W. R. WOLF
U. S. Department of Agriculture, NCL — Building 161, Beltsville MD 20705, U.S.A.
1. Eurofoods interlaboratory trial 1985 on laboratory procedures as sources of discrepancies between food tables
The draft report of this interlaboratory trial was distributed to all participants and was discussed in detail by Peter Hollman (State Institute for Quality Control of Agricultural Products, Wageningen). The between laboratory variations observed in this interlaboratory trial were unacceptably high for total fat, available carbohydrates and fibre. It was felt that differences in methodology were partially responsible for the high variations in total fat. There were a number of different methods used for available carbohydrate. These differences are important contributions to the between laboratory variation. For fibre there are well recognized differences in principle and in resulting analytical values between the several methods used. For example, the neutral detergent method does not determine several components that other procedures include and as a result gives lower values for those samples that contain significant amounts of those components. It was recommended that fibre data be evaluated using only values from the Prosky-Asp methodology as this procedure is being tested for official AOAC approval.
Based on the results of this interlaboratory trial, it was proposed that Eurofoods initiate a programme to develop several suitable reference materials characterized for these important nutrient constituents. Peter Wagstaffe of the Community Bureau of Reference (BCR), CEC, Brussels, presented some discussion on BCR activities and interests in this area. It was agreed that a proposal be developed for a joint programme between Eurofoods (Hollman/Katan at Wageningen) and BCR (Wagstaffe, Brussels) to produce appropriate reference materials in this area.
2. In-house quality assurance programmes
Discussion of this topic led to a number of recommended areas in which Eurofoods should encourage further activity: Development and use of in-house quality control materials to carry out and document proper quality assurance of analytical data; recommendations for inclusion of more analytical detail in publications; informal inter-laboratory exchanges of samples using available inhouse quality control materials, and translation from Swedish of extensive guidelines developed by the Swedish National Food Administration for a Good Laboratory Practices programme for nutrient analysis.
3. Inventory of Eurofoods resources related to food composition
A listing of laboratories and personnel associated with nutrient analysis and food composition should be compiled. Initially this list would include the laboratories involved in the Interlaboratory Study and the participants in this Eurofoods meeting. It should be expanded to include additional laboratories and resources.
4. Check sample programmes
It was generally agreed that appropriate check sample programmes would potentially have significant merit in improving analytical values for nutrient analysis within the Eurofoods community. At present the Swedish National Food Administration has an active check sample programme where two to four samples are sent once a year to 35 laboratories. There would be a possibility of including some additional Eurofoods laboratories in this programme. For details of this programme contact Dr I. A. Torelm, Swedish National Food Administration, Box 622, Uppsala, Sweden.
Dr P. B. Czedik-Eysenberg, (Chairman, Working Party Food Chemistry of the Federation of European Chemical Societies, Vienna) has agreed to conduct a survey of interest and feasibility of establishing a check sample scheme for Eurofoods laboratories.
5. Meeting of analytical committee
It was generally felt that the projects proposed did not require a formal meeting in 1986. Persons responsible for the various proposed projects would handle necessary personal contact, communications and planning to initiate and carry out the projects.
A. M. FEHILY
M.R.C. Epidemiology Unit, 4 Richmond Road, Cardiff CF2 3AS
The problems of handling nutritional data specific to epidemiological surveys were discussed. The main problem is that surveys are conducted on very large numbers of subjects (representative population samples). It is recognized that there is no ideal dietary survey method, but whichever method is used, data must be converted to a form which will enable calculation of nutrient intakes by computer. At present this appears to be done mostly by manual coding. Among the survey team there may be only one experienced nutritionist and it is almost impossible for one person to collect and code all the data accurately within the time available. Therefore, additional trained fieldworkers and coders are required. Manual coding introduces the possibility of a large number of errors, even when the coding is done by an experienced nutritionist, simply due to the volume of data collected. These errors are of course reduced by code-checking conducted by a nutritionist, but because of the number of subjects being studied, this is extremely time-consuming and will probably not eliminate all the errors. Transfer of the coded data to computer is only a minor source of effort since data entry is normally verified.
The possible uses of computers as more efficient tools to overcome these problems were discussed. Entry of data by the coder into a microcomputer for pre-processing prior to transfer to a mainframe computer for storage or nutrient calculation is one possibility. The food code or name could be entered and verified, followed by the amount. Several systems of this type have already been developed and were demonstrated during the poster presentations. These were, however, designed for use by dietitians and nutritionists. For epidemiological surveys a system suitable for use by non-nutritionists would be more valuable.
Where recall dietary methods are used, it may be possible for interviews with subjects to be conducted so as to enter data directly into a microcomputer, as was done in the USA National Health and Nutrition Examination Survey (NHANES). In this study, subjects were asked questions about their food intake, the program being menu-driven i.e. each question was dependent upon the answer to the previous one. This approach eliminates observer variation and the need for data coding, but requires that the subject attend a clinic and that a nutritionist be present to determine portion sizes.
Where dietary recording methods are used, depending upon the purpose of data collection and the nutrient to be studied, the Food Recording Electronic Device (FRED) may be useful (Stockley et al. in press). This device eliminates the need for coding of data, but foods need to be grouped and composite dishes may be a problem.
For both recall and recording methods, data could be entered into a microcomputer by a trained person, who need not be a nutritionist, if an adequate software package were available. In order to be useful, a large part of the skill of the nutritionist would need to be built into it. This requires a first-class liaison between the programmer and the nutritionist. Existing packages are slow if food names rather than code numbers are entered, so that the speed of the search procedure would need to be increased. For foods and composite dishes not present in the food composition tables, one or more substitute codes with the same nutrient composition can often be used and a list of those occurring most frequently kept on file. Amounts of foods could be entered not only as weights but as household measures and converted to weights by the programme. Range checks on the weights of foods entered are an essential part of the pre-processing system to identify possible gross errors, e.g. dry weights coded as liquids or vice versa. Such checks could and should be incorporated into all existing programs where data are coded manually.
The main disadvantage of this proposed pre-processing system, apart from the time and cost required to develop the software, is that nutritionists become more remote from the raw data and errors are more difficult to detect on computer print-outs. The main advantages are that coding errors and the number of hours a nutritionist is required would be reduced. In the longer term, by reducing the clerical load on qualified nutritionists, their efficiency would be enhanced.
L. BERGSTRÖM
The Swedish National Food Administration, Box 622, S-75126 Uppsala, Sweden.
The workshop started with a short survey of the results achieved in the different fields; recipe calculation systems; recipes, analysed, calculation and corrected with NLG factors; NLG research; NLG references; and NLG data base (see these proceedings pp. 8–12).
Dr Bognar described how to use yield factors for energy and nutrients, see appendix.
In the first part of the workshop the following activities for 1985–86 were suggested; collection of retention factors connected with the different cooking methods; collection of food yields for different dishes; collection of 100 recipes for the most common dishes consumed in each European country, and — if possible — inclusion for the recipes of the following parameters; cooking procedures, weight losses, cooking time and cooking temperature.
The discussion then turned to the different parameters in the cooking of food. At the Swedish National Food Administration a pilot data base for storing information on these different parameters has been created. (Mr Hernan Isakson describes the model on pp. 78–79 of these proceedings).
Conclusions
Recommendations for the future work, suggested above, were made. (1) It was suggested that the group should start with the compilation of the 100 recipes for the most commonly consumed dishes in the different European countries. Information on cooking method, procedures, time, temperature, weight loss or food yield should be collected. (2) Analytical and calculated nutrient values for the dishes should also be collected and analysed. (3) Standardized recipe forms should be made. (4) The descriptions of the different recipe calculation systems must be more informative, therefore a new inventory must be performed. (5) The collection of references on NLG factors and related topics must continue.
After discussion of the factors to be taken into account it was finally decided that only 25 recipes for the most common dishes consumed in each European country should be collected. From these recipes ten recipes, representing different food groups and cooking methods, will be chosen. All possible information will be collected about these dishes. The dishes will be analysed and calculated for nutrients on European recipe calculation systems. The results will be analysed.
These results and the results from the NLG project 1985, together mainly with Dr Bognar's research on NLG, will be the base for NLG factors to be suggested at the next Eurofood meeting.
Appendix
Methods for the determination of nutrients and energy in cooked dishes (See: Bognar, Karg: Empirical comparison of different methods for the determination of the energy and nutrient contents of dishes, BFE-Bericht, Karlsruhe 1985)
1) A = analysed = Z (i,p,k)
the content based on chemical analysis, of constituent i in 100 g of recipe p of cooked dish k edible portion.
2) C = calculated = C (i,p,k)
the content of constituent i per 100 g cooked dishes (p, k) are calculated on the basis of the raw food composition-tables (Souci, Fachmann, Kraut, Germany) and the data for the recipe p used for dish k or chemical analysis of a sample of dish p, k in raw state. Then the changes caused by processing in the weight are taken into account using the yield factor for weight for dish p, k.
 | (1) |
 | (2) |
| X | (i,p,k) = | The constituent i in 100 g of recipe p of dish k in the raw state. |
| X | (i,j) = | The content of constituent i in 100 g of ingredient j (edible part) from Nutrition Tables |
| U | (j,p,k) = | Quantity of ingredient j in grams in recipe p for dish k for raw food in the raw state (edible part) |
| J | (p,k) = | Index quantity of the ingredients contained in recipe p for dish k |
| U | (p, k) = | Total quantity of the ingredients in grams in recipe p for dish k in the raw state, where |
 | (3) |
To determine the yield factors for weight the following data are needed:
| U | (p,k) = | Total quantity of the ingredients in grams in recipe p for dish k in the raw state (edible part) |
| V | (p, k) = | Total quantity in grams of the ingredients in recipe p for dish k in the cooked state (edible part) |
Then d (p,k) is the yield factor for weight for recipe p for dish k where
 | (4) |
and d (k) is the average yield factor for weight for dish k where
 | (5) |
Pk = Number of recipes
3) D = calculated from “C” using specific yield factors d (i,k) for nutrients and energy = D (i,p,k):
| D (i,p,k) = C (i,p,k) . d (i,k) | (6) |
To determine the yield factors for constituents the data for Z (i,p,k) and X (i,p,k) for p = 1, . . ., Pk are needed. d (i,p,k) is the yield factor for constituent i in recipe p for dish k where
 | (7) |
and d(i,k) is the average yield factor for constituent i in dish k where
 | (8) |
L. STOCKLEY and C.A. HURREN
ARFC Institute of Food Research, Norwich, Colney Lane, Norwich, UK.
The most usual method for prospective intake measurement is the weighed inventory. However, there are a number of problems associated with the use of this method. These include interference with habitual dietary patterns, low response rates from the less literate members of the population, and the time consuming nature of the subsequent coding of the records and transferral of data to the computer.
A system has been developed which eliminates the need for a subject to read a balance or use a notebook, stores quantitative data for a period up to three weeks, and whereby at the end of the survey period the data can be transferred directly to a host computer holding food composition data, for the calculation of nutrient intake. The system (known by the acronym FRED, Food Recording Electronic Device), consists of an electronic balance, interfaced to a microprocessor. This is housed in a central unit of which a touch sensitive keyboard forms the upper part.
Each key can be labelled with the name of a food, or a food group. Eighty-four food groups have been developed from McCance and Widdowson's The composition of foods' (Paul & Southgate, 1978), and have been validated for the estimation of energy, protein, and fat consumption. Data from 251 24-hour dietary records were used to compare intakes calculated using either the food groups or the tables from which they were derived, with duplicate diet analyses. The mean differences are shown in the Table. The correlation coefficients between analysed and calculated results were similar for both methods of calculation.
The FRED has been tested in 29 subjects, who used the system and at the same time kept a conventional seven day weighed diary. Trials of the method in groups of people who typically do not respond well in diary studies, are currently underway.
This alternative approach may provide a cost-effective means of prospectively collecting quantitative dietary information in large groups of subjects.
Table. Overestimation (+) and underestimation (-) of analysed results using food composition tables or food groups, with coefficient of variation (CV). n = 251.
| Mean analysed value | Food tables | Food groups | |
| Energy | 2168 kcal | -1.9% | -1.5% |
| (CV) | (34) | (33) | (33) |
| Protein | 71.0g | +2.0% | +0.3% |
| (CV) | (30) | (30) | (32) |
| Fat | 96.6g | -8.7% | -8.5% |
| (CV) | (35) | (36) | (36) |
Reference
Paul, A. A. & Southgate, D. A. T. (1978): McCance and Widdowson's The composition of foods', 4th edn. London: H.M. Stationery Office.
J.C. RIGG
Pudoc, P. O. Box 4, 6700 AA, Wageningen, The Netherlands.
What can a linguistic data bank do? An electronic and organizational network can combine the functions of:
In the Netherlands, we have about 100 agricultural (including food industry) institutions, often working in several languages for scientific communication, trade and policy matters. Many individuals and institutions are involved in terminological commissions, in the compilation of glossaries, thesauri and classifications, and in communication skills.
The extent of institutional involvement is being surveyed by questionnaire. Various files have been tested with a data base management system (Synergy on DEC 350). We hope to test termbank software being developed at Manchester (UMIST) and at Heidelberg.
Ultimately the linguistic resources should be interchangeable between active users of the system, with mechanisms for interaction between different groups of experts and with continuous improvement in the content of the files. Active and passive users could have access on line or by exchange of discs.Experiments will continue on DES 350, IBM Professional and DEC VAX.
W. HARE, W. DOYLE, P. DRURY and M. CRAWFORD
Nuffield Laboratories of Comparative Medicine, Institute of Zoology, Regent's Park, London, NW1 4RY.
The accurate calculation of the fatty acid content of foods faces two important problems. Firstly, there is dearth of information on the fatty acid composition of foods, moreover, much of the published data on the fatty acid composition of foods is inaccurate, and secondly, there are enormous variations in fatty acid composition between the same types of food.
Data base accuracy
Published information on the fatty acid composition of foods frequently contains important inaccuracies.
(a) Long-chain n-6 and n-3 essential fatty acids (EFA). The long-chain EFA are sometimes listed when almost certainly absent. This is likely to be due to the fact that the fatty acid composition of the food fat is usually determined by gas-liquid chromatography (GLC). Slow running components on GLC (long-chain fatty acids with unusual groups of configuration, or possibly compounds other than fatty acids) may be confused with the long chain EFA. Equally, the long-chain EFA may not be listed when present, e.g. in animal products such as milk, meats and meat dishes in McCance & Widdowson's tables of food composition. The high biological potency of the long chain EFA and their relevance to questions concerning circulatory and other diseases mean that these inaccuracies in their assessment should not be ignored.
(b) Positional and geometric isomers. Data on the contents of fatty acid isomers of foods are absent from published tables of food composition. Other sources of information give a very incomplete picture of this significant aspect of commonly consumed foods. The recommendations of the COMA (1984) report have emphasized the need to monitor the intakes of fatty acid isomers, which it suggests should be considered together with the saturated fatty acids. However, conventional analytical techniques do not separate the isomers from other unsaturated fatty acids.
There is no solution to problems (a) and (b) other than to carry out detailed analyses of the fatty acid content of those foods for which information is required but unavailable.
Fat subcoding
Many foods incorporate large amounts of added fat, e.g. fried foods such as chips, and home or commercially prepared foods such as cakes. The nature of the fat used in these foods will vary with individual choice of cooking fat, or with restaurant, or with the brand of food.
To accomodate this variability, the NLCM diet analysis computer programs incorporate a ‘fat subcoding’ routine. All foods containing added fat allow selection for both the amount and type of added fat. The amount is expressed as a percentage, e.g. 11% for chips indicates that 11% of the weight of cooked chips is added fat. Standard % values for most common foods may be read from file automatically.
The fat subcoding is done during entry of a subject's food record onto the computer. For example, entering the code and amount for chips causes the additional questions to follow: Code for fat? % fat?
With this information, the correct amounts of the food and the selected fat are computed, and the energy and fatty acids etc. derived from the fat are added to the data for a ‘no fat added’ chip (the latter data obtained by calculation).
The use of this technique answers the problem of the variability of the fat added to foods, and importantly, also greatly simplifies the problem of updating the fatty acid data for these foods.
Committee on Medical Aspects of Food Policy (1984): Diet and cardiovascular disease. London: HMSO.
B.L. NYGARD
KLEM, Akerschus Sentral Hospital, Oslo, Norway.
KLEM (Institute for Clinical Nutrition and Medical Research) and NUME (Nutrition Medical Products) have in co-operation with Elisabeth Kindt, dietitian at the National Hospital and Professor Sverre Halvorsen at the Oslo City Hospital, developed a computer program for children with phenylketonuria (PKU). A prototype of this programme was demonstrated at the Eurofoods meeting.
PKU children are strongly dependent on a phenylalanine controlled diet. If the child's diet is not monitored as early as possible, mental retardation is likely to occur. The recommended intake of phenylalanine for these children will vary, as the same intake of phenylalanine may have different effects on plasma concentration of phenylalanine in different children. Each child has to be treated individually and is dependent on continuous medical and dietary support.
The intention of the PKU program is to help parents and children to self-monitor the phenylalanine intake in a simpler and better manner. Besides, the program gives the child an opportunity to take on the responsibility for his or her diet. This is even more important when the child grows older and the diet becomes less restrictive.
The program is developed for Commodore 64 and is available on tape as well as diskette. The program monitors the phenylalanine content of the diet and relates this to the child's allowance of phenylalanine. It is also possible to monitor the energy, fat and protein content if desired.
KLEM is an independent Norwegian research institute. An important aspect of KLEM's work is to develop new tools in the treatment of dietary disorders. KLEM co-operates with NUME, which is responsible for production and marketing of the computer programs. These will be tailored for various other countries, in which KLEM is collaborating with nutrition consultants.
S. BASSHAM and L.R. FLETCHER
Department of Mathematics and Computer Science, University of Salford, Salford MS 4WT.
In 1982 work started on the nutrient data access system ‘Microdiet’. Specifications were drawn up according to dietitians' and nutritionists' requirements and software developments have been undertaken according to their requests and in line with advances in nutritional science. Early stages of the work are described and some recent developments outline by Bassham, Fletcher & Stanton (1984).
The availability of specially designed, easy-to-use microcomputer software and experience in its use have led to changed expectations of what dietitians and nutritionists can achieve and also enabled them to deal with heavier workloads at a time of increasing public awareness and medical interest in the significance of diet.
Inter-active microcomputer access has also changed users' qualitative expectations. For example, in dietary assessment work, comprehensive and detailed nutrient analysis is now taken for granted and extra facilities are requested as a result of learning about computers' capabilities both from experience and from attendance at training courses.
The first users of ‘Microdiet’ who were involved at the project's inception already did simple dietary assessments (usually for as few nutrients as possible) by hand. The original software specification translated familiar manual processes to microcomputer. Nutrient totals and daily averages were required for comparison with appropriate intake recommendations. Soon, users asked that recommendations be computerised for easy direct comparison and software was written so that standards could be entered and then displayed alongside actual nutrient intakes.
Further analysis options have been added making a total of twelve in the nutrient accumulation module of ‘Microdiet’. In response to the NACNE recommendations (Health Education Council, 1983), analysis of energy intake from fat, protein, carbohydrate and alcohol is often required. An energy profile is given and the user may alter the conversion factors if this is appropriate. Similarly the p:s ratio for fat consumed is displayed in addition to an intake analysis by individual fatty acids. The latest routine allows nutrient totals to be filed for future use and for manipulation by other software packages. Within a month of this facility being available, requests were received for further developments of this type.
Summary. Initial programme design and later developments have been undertaken with the aim of meeting dietitians' and nutritionists' requirements and frequent continuing contact with many endusers have maintained the professional value and usefulness of the software. Progress in hardware availability and software use has made end-users more aware of the possibilities offered by microcomputers. There are now tremendous opportunities for innovation and development in this area.
Bassham, S., Fletcher, L. R. & Stanton R. H. J. (1984): Dietary analysis with aid of a microcomputer Journal of Microcomputer Applications 7, 279.
Health Education Council (1983): A discussion paper on proposals for nutritional guidelines for health education in Britain.
K. LOUEKARI
Ministry of Agriculture and Foresty of Finland, Food Research Program, Viikki 22, SF-00710 Helsinki, Finland.
Intakes of heavy metals are estimated in many countries, due to relative closeness of intake rates to the acceptable daily intakes (ADI) established by WHO. However, the elevated heavy metal exposure is considered as a consequence and an indication of environmental pollution. Methods used in estimating food contaminant intakes and results obtained vary considerably. It is seldom discussed whether the variation of intake rates between countries is due to differences in food consumption or in environmental contamination rate between countries.
The present comparison of heavy metal intakes between Finland, West Germany and Japan suggests that intake of lead, cadmium and mercury is lower in Finland than in the other two countries. Compared to West Germany the intakes in Finland are three to four times lower. The intake of cadmium in West Germany and in Japan are 70 and 65% of the ADI-value. In these countries among some groups of consumers the ADI is probably approached or exceeded, due to the fact that the intake of heavy metals varies considerably within the population. The content of lead and cadmium in single food items were compared and again Finnish food seems less contaminated than food of other countries. The lead content of wheat, potato and meat is one and a half to six times higher in West Germany and Japan compared to Finland. The largest difference is observed in lead content of bovine liver. Liver is known to accumulate lead and cadmium.
Vegetables and fruits contribute substantially (33–38%) to the total intake of lead in all the three countries. Vegetables are exposed to lead contamination from exhaust fumes of cars. Although the consumption of these food groups is almost similar in Finland and West Germany, the lead intake through vegetables and fruits is three times higher in West Germany than in Finland (70 and 22 μg/ d respectively). This illustrates the effect of environmental contamination.
On the other hand a part of the differences in the intake of lead and cadmium is explicable by food consumption habits. Fish and seaweed cause remarkable exposure in Japan, because they are an essential part of Japanese diet. Germans consume much beer and this is reflected in the intake of lead. Finns prefer milk and cultured milk instead, which contain ten times less lead than the German beer.
A. M. FEHILY, B. K. BUTLAND, R. M. HOLLIDAY and J. W. G. YARNELL
M. R. C. Epidemiology Unit (South Wales), 4 Richmond Road, Cardiff, CF2 3AS.
The Caerphilly Heart Disease Survey comprises a whole programme of studies based in Caerphilly, South Wales (The Caerphilly and Speedwell Collaborative Group, 1984). A considerable emphasis is being placed on the investigation of dietary factors of possible importance. Every subject was asked to complete a food frequency questionnaire and a representative subsample (665) completed a seven day weighed dietary record. Preliminary results of the investigation of dietary determinants of plasma lipoproteins, fibrinogen and blood pressure using weighed intake data from 117 men were presented (Fehily et al., 1982). In addition, nutrient intakes calculated from questionnaire data were compared with those from weighed intake data. Since the preliminary report on 119 subjects (Yarnell et al., 1983), the nutrient calculations have been altered to include all foods on the questionnaire which contain a particular nutrient, rather than just the main sources. This has reduced the mean differences between the two methods (see Table). These differences are now small but, except for fat, starch and ascorbic acid, still statistically significant. The correlation coefficients and percentage of subjects classified in the same and in opposite thirds of the distribution were similar to those reported previously.
Table. Comparison of nutrient intakes calculated from questionnaire with those from 7 day weighed intake records of 665 men
| Nutrient (g/d) | Questionnaire | Weighed intake | ||
| Mean | (SD) | Mean | (SD) | |
| Energy | 2300 | (592) | 2405 | (588)* |
| Protein | 73 | (18) | 82 | (19)* |
| Fat | 100 | (30) | 99 | (28) |
| Starch | 156 | (58) | 160 | (51) |
| Sugars | 95 | (39) | 111 | (56)* |
| Cereal fibre | 8 | (5) | 9 | (5)* |
| Fruit and vegetable fibre | 8 | (3) | 11 | (4)* |
| Ascorbic acid (mg/d) | 51 | (21) | 52 | (29) |
| Alcohol | 29 | (41) | 26 | (33)* |
The Caerphilly and Speedwell Collaborative group (1984): Caerphilly and Speedwell Collaborative heart disease studies. J. Epidemiol. Commun. Health 38, 259.
Fehily, A. M., Milbank, J. E., Yarnell, J. W. G., Hayes, T. M., Kubiki, A. J. and Easthem, R. D. (1982): Dietary determinants of lipoproteins, total cholesterol, viscosity, fibrinogen and blood pressure. Am. J. Clin. Nutr. 36, 890.
Yarnell, J. W. G., Fehily, A. M., Milbank, J. E., Sweetnam, P. M. and Walker, C. L. (1983): A short dietary questionnaire for use in an epidemiological survey: comparison with weighed intake records. Hum. Nutr.: Appl. Nutr. 37A, 103.
H. ISAKSON
National Food Administration, Box 622, S-751 26 Uppsala, Sweden.
It is known that food treatments, including food preparation, storage, preservation, industrial processing, transport and even “no-action”, may affect the nutrient content of foods. It is also known that even bioavailability of certain foods is affected in the presence of other foods and substances, or when foods are treated in different ways. It is in the nutritionists' interest to ‘optimize’ food treatment, and find methods which destroy less nutrients and/or eliminate most toxic substances contained in foods. The food industry is also interested in optimizing industrial processes and researchers need more and better information on foods, nutrients and how these are affected by treatment or by the presence of other foods or substances. The model presented here is an attempt to structure data in a suitable way so that it can be available for all user groups who need information on nutrient contents and modifications, recommendations, bibliographical references, etc. The goal is an ‘infological model’, based on a logical description of data from the point of view of the users. The concepts of ‘objects’, ‘properties’ and ‘relations’ so common in datamodelling, were used, and all the roles of Relational Algebra apply here. Software to manage tables according to these rules is available in the form of Relational Database Management systems.
An experiment, according to the model, covers one or more foods F1, F2… Fi… (in a certain stage s), each one containing nutrients N1, N2, Nj, affected by a treatment T, which is defined as a sequence of processes, P1, P2, Pj… where each stage s (process Pj) is defined by physical, chemical and biological parameters.
As a result of every stage (process P) of treatment T, acting on food (i) in status (s), the food is transformed becoming a new food:
and the nutrient contents may also be modified, so the nutrient loss or gain is the difference:
Extensions of the basic model and validation. Other objects and relations may also be defined and represented by tables. The model covers food groups, recommendations, references and literature searching, losses and gains in different conditions, bioavailability, dishes and recipes; it may be extended to cover additives and other non-nutrients. It was constructed and validated at the NFA by answering some 40 questions in the fields mentioned above. The language MIMER/QL was used in the project.
S. BINGHAM, P. R. MURGATROYD and J. H. CUMMINGS
Medical Research Council, Dunn Clinical Nutrition Centre, 100 Tennis Court Road, Cambridge, CB2 1QL.
Conventional methods of assessing dietary intake are difficult. In the first place subjects have to weigh and keep accurate records of everything they eat. This is a problem when the subjects under review are unfamiliar with measuring equipment, and they may find accurate record keeping not just onerous, but impossible. It is time-consuming, and the tedious nature of the task may often lead to forgetfulness and incomplete records. In addition, subjects have to literate.
To overcome these problems, we have developed a new Portable Electronic Tape Recording Automated (PETRA) Scale. To use the scale subjects simply press a button and describe each item of food as it is served on to a plate. The description and a simultaneous record of the weight is recorded on a tape cassette. The system is simple and quick to use, extremely accurate (weights are recorded to the nearest lg), robust, and fully portable. The weight data is concealed from the subject and cannot be removed during the recording period.
To validate (Bingham & Cummings, 1984) the PETRA system 17 free-living healthy students and volunteers, aged 20–40 years used PETRA for 14 days and collected 24 hourly urine samples for nitrogen (N) determination. The completeness of the 24 hourly urine samples was verified by the PABA check test (Bingham and Cummings, 1983). The results were compared with records kept by 10 other healthy volunteers using conventional spring balances. Mean urine N was 78 ± 9% (range 60–92%) of dietary intake, which was not significantly different from that of 79 ± 8% (range 63–89%) with conventional weighed records. PETRA is available from Cherlyn Electronics, 22 High Street, Histon, Cambs., U.K.
Bingham, S. A., Cummings, J. H. (1983): The use of 4-amino benzoic acid as a marker to validate the completeness of 24th urine collections in man. Clin. Sci. 64, 629–635.
Bingham, S. A., Cummings, J. H. (1984): Urine 24th nitrogen excretion as an independent measure of the habitual dietary protein intake in individuals. Proc. Nut. Soc. 43, 80A.
A. SCHUBERT, J. M. HOLDEN and W. R. WOLF
Nutrient Composition Laboratory — Building 161, Beltsville Human Nutrition Research Center, BARC-East, ARS, USDA, Beltsville, MD 20705, USA.
References published since 1960 which report analyses of selenium (Se) in foods were collected and evaluated according to five criteria: Number of Samples, Analytical Method, Sample Handling, Sampling Plan, and Analytical Quality Control. Data were grouped by food item and rated according to the criteria requirements which had been developed specifically for evaluating the quality of Se data. Ratings assigned to the data from each study yielded a Quality Index, indicating which data would be included in the calculation of the mean Se value for each food item. The Quality Indices for acceptable studies were summed to determine a Confidence Code intended to indicate the relative degree of confidence the user can have in each mean Se value. The selection of Se core foods was based on their ranked Se contribution to U.S. diets. Foods were ranked by multiplying Se contribution by the amount consumed by the 28 030 individuals who provided 3-day dietary intake data in the U.S. Department of Agriculture's 1977–78 Nationwide Food Consumption Survey. Mean, minimum, and maximum Se values, Confidence Codes, ranks, and references have been compiled for 111 food items. The five most highly ranked food aggregates (beef, white bread, eggs, pork, and chicken) provided half of the Se accounted for in the diets of the survey respondents.
A. E. BLACK, A. M. PRENTICE, W. A. COWARD
Dunn Nutrition Unit, Downham's Lane, Milton Road, Cambridge CB4 1XJ.
The recently developed double-labelled water (stable isotope) technique of measuring total energy expenditure has been used to validate the 7-day weighed diet record as a measure of energy requirements (Coward et al., 1984).
The following were measured in ten normal-weight (mean ideal body weight [IBW] 99%) and eight obese (mean IBW 159%) women: (a) total energy expenditure (TEE) over 14 days (normal weight) or 28 days (obese) women; b) one (normal weight) or two (obese) 7-day weighed diet records (DIET); c) body weight and total body water (TBW) at both ends of the study period to assess changes in energy stores (body fat).
In five normal-weight subjects the agreement between DIET and TEE was within 10 per cent (see Table 1). In two subjects DIET overestimated TEE by 13 and 14% and in three subjects DIET underestimated TEE by 16, 22 and 28%. Allowing for changes in body composition did not improve the agreement; there are however many problems in assessing changes in body fat over such a short period of time. The discrepancy of 28% between DIET and TEE was in a subject who was mildly overweight (115% IBW) and a member of a slimming club. This tends to confirm suspicions that chronic ‘dieters’ will diet while keeping food records. With the limited data available so far, there is no indication in the remaining subjects of a constant bias in the weighed intake technique.
Table 1. Normal-weight women.
| Subject | Diet | Tee |  | Energy from stores (EFS) | Diet + EFS |  | |
| (kcal) | (kcal) | (kcal) | (kcal) | ||||
| 1 | 1807 | 1848 | 0.98 | 0 | 1807 | 0.98 | |
| 2 | 2390 | 2461 | 0.97 | 0 | 2390 | 0.97 | |
| 3 | 2024 | 1791 | 1.13 | + | 61 | 2085 | 1.16 |
| 4 | 1842 | 1693 | 1.09 | - | 215 | 1627 | 0.96 |
| 5 | 1673 | 1466 | 1.14 | + | 133 | 1808 | 1.23 |
| 6 | 1421 | 1805 | 0.78 | - | 62 | 1352 | 0.75 |
| 7 | 1301 | 1807 | 0.72 | NA | NA | NA | |
| 8 | 2516 | 3009 | 0.84 | - | 167 | 2349 | 0.78 |
| 9 | 2311 | 2490 | 0.93 | + | 171 | 2482 | 1.00 |
| 10 | 2608 | 2468 | 1.06 | - | 295 | 2313 | 0.94 |
| Mean | 1989 | 2084 | 0.96 | - | 41 | 2024 | 0.97 |
| ± s.d. | ±457 | ±488 | ±0.13 | ±394 | ±0.15 | ||
In obese subjects DIET only accounted for 67% of TEE (see Table 2). In seven out of eight subjects, energy was taken from fat stores, indicating that they were dieting while recording intake. However DIET + Energy from stores still only accounted for 84% of TEE indicating that subjects either are more on days when not recording intake and/or under-recorded actual food intake on recording days.
Table 2. Obese women.
| Subject | | Energy from stores (EFS) (kcal) | | ||
| 11 | 0.96 | + | 402 | 1.16 | |
| 12 | 0.40 | + | 654 | 0.71 | |
| 13 | 0.71 | + | 240 | 0.79 | |
| 14 | 0.64 | - | 107 | 0.60 | |
| 15 | 0.49 | + | 854 | 0.85 | |
| 16 | 0.79 | + | 53 | 0.81 | |
| 17 | 0.83 | + | 568 | 1.06 | |
| 18 | 0.52 | + | 418 | 0.72 | |
| Mean | 0.67 | + | 385 | 0.84 | |
| ± s.d. | ±0.19 | ±0.19 | |||
Coward, W. A., Prentice, A. M., Murgatroyd, P. R., Davies, H. L., Cole, T. J., Sawyer, M., Goldberg, G. R., Haliday, D. and Macnamara, J. P. (1984): Euro/Nut Workshop, Hum. Energy Metab. Wageningen.
