling regulations, the agency there will be sufficient analytical data will work with the firms responsible for available in five to ten years' time to the product in question and with the comply fully with the different criteria appropriate authorities who are given in the manual. maintaining the applicable nutrient Historically, label values based on database to correct the problem before calculation of nutrient content from initiating compliance provision actions.” ingredients were considered The policy given in that 1979 notice is unacceptable for a mixed product for the the same policy that is in effect today. following reasons: 1) There are no With the Nutrition Labeling and quality indicators of the data for the Education Act of 1990 (4) expanding components; 2) there are no indicators mandatory nutrition labeling to nearly all of the methods of analysis and sampling foods regulated by FDA, greater interest used to obtain the data for the has been expressed in using components; 3) there are no indicators industrywide databases for some food of the design and execution of quality products. Some manufacturers of food control procedures used to monitor the products not currently labeled have sampling and analysis of the com expressed interest in using data Nutrition Labeling in the United States available from other sources, for ponents; and 4) there are no indicators example, the open scientific literature, of nutrient loss during the processing as the basis for labeling their products. and handling of the mixed product. In The policy of the Food and Drug addition, inclusion of sugars as Administration is that the choice of a mandatory in nutrition labeling and a data source is the prerogative and the change in the definition of fat limit the responsibility of the firm or organization use of most ingredient databases. that provides a nutritionally labeled Table I. Formulas for calculating label values (1) Class I nutrients (fortified): 1/2 Computed value=mean-t(0.95;df)(1/k + 1/n) (s) (2) Class II nutrients (naturally occurring)•80% label value: 1/2 Computed value = [mean - t(0.95;df)(1/k + 1/n) (s)](5/4) (3) Class II nutrients (naturally occurring): calories, sugars, total fat, saturated fat, cholesterol, and sodium, ”120% label value: 1/2 Computed value = [mean + t(0.95;df)(1/k + 1/n) (s)](5/6) where mean = sample mean t(0.95;df) = 95th percentile of the t-distribution df = n-1 degrees of freedom n = sample size used to calculate the mean k = number of future units making up the future mean.

Use of data from the open literature on the nutrient classes (i.e., Class I or and use of ingredient databases present II). The computed limit for Class II similar problems because the values nutrients is adjusted for the 20 per cent given are generally averages based on margin of allowance in the FDA an undetermined number of analyses. compliance evaluation. Average values based on numerous The computing formulas are given in analytical values representing the Table I. Label values computed using different variables associated with a these equations have the highest nutrient may be sufficient if they are assurance of meeting FDA within the range (per cent CV) requirements. represented by the coefficient of Ŷ Class I and Class II Nutrients and variation given in the manual. If the Compliance Policy coefficient of variation is large, then the equation given in the manual should be Compliance with nutrition labeling is used to ensure that the regulatory determined in the following manner: A requirements can be met. This applies collection of primary containers or units to indigenous nutrients only; for of the same size, type, and style fortification nutrients the value must be produced under conditions as nearly at least 100 per cent of the label value. uniform as possible, designated by a If average values are used, then the common container code or marking, or value will be correct only 50 per cent of in the absence of any common the time. The manual gives equations container code or marking, a day's for calculating label values for production, constitutes a “lot.” indigenous and fortification nutrients that The sample for nutrient analysis shall would have the highest probability of consist of a composite of 12 meeting the regulatory requirements subsamples (consumer units), one from which the agency must enforce. each of 12 different randomly chosen shipping cases, taken to be Ŷ Equations for Use in Calculating representative of a lot. Unless a Label Values particular method of analysis is For label values computed to reflect the specified, composites shall be analyzed variability of individual units, the by appropriate methods as given in the onesided 95 per cent prediction interval Official Methods of Analysis of the is constructed to contain the result of a Association of Official Analytical single (k = 1) future retail unit or to Chemists, 15th Edition (1990) (6), or in contain the mean of k = 12 future retail the supplements issued annually. If no units. Suppose there are n individual AOAC method is available or values of a nutrient in the database, appropriate, other reliable and from which the sample mean and SD appropriate analytical procedures are computed. The calculation of a should be used. lower or upper limit of the one-sided 95 Two classes of nutrients are defined per cent prediction interval also depends for purposes of compliance: Class I nutrients: added nutrients in fortified or Reasonable excesses over labeled fabricated foods; and, Class II nutrients: amounts of a vitamin, mineral, protein, naturally occurring (indigenous) total carbohydrate, dietary fiber, other nutrients. If any ingredient which carbohydrate, poly- or monounsaturated contains a naturally occurring nutrient is fat, or potassium are acceptable within added to a food, the total amount of current good manufacturing practice. such nutrient in the final food product is Reasonable deficiencies under labeled subject to Class II requirements unless amounts of calories, sugars, total fat, the same nutrient is also an added saturated fat, cholesterol, or sodium are ingredient. acceptable within current good A food with a label declaration of a manufacturing practice. vitamin, mineral, protein, total Compliance with these provisions carbohydrate, dietary fiber, other may be provided by use of a database carbohydrate, poly- or monounsaturated that has been compiled by following fat, or potassium shall be deemed to be FDA guideline procedures and misbranded under Section 403(a) of the submitted to FDA for approval and with Federal Food, Drug, and Cosmetic Act foods that have been handled in (the Act) (7) unless it meets the accordance with current good following requirements: manufacturing practice to prevent ƒClass I vitamin, mineral, protein, nutrient loss. Guidance in the use of dietary fiber, or potassium: The databases may be found in the FDA nutrient content of the composite is Nutrition Labeling Manual — A Guide for at least equal to the value for that Developing and Using Databases (5). nutrient declared on the label Ŷ Nutrient Database Development ƒClass II vitamin, mineral, protein, and Use total carbohydrate, dietary fiber, The development of a database (a other carbohydrate, poly- or collection of related information) is a monounsaturated fat, or potassium: complex task that consists of several The nutrient content of the general steps such as development of a composite is at least equal to 80 per sampling plan, collection of the samples, cent of the value for that nutrient analysis of the laboratory samples, and declared on the label statistical analysis and interpretation of ƒProvided: That no regulatory action the results. Each of the steps can be will be based on a determination of a performed in several different ways, and nutrient value that falls below this decisions made regarding the level by a factor less than the alternatives may directly affect the variability generally recognized for available resources, the quality of the the analytical method used for that data, and the risk of making incorrect food at the level involved (8). decisions. A food with a label declaration of The process of developing a calories, sugars, total fat, saturated fat, sampling plan involves the resolution of cholesterol, or sodium shall be deemed a series of interrelated tasks that may to be misbranded under section 403(a) be broadly classified as follows: of the Act if the nutrient content of the ƒdefining the sampling objective composite is greater than 120 per cent defining the target product of the value for that nutrient declared on ƒ population the label. Again, the same statement on analytical variability applies. ƒdeveloping the sampling frame ƒselecting the sampling methods (i.e., which is expressed in terms of four stratified multistage plan) characteristics: ƒselecting the analytical methods. ƒprecision (magnitude of the error of To increase the chance that the data the estimate) will be of the desired quality, it is ƒrepresentativeness (how accurately essential that these tasks, as a samples reflect the population) minimum, be given careful ƒcomparability (similarity of data from consideration, and that specific different sources) questions be addressed and resolved in the planning stage of the data collection ƒcompleteness (amount of effort. information collected). It is necessary that satisfactory In using a database for the purpose of degrees of these characteristics be labeling, consideration has to be given reflected in the data. to: Ŷ Calculation of Label Values ƒthe variability of the factors that Once an acceptable amount of influence nutrient content analytical data of satisfactory quality has ƒthe distribution that the nutrient been accumulated, a value must be values follow determined for the label, which will ƒthe statistical methodology that is reflect the nutrient content of the applied in deriving label values. product. This number may be calculated To resolve these tasks effectively, in several ways. information on the variability of the The first and perhaps the most nutrient levels in the product is needed. straightforward way is to use the mean Variables such as variety or geographic of the analytical data. If the analytical growing area for fruits and vegetables data are within certain limits indicated by need to be determined. For mixed the coefficient of variation (usually 11 products and/or products requiring per cent or less), then the mean value processing, the nutrient content may may be used. This applies only to Class change during processing or during II nutrients. The coefficient of variation is storage before sale. Information on the the standard deviation times 100 divided variability of the analytical method for by the mean. The reason for using the the nutrients of interest is also needed. coefficient of variation is that numbers If sufficient information is not which are applicable to all available, it will be necessary to perform concentrations can be given in the a pilot study or perform a literature guidelines. search to obtain the necessary For nutrients that are highly variable, information before developing the i.e., that have coefficients of variation sampling plan. higher than the maximum limits given in A database that would be adequate the guidelines, there is an equation for for the purpose of nutrient labeling will Class I nutrients (which must be at least reflect a satisfactory degree of data 100 per cent of the label claim) and two quality, and hence database accuracy. for Class II nutrients, one for those Data quality (determined by the amount nutrients that must be 80 per cent of the of error that is contained in data) label claim, or greater, and one for those depends primarily on the effectiveness nutrients that may not exceed 120 cent of the activities stated earlier for of the label claim. database development. The accuracy of Once the nutrient content has been a database depends on its data quality, calculated, various rounding rules must be used, such as rounding to the limited by the products for which it was nearest gram or half-gram, etc. developed. FDA believes that in the future Ŷ “Recipe Databases” ingredient composition databases may Although FDA has stated that it would have the necessary validation to be not accept label values generated by used in calculating the final composition computer from ingredient or component of mixed products for an acceptable composition values or “recipe range of food products. At this time, databases,” it was stated that FDA however, the agency believes that the would work with various trade data making up ingredient databases organizations and companies to develop are of mixed quality and, therefore, of a model which could be used to limited value. Companies that wish to calculate label values based on use ingredient databases must evaluate ingredient composition and the individual analytical values for each determination of nutrient loss (if any) ingredient to assure themselves that the during processing. To date we have data are representative, valid from an worked with several organizations but analytical perspective, and sufficient to have not yet accepted any models. In account for any variation in the the comments received after the manual ingredient. was announced in the Federal Register, During the comment period several a large number stressed the need for principles were given by companies and recipe databases. FDA still believes that trade associations that would be useful most ingredient composition databases and instructive for developing recipe are not useful for calculating the final databases. Keeping in mind the general composition of a mixed product for the requirements for databases, these following reasons: proposed principles are given as ƒsuch databases are usually not general guides: accompanied by indicators of the ƒConfidence in the quality of data, quality of the data for the supported by documentation of data components sources. Companies maintaining or ƒsuch databases have no indicators using ingredient composition of the methods of analysis and databases must be able to sampling used to obtain the data for demonstrate the data source used the components for each type of product and each nutrient for which ingredient ƒsuch databases have no indicators composition databases are utilized. of the design and execution of quality control procedures used to ƒProper maintenance of the monitor the sampling and analysis of database. Companies developing or the components using ingredient composition databases must have procedures in ƒsuch databases have no indicators place to ensure that the values in the of nutrient loss during the processing ingredient composition databases and handling of the mixed product. are reviewed and updated as To successfully validate a model to needed and on a regular basis. calculate final composition based on ingredient composition, extensive ƒSpecificity with respect to analyses of ingredients and final product ingredients, product formulations, composition would be required. The and processes. Companies using application of this model would be ingredient composition databases must have procedures in place to ensure that the nutrient values are confidential company information, and used only for specific applications. there should be some assurance that For example, a company should this information will not be available to have a procedure to ensure that anyone who requests it. The agency is nutrient data specific for one product aware that the development of a formulation or process are not used database is costly, and that it may to prepare nutrient declarations for contain confidential information. FDA similar product formulations or agrees that release of a database could processes, when there is no reveal substantial proprietary interests in assurance that the data are documents that have been submitted to applicable to those products or the agency. Furthermore, it has never processes. been the agency's intent, nor does it ƒValidation of the database. have the resources, to maintain and Companies developing or using manage databases that are developed ingredient composition databases by manufacturers or associations. The must have procedures in place to agency believes that the availability of a ensure that nutrient values receive database is, therefore, the primary reviews, audits, and confirmation responsibility of the developer. through nutrient analyses as often FDA will continue with the policy of as necessary. assisting the developers of databases, providing guidance to those who ask for Several other companies offered it, and evaluating databases for the equivalent guidelines and one other products submitted for review. guide that is important: Confidentiality of such data will be determined and maintained in ƒCompanies making nutrient content accordance with regulations. or health claims should substantiate Those database developers who the claim by analysis of the final choose to do so are encouraged to product. This applies to products make their information available through labeled with a health message or such compilations as the USDA those making claims such as low or Handbook No. 8 (9), so that all may reduced in certain nutrients. benefit from the additional analytical Ŷ Confidentiality of Databases information. In the long run, recipe During the comment period many databases will be useful after extensive companies/trade associations objected information is gathered and placed in to the lack of confidentiality of submitted these compilations of public information. The information and procedures databases. They did not want the given in the manual are generic, and information gained through analyses of products and ingredients to be released only the parts that pertain to the food or through freedom of information requests product under study should be used. or used in unacceptable ways or for Because preparation of a different manual for each food type was not inappropriate products. In addition, in practical, all the information is contained developing databases, costs are shared among the participating companies; in one manual. FDA is always available these companies sought assurance that to help companies with problems or the data would not be available without those needing assistance in determining cost to companies that did not how to proceed in attaining the best label possible while continuing to satisfy participate in development. Formulations that are used to produce the regulations that the agency must mixed products are also regarded as enforce. Ŷ References Center for Food Safety and Applied (1) Anon. (1993) Federal Register 38, Nutrition, U.S. Food and Drug 20702 Administration, Washington, DC (2) White House Conference on Food, 20204 Nutrition, and Health (1969) U.S. (6) Official Methods of Analysis (1990) Government Printing Office, 15th Ed., AOAC, Arlington, VA Washington, DC (7) Federal Food, Drug, and Cosmetic (3) Anon. (1979) Federal Register 44, Act (1938), P.L. 75–717 75990 (8) Horwitz, W., Kamps, L.R., & Boyer, (4) Nutrition Labeling and Education K.W. (1980) J. Assoc. Off. Anal. Act (1990), P.L. 101–535 Chem. 63, 1344–1358 (5) FDA (1993) Nutrition Labeling (9) US Department of Agriculture Manual—A Guide for Developing (1976- ) Composition of Foods: and Using Databases; available by Raw, Processed, Prepared, Agric. sending a self-addressed mailing Handbook No. 8 series, USDA, label to: James T. Tanner, HFS451, Washington, DC Copyright, Food Industry and Food Safety Considerations

Functional Foods for Specific Health Use — The Needs for Compositional Data

Kazuki Shinohara

National Food Research Institute, MAFF, Tsukuba, Ibaraki, Japan

National projects on the physiological functions of foods, which have been carried out by university groups and the Ministry of Agriculture, Forestry and Fisheries in Japan, revealed that foods have functions controlling homeostasis in the body, as well as nutritional and sensory functions, leading to the introduction of the terminology of functional foods, and in 1991, a system for licensing “Foods for Specified Health Use”. To obtain a license for a “Food for Specified Health Use”, the following evidence is required: a) potential to contribute to the improvement of dietary habits and the maintenance and enhancement of health; b) medical and nutritional data on specific health aspects; c) data on intakes determined medically and nutritionally; d) safety data; e) physicochemical data; f) data on nutritional composition, and others. As such foods proliferate there will be a need to collect comprehensive data on them and to make these data systematically available to doctors, dietitians and others.

t present, one can obtain nutritious and appetizing foods whenever necessary in industrialized countries and the daily diet is becoming satisfactory in terms of Aquantity. Longevity has increased considerably due to improved nutrition and medical treatment. However, with a change in food habits, food-related diseases such as allergy, obesity and geriatric diseases are increasing in Japan. Because of the costs of treatment of geriatric diseases, total medical expenditure is expected to increase. Furthermore, the aged population is expected to reach a maximum in the next century. From these points of view, new health aspects of foods are a public concern. In order to respond to the need for improved foods, it is important to elucidate the physiological functions of food components. Ŷ Research on the Physiological ƒsystematic evaluation of the Functions of Food Components physiological functions of foods (1984– 1986) National projects on the physiological functions of food components have ƒevaluation of bioregulatory functions been carried out by university groups in of foods (1988–1990) Japan under the following headings: ƒanalysis of functional foods and molecular design (1992–1994). These projects have revealed that chemical structure, clear mode of action foods have biomodulating functions at cellular level, proven effectiveness by (tertiary) which control homeostasis in oral administration, safety, stability in the body, as well as nutritional (primary) foods, acceptability as food, and and sensory (secondary) functions. The potential for use in a range of diverse biomodulating functions include food products. In comparison with most regulation of immunological functions, conventional nutritious foods, functional regulation of biorhythms, prevention of foods have more potential efficacy for ageing, prevention of food-related health. Functional foods claim to control disease, and facilitation of recovery from homeostasis, resulting in the prevention food-related disease. In the Ministry of of geriatric diseases. In contrast to Agriculture, Forestry and Fisheries, the conventional healthy foods, these foods following projects have been carried out: have direct scientific evidence about ƒthe Japanese R&D Association for a their functionality. The member new separation system in the food companies of the Japan Health and industry (1989–1991) Nutritional Food Association collaborate in working groups which are responsible ƒthe Japanese R&D Association for for collecting scientific evidence about new food materials (1990–1993) 11 identified categories of functional ƒelucidation of the functions and components: dietary fibers; molecular structures of food oligosaccharides; sugar alcohols; components (1989–1992) polyunsaturated fatty acids; peptides ƒintegrated research program for and proteins; glycosides, isoprenoids effective use of biological activities and vitamins; alcohols and phenol; to create new demand (1991–2000) choline; lactobacteria; minerals; and others. ƒdevelopment of technology for evaluation and utilization of Ŷ Physiological Functionality of Food functional properties of agricultural Components products in relation to health (1993– 1999). The physiological functions of food From these projects, the components are as follows: physiological functions of components of Dietary Fibers vegetables, soybeans, milk and other foods or their metabolites have been Dietary fibers such as polydextrose, discovered, such as antimutagenicity, wheat, bran, corn, apple, soybean, and anticarcinogenicity, antioxidant activity, beet fibers have been recognized to immunostimulation, modifying endocrine have beneficial effects. Fibers are function, hypotensive effect, cholesterol considered to be promising candidates control, intestinal control, and others. for functional components of foods. The physiological functions of dietary fibers Ŷ Functional Foods are claimed to be: a) regulation or These projects exerted a great influence control of intestinal functions, including on the Japanese food industry and the prevention of constipation, improvement Ministry of Welfare and Health (MWH) of bacterial flora in the intestine, leading to the introduction of the inhibition of absorption of harmful terminology of “functional foods”. A substances and promotion of their functional food was defined as a food excretion, prevention of colon cancers, containing compounds which satisfy the and immuno-stimulation, b) regulation or following criteria: clear effectiveness for control of blood sugar content, including a specific health purpose, defined inhibition of insulin secretion, inhibition of glucagon secretion, and prevention of Peptides and Proteins diabetes mellitus, and c) regulation or These include casein phosphopeptide control of cholesterol levels, prevention (CPP), lactoferrin, and peptides from of gallstone formation, decrease in fat casein, fish, and soybean. CPP deposition, prevention of obesity, and prepared from milk is a promising hypotensive effects. Dietary fibers can functional ingredient which enhances be used in foods (insoluble fibers) and the absorption of calcium and iron. beverages (soluble fibers). Polydextrose Recently, hypotensive peptides have has been used in beverages as a been recovered from milk casein, fish substitute for sugar. proteins and soybean. Their functional Oligosaccharides properties are claimed to be: Oligosaccharides such as lactulose, hypotensive effect, control of cholesterol fructo-, galacto-, isomalto-, and xylo- level, detoxification of harmful oligosaccharides, and cyclodextrins are substances, antiviral effects, and considered to have potential for use as promotion of bone and tooth growth. functional ingredients in foods. They can Glycosides, Isoprenoids and be used as food modifiers that do not Vitamins affect the texture and physicochemical The functions of this class which properties of foods and can be includes saponins, carotenoids, replacements for sugar. The flavonoids, and tocopherols are claimed physiological functions are claimed to to be: antioxidant activity, intestinal be: low energy, prevention of tooth control, improvement of stomach, liver, decay, intestinal control, and and kidney metabolism, decrease in bifidobacterium activation. Products blood sugar and cholesterol levels, and containing oligosaccharides are hypotensive effects. believed to have great sales potential in Japan. Alcohols and Phenols Sugar Alcohols The functions of this class which includes tea polyphenols, oryzanol, and Sugar alcohols such as malcitol, octacosanol are claimed to be: erythritol, and reduced palatinose are prevention of tooth decay, deodorant promising materials which have low effect, hypotensive effect, and decrease energy content and preventative effects in plasma cholesterol. Among tea against obesity and tooth decay. polyphenols, catechin, epicatechin, Polyunsaturated Fatty Acids epicatechin gallate, and epigallocatechin Polyunsaturated fatty acids, especially gallate have received considerable eicosapentaenoic acid (EPA), attention. docosahexaenoic acid (DHA), and Cholines linolenic acid, have great potential for This class of compounds, including sales and development into functional soybean and egg yolk lecithins, is foods. EPA and DHA are derived from claimed to have functions such as fish body oils. Their functionalities are improvement of plasma lipid claimed to be: decrease in fat metabolism, prevention of deposition, decrease in plasma arteriosclerosis, and prevention and cholesterol, hypotensive effect, improvement of fatty liver. decrease in viscosity of blood, and prevention of cancers of the breast, Lactobacteria colon and prostate gland. Fermented foods with lactobacteria and obtain the health use through the bifidobacteria are popular in Japan, consumption thereof” (1). because they have the function of “Foods for Specified Health Use” are intestinal control, decrease in defined as a category of food for special cholesterol, and immuno-stimulation. dietary uses within the establishment Minerals frame of the Special Nutritive Foods under article 12 of the Nutrition The claimed benefits of minerals (e.g., Improvement Act (Figure 1). calcium salts, heme-iron, magnesium To obtain a license for a “Food for include: promotion of bone and teeth Specified Health Use”, evidence that the growth, prevention of osteoporosis, and food meets the following requirements prevention of anemia. must be produced: Others ƒpotential to contribute to the The last category are classified as improvement of dietary habits and health-enhancing foods which cannot be the maintenance and enhancement assigned to any of the categories above. of health It includes fermented vinegar, and ƒmedical and nutritional data on Chlorella. The claimed benefits are: specific health aspects improvement of plasma lipid ƒappropriate serve size of the food or metabolism, immuno-stimulation, and relevant compound, based on anticarcinogenic effects. medical and nutritional information Ŷ Foods for Specified Health Use ƒsafety data of the food or relevant In 1991, a system for licensing “Foods compounds for Specified Health Use” was ƒphysicochemical properties, test established by MWH. According to this methods and methods of qualitative system a food for specified health use is and quantitative determination “a food which, based on knowledge data on nutritional composition, concerning the relationship between ƒ since the nutritional composition of foods or food components and health, is the product should not be defective expected to have certain health benefits, when compared with the and has been licensed to bear a label composition of the regular food claiming that a person who uses it for product that it may replace. specified health use may expect to Figure 1. Types of special nutritive foods In addition to these requirements, the pills or capsules. Only functional foods product should preferably be a food that which satisfy these requirements will be is consumed regularly in the general approved by MHW as a Food for diet, rather than a food consumed only Specified Health Use. occasionally. The product should be in Official permission for licensing a the form of ordinary food, rather than Food for Specified Health Use is under the control of the MHW. In the system for patients with atopic disease. Low for licensing, overseas applicants need phosphate content milk is effective for to apply directly to the MHW (Office of patients with chronic renal diseases who Health Policy on Newly Developed are instructed to reduce the intake of Foods, Environmental Health Bureau). phosphate. The phosphate content of The application for licensing by the the product is 20 per cent of that of MHW must be in writing and must be regular milk while the protein content accompanied by a product sample. The remains the same. In addition to these, MHWs minimum requirements for an foods such as beverages containing application are: brand name of food, list soybean-, fructo-, xylo-oligosaccharides, of ingredients by percentage, details of or calcium salts, table sugars containing manufacturing procedure, analysis of soybean- or fructo-oligosaccharides, ingredients, matters for which and gums containing malcitol and permission or approval is sought, name, palatinose have also been approved as address, and date of birth of applicant, Foods for Specified Health Use. Foods name and address of main office, for Specified Health Use containing reason for seeking permission or dietary fibers, oligosaccharides, approval, statement of energy value, list lactobacteria and others will be licensed. of nutritive elements by amount, As such foods proliferate there will be a instructions for storage, preparation, and need to collect data on them and to administration, and any precautions to make this available to doctors, dietitians be observed in use of such foods. and others, particularly those concerned Future Prospects with nutritional epidemiology and controlled dietary intervention trials. Recently, allergen-free rice (fine rice) and low phosphate (LPK) were Ŷ Reference approved by the HMW as the first Foods (1) Nutrition Improvement Act (1991) for Specified Health Use. Allergen-free Article 12, Ministry of Health and rice in which 99 per cent of allergic Welfare, Tokyo globulin protein is removed is effective The Second International Food Data Base Conference

Food Composition Research: The Broader Context

August 28–30, 1995 Lahti, Finland

Eurofoods-Enfant Meeting as a satellite August 31 For further information: The Second International Food Data E-Mail: [email protected] Base Conference: The Eurofoods-Enfant Meeting:

Prof. Lea Hyvönen Prof. Clive West Dept. Food Technology Dept. Human Nutrition P.O. Box 27 (Viikki, Building B) Wageningen Agricultural University FIN-00014 University of Helsinki P.O. Box 8129 Wageningen 6700 EV Finland The Netherlands Tel: +358 0 7085215 Tel: +31 8370 8 25 89 Fax: +358 0 7085212 Fax: +31 8370 8 33 42