Introduction to Human Nutrition 2nd Edition ( PDFDrive )

Table 9.21 Absorption, transport and storage characteristics of the ultratrace elements Element Major mechanism(s) for Means of absorption Aluminum homeostasis Uncertain; some evidence for passive diffusion throug Arsenic Absorption the paracellular pathway; also, evidence for active Boron absorption through processes shared with active Urinary excretion; inorganic processes of calcium; probably occurs in proximal arsenic as mostly duodenum; citrate combined with aluminum enha dimethylarsinic acid and absorption organic arsenic as mostly arsenobetaine Inorganic arsenate becomes sequestered in or on mucosal tissue, then absorption involves a simple Urinary excretion movement down a concentration gradient; organic arsenic absorbed mainly by simple diffusion throug lipid regions of the intestinal boundary Ingested boron is converted into B(OH)3 and absorbe this form, probably by passive diffusion Cadmium Absorption May share a common absorption mechanism with ot Germanium metals (e.g. zinc) but mechanism is less efficient fo Lead Urinary excretion cadmium Absorption Lithium Has not been conclusively determined but probably i Nickel Urinary excretion passive diffusion Both absorption and urinary Uncertain; thought to be by passive diffusion in smal excretion intestine but evidence has been presented for an active transport perhaps involving the system for calcium Passive diffusion by paracellular transport via the tigh junctions and pericellular spaces Uncertain, evidence for both passive diffusion (perha as an amino acid or other low molecular weight complex) and energy-driven transport; occurs in th small intestine Rubidium Excretion through kidney Resembles potassium in its pattern of absorption; and intestine rubidium and potassium thought to share a transp Silicon system Tin Both absorption and urinary Vanadium excretion Mechanisms involved in intestinal absorption have no been described Absorption Mechanisms involved in intestinal absorption have no Absorption been described Vanadate has been suggested to be absorbed throug phosphate or other anion transport systems; vanad has been suggested to use iron transport systems; absorption occurs in the duodenum Reproduced from Nielsen (1999) in Sadler et al. Encyclopaedia of Human Nutrition, copyright 1999 w

gh Percentage of ingested absorbed Transport and storage vehicles e <1% Transferrin carries aluminum in plasma; bone ances Soluble inorganic forms, >90%; a possible storage site slightly soluble inorganic forms, c 20–30%; inorganic forms with Before excretion inorganic arsenic is gh foods, 60–75%; methylated converted into monomethylarsonic acid ed in forms, 45–90% and dimethylarsinic acid; arsenobetaine not biotransformed; arsenocholine ther >90% transformed to arsenobetaine or is by 5% Boron transported through the body as ll undissociated B(OH)3; bone a possible >90% storage site Adults 5–15%, children 40–50% Incorporated into metallothionein which probably is both a storage and transport vehicle None identified Bone is a repository for lead ht Lithium chloride highly absorbed: Bone can serve as a store for lithium Minerals and Trace Elements 233 aps >90% he Transported in blood principally bound to <10% with food serum albumin with small amounts bound port to L-histidine and α2-macroglobulin; no ot Highly absorbed organ accumulates physiological amounts ot of nickel Food silicon near 50%; insoluble gh or poorly soluble silicates ~1% None identified dyl ; ~3%; percentage increases when Silicon in plasma believed to exist as very low amounts are ingested undisassociated monomeric silicic acid <10% None identified; bone might be a repository Converted into vanadyltransferrin and vanadyl-ferritin; whether transferrin is the transport vehicle and ferritin is the storage vehicle for vanadium remains to be determined; bone is a repository for excess vanadium with permission of Elsevier.

Table 9.22 Excretion, retention, and possible biological roles of the ultratrace elements Element Organs of high content Major excretory route af ter Molecules of bi Aluminum (typical concentration) ingestion Arsenic Aluminum bind Bone (1–12 μg/g) Urine; also significant phospholipid Boron Lung (35 μg/g) amounts in bile is a transpor Bromine Hair (0.65 μg/g) Urine Methylation of Cadmium Nails (0.35 μg/g) organisms ra Germanium Skin (0.10 μg/g) methylated e Lead arsenobetain Lithium Bone (1.6 μg/g) Urine acid; arsenite Fingernails (15 μg/g) monomethyl Hair (1 μg/g) Urine adenosylmet Teeth (5 μg/g) Urine and gastrointestinal Boron biochem Hair (30 μg/g) tract forms ester c Liver (40 μg/g) preferably th Lung (6.0 μg/g) Urine compounds; Testis (5.0 μg/g) antibiotics) s Kidney (14 μg/g) Urine; also significant characterized Liver (4 μg/g) amounts in bile Bone (9 μg/g) Exists as Br− ion Liver (0.3 μg/g) Urine acids Pancreas (0.2 μg/g) Testis (0.5 μg/g) Metallothionein Aorta (1–2 μg/g) involved in r Bone (25 μg/g) Kidney (1–2 μg/g) None identified Liver (1–2 μg/g) Adrenal gland (60 ng/g) Plasma lead mo Bone (100 ng/g) mostly to he Lymph nodes (200 ng/g) molecular w Pituitary gland (135 ng/g) None identified

iological importance Possible biological role 234 Introduction to Human Nutrition Enzyme activator ds to proteins, nucleotides and ds; aluminum-bound transferrin apparently Metabolism of methionine, or involved in rt molecule labile methyl metabolism; regulation of gene expression inorganic oxyarsenic anions occurs in anging from microbial to mammalian; Cell membrane function or stability such that it end-products include arsenocholine, influences the response to hormone action, ne, dimethylarsinic acid and methylarsonic transmembrane signaling or transmembrane e methyltransferase and movement of regulatory cations or anions larsonic acid methyltransferase use S- thionine for the methyl donor Electrolyte balance mistry essentially that of boric acid, which complexes with hydroxyl groups, hose adjacent and cis, in organic five naturally occurring boron esters (all synthesized by various bacteria have been d n in vivo, binds to proteins and amino n, a high sulfhydryl-containing protein Involved in metallathionein metabolism and regulating cadmium distribution utilization d Role in immune function ostly bound to albumin; blood lead binds Facilitates iron absorption and/or utilization emoglobin but some binds a low Regulation of some endocrine function weight protein in erythrocytes d

Nickel Adrenal glands (25 ng/g) Urine as low molecular Binding of Ni2− Bone (33 ng/g) weight complexes (especially h Rubidium Kidney (10 ng/g) albumin) and Silicon Thyroid (30 ng/g) Urine; also significant important in Tin amounts excreted through of urease; N Vanadium Brain (4 μg/g) intestinal tract desulfurizatio Kidney (5 μg/g) anaerobic m Liver (6.5 μg/g) Urine Testis (20 μg/g) None identified Aorta (16 μg/g) Urine; also significant Bone (18 μg/g) amounts in bile Silicic acid (SiO Skin (4 μg/g) plasma; mag Tendon (12 μg/g) Urine; also significant of silicon in Bone (0.8 μg/g) amounts in bile never been r Kidney (0.2 μg/g) Liver (0.4 μg/g) Sn2+ is absorbe Bone (120 ng/g) Kidney (120 ng/g) Vanadyl (VO2+), Liver (120 ng/g) peroxovanad Spleen (120 ng/g) especially th Testis (200 ng/g) hemoglobin) Reproduced from Nielsen (1999) in Sadler et al. Encyclopaedia of Human Nutrition, copyright 19 None of the suggested biological functions or roles of any of the ultratrace elements has been c

− by various ligands including amino acids Cofactor or structural component in specific histidine and cysteine), proteins (especially metalloenzymes; role in a metabolic pathway involving vitamin B12 and folic d a macroglobulin called nickeloplasmin acid; role similar to potassium; n transport and excretion; Ni2+ component neurophysiological function Ni3+ essential for enzymic hydrogenation, Role similar to potassium; neurophysiological on and carboxylation reactions in mostly function microorganisms d OH4) is the form believed to exist in Structural role in some mucopolysaccharides or gnesium orthosilicate is probably the form collagen; role in the initiation of calcification urine. The bound form of silicon has and in collagen formation rigorously identified ed and excreted more readily than Sn4+ Role in some redox reactions , vanaclate (H2VO−4 or VO−3) and Lower forms of life have haloperoxiclases that dyl [V-OO]; VO2+ complexes with proteins, require vanadium for activity; a similar role may exist in higher forms of life hose associated with iron (e.g. transferrin, Minerals and Trace Elements 235 ) 999 with permission of Elsevier. conclusively or unequivocally identified in higher forms of life.

236 Introduction to Human Nutrition Table 9.23 Human body content and deficient, typical, and rich sources of intakes of ultratrace elements Apparent deficient intake Typical human daily (species) Element Human body content dietary intake Rich sources Aluminum 160 μg/kg (goat) 30–50 mg 2–10 mg Baked goods prepared with chemical leavening Arsenic <25 μg/kg (chicks) 1–2 mg 12–60 μg agents (e.g. baking powder), processed <35 μg/kg (goat) cheese, grains, vegetables, herbs, tea, Boron <15 μg/kg (hamster) 10–20 mg antacids, buffered analgesics <30 μg/kg (rat) Bromine <0.3 mg/kg (chick) 200–350 mg Shellfish, fish, grain, cereal products Cadmium 0.25–0.35 mg/day (human) 5–20 mg Germanium <0.3 mg/kg (rat) 0.5–3.5 mg Food and drink of plant origin, especially Lead 0.8 mg/kg (goat) 3 mg noncitrus fruits, leafy vegetables, nuts, <5 μg/kg (goat) Children less than 10 2–8 mg pulses, legumes, wine, cider, beer Lithium <4 μg/kg (rat) 10–20 μg 0.7 mg/kg (rat) years old 2 mg, 0.4–3.4 mg Grain, nuts, fish Nickel <32 μg/kg (pig) Adults 120 mg 15–100 μg Shellfish, grains, especially those grown on Rubidium <45 μg/kg (rat) 350 μg Silicon 200–600 μg high-cadmium soils, leafy vegetables Tin <1.5 mg/kg (goat) 1–2 mg Wheat bran, vegetables, leguminous seeds Vanadium <15 μg/kg (rat) 360 mg 70–260 μg Seafood, plant foodstuffs grown under high- 1–5 mg <100 μg/kg (goat) 2–3 g 20–50 mg lead conditions <20 μg/kg (rat) 7–14 mg 1–40 mg 180 μg/kg (goat) 100 μg 10–30 μg Eggs, meat, processed meat, fish, milk, milk products, potatoes, vegetables (content <20 mg kg (chick) varies with geological origin) <4.5 mg/kg (rat) <20 μg/kg (rat) Chocolate, nuts, dried beans and peas, grains <10 μg/kg (goat) Coffee, black tea, fruits and vegetables (especially asparagus), poultry, fish Unrefined grains of high fiber content, cereal products Canned foods Shellfish, mushrooms, parsley, dill, seed, black pepper, some prepared foods Reproduced from Nielsen (1999) in Sadler et al. Encyclopaedia of Human Nutrition, copyright 1999 with permission of Elsevier.. hypertension or diabetes. However, for many nutri- status that merit further evaluation. The judicious ents there are as yet no functional indicators that application of methods in molecular biology respond to dietary intake and, in such cases, nutri- (including genomics and proteomics) and nonin- ent requirements are established using more tradi- vasive imaging techniques is likely to provide new tional approaches, such as balance data. The lack of breakthroughs and rapid advances in the nutrition functional markers of mineral and trace element and biology of trace elements status is a significant disadvantage for studies relat- ● evaluate further the specific health risks associated ing their intake or status to health outcomes such with marginal deficiencies of various minerals and as hypertension, cardiovascular disease, osteoporo- trace elements. There is a need to determine reliable sis, diabetes, and other disorders. For example, relationships between mineral status and disease widely used biochemical indicators of essential and then to demonstrate that the incidence or trace element status generally lack both the sensi- severity of specific diseases is reversible by repletion tivity and the specificity that are required to define of mineral status. The development and validation optimal intake at various stages of the life cycle. of reliable assessment tools and functional markers Recent efforts have provided a number of potential of mineral status are the utmost priority for this “sensors” of cellular copper, zinc, and manganese field

Minerals and Trace Elements 237 ● define the adverse effects of acute and chronic high and Nutrients in the UK. HMSO, Department of Health, London, intakes of some minerals and trace elements. 1991. Interest in mineral fortification of foods is higher Hallberg L, Sandstrom B, Aggett PJ. In: JS Garrow, WPT James, A than ever before. Governments worldwide are Ralph, eds. Human Nutrition and Dietetics, 9th edn. Churchill increasingly tackling the common deficiencies of Livingstone, London, 1993. iron and iodine by adding these minerals to widely Holland B, Welch AA, Unwin ID, Buss DH, Paul AA, consumed staple foods such as cereal flours, sugar, Southgate DAT, eds. McCance & Widdowson’s The or soy sauce. The food industry in industrialized Composition of Foods, 5th edn. Royal Society of Chemistry and countries is manufacturing an increasing number Ministry of Agriculture, Fisheries and Food. London: HMSO, of functional foods designed to provide the con- 1995. sumer with protection against diseases of major Institute of Medicine (USA). Dietary Reference Intakes for Calcium, public health significance, such as osteoporosis, Phosphorus, Magnesium, Vitamin D, and Fluoride. National cancer, and heart disease, and fortified with miner- Academy Press, Washington DC, 1997. als such as calcium, selenium, zinc, magnesium, and Institute of Medicine (USA). Dietary Reference Intakes for Vitamin copper. The same minerals are added to dietetic A, Vitamin K, Arsenic, Boron, Chromium, Copper, Iodine, Iron, products, including infant foods, foods for preg- Manganese, Molybdenum, Nickel, Silicon, Vanadium, and Zinc. nant and lactating women, and enteral feeds for National Academy Press, Washington DC, 2001. hospital patients, all designed to cover the nutri- Mills CF, ed. Zinc in Human Biology. Springer-Verlag, London, tional requirements of specific consumers. This 1989. raises the issues not only of the possible health ben- Nielsen F. In: MJ Sadler, JJ Strain, B Caballero, eds. Encyclopedia of efits of fortification but also of possible toxicity. Human Nutrition. Academic Press, London, 1999. Therefore, there is an ever-increasing emphasis Rayman MP. The importance of selenium to human health. 2000; placed on upper safe levels of mineral intake and Lancet 356: 233–241. on fortification legislation Reilly C. Selenium in Food and Health. Blackie, London, 1996. Reilly C. The Nutritional Trace Metals. Blackwell Publishing, ● elucidate the impact of single nucleotide polymor- Oxford, 2006 phisms in the human genome on mineral and trace Sánchez-Castillo CP, James WPT. In: MJ Sadler, JJ Strain & B element dietary requirements. The key to future Caballero, eds. Encyclopedia of Human Nutrition. Academic applications of the DNA polymorphisms will be to Press, London, 1999. mine the human genome for DNA sequence infor- Strain JJ, Benzie IFF. In: MJ Sadler, JJ Strain & B Caballero, eds. mation that can be used to define biovariation in Encyclopedia of Human Nutrition. London: Academic Press, nutrient absorption and use. However, before this 1999. can be accomplished, a vast amount of nutritional biology research is needed to correlate gene poly- Further reading morphism with nutritional outcomes. Bowman B, Russel R, eds. Present Knowledge in Nutrition, 8th edn. References ILSI Press, Washington DC, 2001. Committee on Medical Aspects of Food Policy. Report on Health Passmore R, Eastwood MA, eds. Davidsons and Passmore Human and Social Subjects 41. Dietary Reference Values for Food Energy Nutrition and Dietetics, 8th edn. Churchill Livingstone, London, 1986. Optimal Nutrition Symposium. A series of papers. Proceedings of the Nutrition Society 1999; 58: 395–512. Sadler MJ, Strain JJ, Caballero B, eds. Encyclopedia of Human Nutrition, Vol. 3, Parts 1–3. Academic Press, London, 1999. Website Online Mendelian Genetics in Man (OMIM) website at the National Institute for Biotechnology Information: http://www. ncbi.nlm.nih.gov/Omim

10 Measuring Food Intake Una E MacIntyre Key messages • To increase our understanding of the error associated with mea- surements of food intake it is also necessary to develop and use • Measuring the food intake in free-living individuals is a complex physiological and biochemical markers of food intake. task. • To evaluate food intake data effectively it is important to collect • All measurements of food intake are subject to sources of error. sufficient additional data to allow individuals to be identified not • The dietary assessment method used depends on the purpose of only by age and gender, but also by body mass index, physical activity, and supplement use. the study. • The existence of error means that it is always important to be aware of and, whenever possible, to assess the nature and magnitude of the error. 10.1 Introduction Assessment of nutritional status The purpose of this chapter is to describe the various Nutritional health is maintained by a state of equilib- ways in which one can determine what people eat. The rium in which nutrient intake is balanced by nutri- task may be to find out about the national food supply, tional requirements. Malnutrition occurs when net the usual intake of a group or a household, or the nutrient intake is less than requirements (undernutri- intake of a given individual over a specified period. tion) or exceeds requirements (overnutrition). Both under- and overnutrition lead to metabolic changes The many reasons for finding out about the food which have acute and chronic consequences for that people eat fall into three broad categories: health. 1 Public Health: to evaluate the adequacy and safety There is no ideal tool to measure a person’s nutri- of the food that people eat at national or commu- tional status accurately. Attempts to predict the influ- nity level and to identify the need for or to evaluate ence of malnutrition based on single measurements nutrition-based intervention programs. fail to consider the many interacting factors between nutrition and disease state. For this reason, it is 2 Clinical: to assist with the prevention, diagnosis, necessary to look at several different measurements and treatment of diet-related conditions. in order to assess a person’s nutritional status. This process is known as the A, B, C, D of nutritional 3 Research: to study the interrelationships between assessment: food intake and physiological function or disease conditions under controlled conditions or in field ● Anthropometry (discussed in detail in Chapter 2) conditions. ● Biochemical and hematological variables ● Clinical and physical assessment The kind and amount of food intake data required ● Dietary intake. differ in each situation and may require data at the national, community, household, or individual level. © 2009 UE MacIntyre.

Measuring Food Intake 239 The rest of this chapter will concentrate on the advantages, and disadvantages. It is, therefore, measurement of food (dietary) intake, but more essential that the purpose of collecting dietary data detailed descriptions of the assessment of nutritional is clearly defined, so that the most appropriate status, at the individual and population level, can be dietary assessment method is used. found in the Public Health Nutrition and Clinical Nutrition textbooks in this series. It is also essential to recognize that finding out what people eat requires adequate resources. Appropriately Obtaining data on food intake is probably the most trained personnel must be employed not only for the difficult aspect of nutritional assessment and is asso- period of data collection but also for the time it takes ciated with several problems: to review, enter, and analyze the data. It may not always be necessary to obtain detailed data on food ● “Food intake” is not a simple measure of one vari- intake in order to answer a particular question. When able, such as weight or height, but requires data on resources are limited it is probably more useful to the intake of many different food items. collect limited data of high quality than to attempt to collect comprehensive dietary data with inadequate ● Food intake data are subject to many sources of resources. Being able to recognize this situation is variability, since even the same individuals eat dif- important for maximizing available resources. Table ferent foods, at different times, in different places, 10.1 lists the different approaches to measurement of in many different combinations, and with many food intake that are described in this chapter. different preparation methods. The net effect of all these sources of variability is that more data are Finally, it is important that the interpretation and needed to generate reliable results than would be application of data derived from food intake studies the case with a less variable measure. take into account the limitations of the data. This clearly does not improve the quality of the data per ● We are rarely in a position to know the truth about se, but maximizes their usefulness for the purpose for food intake. With many biological measurements it which they were originally collected. Recognition of is possible to check the results obtained against a the limitations of dietary data involves more than reference method that is known to give accurate simply stating the limitations. External comparisons results or by means of an independent measure. For to check whether the data are consistent with inde- example, we can check an infant’s birth weight by pendent sources of information on food intake and checking the accuracy of the weighing equipment to determine the likely direction and magnitude of used to measure it by means of a standard weight any bias are an integral part of the interpretation of or, if the information was obtained by means of a dietary data. Relevant sources of comparative infor- questionnaire, we may be able to check the data mation may include food supply and expenditure from official records. With food intake data we have data and physiological or biochemical measures to rely on the individuals who eat the food to related to nutrient intake. provide us with the answers to our questions. We ask individuals to remember what and how much 10.2 Indirect measurement of they ate, to estimate how often they eat particular food intake foods, or even, in some situations, to weigh or measure their food intake for a number of days. For Indirect measurements of food intake make use of this reason one of the most important consider- information on the availability of food at national, ations, when obtaining information on food intake regional, or household levels to estimate food intakes, from individuals, is to take all possible steps to rather than using information obtained directly from obtain their full cooperation. It is also extremely individuals who consume the food. Indirect methods important that individuals understand the purpose are most useful at the population and household of the process and what is expected of them. This levels for determining the amount and types of may well involve much time and effort on the part foods: of the investigator(s), but is essential for high- quality data. ● available for consumption at national level (commodity-level food supply data) ● There are a number of different methods to obtain dietary intake data. Each method has its purposes,

240 Introduction to Human Nutrition Table 10.1 Approaches to the measurement of food intake in population groups, households, and individuals Type and nature of data Name of method Used for assessing differences between Commodity-level food supply data, e.g., production, Food balance sheets Countries and regions of the world imports, and exports (indirect) Food disappearance data Country, locality, and season Product-level food supply data, e.g., retail and wholesale sales data (indirect) Household expenditure surveys Country, locality, season, and type of household Household food expenditure, e.g., money spent on food Household budget surveys (indirect) Household food account Country, locality, season, and type of Household food procurement household Household food acquisition, e.g., amount of food Household food inventory entering the household (indirect) Household food records Country, locality, season, and type of household Household food consumption (direct) Geographical, seasonal, and demographic Qualitative record of foods (but not amounts) eaten Menu records subgroups and individuals over the course of 1–7 days by individuals (direct) Weighed records and records Geographical, seasonal, and demographic Quantitative record of food intake, e.g., record of foods estimated in household measures subgroups and individuals eaten over the course of 1–7 days by individuals Food frequency questionnaires (direct) Geographical, seasonal, and demographic Single or multiple 24 hour recalls subgroups and individuals Qualitative or semiquantitative recall, usually of a specified list of foods, eaten in the previous month Diet history Geographical, seasonal, and demographic or year by individuals (direct) subgroups and individuals (if multiple recalls obtained) Quantitative recall of foods eaten on the previous day, usually obtained from individuals by interview Temporal and demographic subgroups and (direct) individuals Quantitative recall of habitual intake in the immediate past obtained from individuals by interview (direct) ● traded at wholesale or retail levels (product-level human consumption by relevant population statis- food supply data) tics. An analysis of the energy, protein, and fat pro- vided by the food item may also be given. ● purchased at household level (household-based budget/expenditure data). The Food and Agriculture Organization (FAO) has compiled and published food balance sheet data for Commodity level food supply data most countries in the world since 1949. Regularly updated food balance sheet data are available online Food supply data are usually produced at national at www.fao.org for most countries for about 100 level from compilations of data from multiple sources. primary crop, livestock, and fishery commodities and The primary sources of data are records of agricul- some products such as sugar, oils, and fats derived tural production and food exports and imports from them. Table 10.2 shows an extract from the food adjusted for changes in stocks and for agricultural balance sheet for the Southern African Development and industrial use of food crops and food products. Community for 2003. National food supply data are usually referred to as “food balance sheets” or as “apparent consumption The accuracy of food balance sheets and apparent data.” Food balance sheets give the total production consumption data depends on the reliability of the and utilization of reported food items and show the basic statistics used to derive them, i.e., population, sources (production, stocks, and imports) and utiliza- supply, utilization, and food composition data. These tion (exports, industrial use, wastage, and human can vary markedly between countries not only in consumption) of food items available for human con- terms of coverage but also in terms of accuracy. sumption in a country for a given reference period. The amount of each food item is usually expressed Several internal and external consistency checks are per caput (per person) in grams or kilograms per year built into the preparation of the FAO food balance by dividing the total amount of food available for sheets, but users still need to evaluate the data for themselves in the context of the purpose for which

Table 10.2 Extract from the food balance sheet for the Southern African Development Community, 200 Items Production Import Stock Export Domestic Feed Seed quantity quantity variation quantity supply quantity quantity Cereals, (1000 (1000 (1000 (1000 (1000 (1000 (1000 excluding tonnes) tonnes) tonnes) tonnes) tonnes) tonnes) tonnes) beer 24862.65 8500.95 1847.63 1622.89 33588.34 4823.37 473.57 Wheat Rice (milled 2041.13 4256.29 183.54 381.07 6099.89 22.68 68.04 857.34 1432.74 70.97 38.06 2322.99 1.2 42.84 equivalent) Barley 277.4 355.47 18.64 20.39 631.12 18.17 6.54 Maize 19463.18 2271.36 1348.92 1164.19 21919.27 4485.34 309.62 Rye Oats 0.23 13.39 0 0.48 13.14 5.93 0.08 Millet 33.91 34.63 31.07 1.08 98.52 44.7 2.04 Sorghum 629.77 11.22 1.91 648.09 16.2 16.67 Cereals, other 1521.78 81.1 9 4.11 1779.27 216.22 27.69 37.9 44.75 180.5 11.61 76.05 12.92 0.07 5.01 Reproduced with permission of FAO (http://fao.org).

03 Waste Other uses Food Food/ Calories/ Proteins/ Fat/capita/ quantity quantity quantity capita/year capita/ capita/ day (g) Food (1000 (1000 (1000 (kg) day day (g) manufacture tonnes) tonnes) tonnes) 7.11 (1000 118.4 960.06 24.82 tonnes) 1317.2 356.34 24984.3 1634 19.05 105.43 2.47 5882.29 27.88 171.31 5.18 0.68 2.57 58.87 4.67 2213.09 10.49 103.23 1.98 0.15 559.87 10 26.33 10.22 0.05 0.34 0.01 0 400.67 961.91 322.87 15438.9 73.16 627.56 16.05 5.79 0 0 0.01 0 7.12 0.03 0.26 0.01 0.03 0 1.67 0 50.17 0.24 1.49 0.05 0.12 156.42 48.45 0 410.35 1.95 16.46 0.38 0.33 495.46 130.45 0 909.45 4.31 37.21 1.1 0.01 0 0.39 0 62.76 0.3 2.2 0.06

242 Introduction to Human Nutrition they are being used. One of the crucial factors in using Over a period of years, food balance sheets show the data appropriately is to understand the terminol- trends in national food supply and food consumption ogy used. Box 10.1 provides “in principle” definitions patterns. They may be used for population compari- of the key terms. sons such as comparing population estimates of fat intake with cardiovascular disease rates. Food balance sheets provide important data on food supply and availability in a country and show In practice, the data needed to compile food balance whether the food supply of the country as a whole is sheets are not always available and imputations or adequate for the nutritional needs of its population. estimates may have to be used at each stage in the calculation of per caput food and nutrient availability. Box 10.1 In most industrialized countries reliable data are usually available on primary commodities, but this Commodity coverage: all potentially edible commodities whether is not necessarily the case for the major processed used for human consumption or used for nonfood purposes. products. For example, data may be available on flour Exports: all movements out of the country during the reference but not on products such as bread and other cereal period. products made from flour that may have quite differ- Feed: the quantity of the commodity available for feeding livestock ent nutrient characteristics. The overall impact of and poultry. incomplete data will vary from country to country, Food quantity: the amounts of the commodity and any commod- but it has been suggested that in general underestima- ity derived therefrom available for human consumption during the tion of per caput availability of nutrients is more reference period, e.g., for maize includes maize, maize meal, and likely in less developed countries and overestimation any other products derived from maize that are available for in countries where most of the food supply is con- human consumption. sumed in the form of processed products. Imports: all movements of the commodity into the country, e.g., commercial trade, food aid, donated quantities, and estimates of It is also very important to keep in mind that food unrecorded trade. balance sheets show only data on foods available for Industrial uses: commodities used for manufacture for nonfood consumption, not the actual consumption of foods; purposes, e.g., oils for soap. nor do they show the distribution of foods within the Per caput supply: adjustments are made when possible to the population, for example among different regions resident population for temporary migrants, refugees, and tourists. or among different socioeconomic, age, and gender The figures represent only the average supply available for the groups within the population. Food balance sheets population as a whole and not what is actually consumed by indi- also do not provide information on seasonal varia- viduals. Many commodities are not consumed in the primary form tions in food supply. in which they are reported in food balance sheets. To take this into account the protein and fat content shown against primary com- Product-level food supply data modities are derived by applying appropriate food composition factors to the relevant amounts of processed foods, and not by In some countries (e.g., Canada and the USA) data multiplying the quantities shown in the food balance sheet by food on per caput food availability are prepared from composition factors relating to the primary commodity. information on raw and processed foods available at Production: total domestic production whether produced inside or the retail or wholesale level. Such data are derived outside the agricultural sector, i.e., includes noncommercial pro- mainly from food industry organizations and firms duction and production from home gardens. engaged in food production and marketing such Seed: quantity of the commodity set aside for sowing or plant- supermarkets. Errors arise mainly from inappropriate ing or any other form of reproduction for animal or human conversion factors for processing, the absence of data consumption. for some processed products, and the lack of data on Stock variation: changes in stocks occurring during the reference food obtained from noncommercial sources such as period at all levels between the production and retail levels, i.e., home gardens, fishing, and hunting. stocks held by the government, manufacturers, importers, export- ers, wholesale and retail merchants, and distributors. In practice, Commercial databases such as those produced by the available information often only relates to government stocks. the AC Neilsen Company and the electronic stock- Waste: commodities lost through all stages between postharvest control records from individual supermarkets, from production and the household, i.e., waste in processing, storage, which they are compiled, have the potential for moni- and transportation, but not domestic waste. Losses occurring toring national-, regional-, and local-level trends in during the manufacture of processed products are taken into account by means of extraction/conversion rates.

Measuring Food Intake 243 the food supply, at a product-specific level. Their groups, to identify high-risk groups for nutrition- principal disadvantage at present lies in the costs related conditions, to monitor trends in food patterns associated with processing or otherwise accessing, on over time, and for developing nutrition policy. a regular basis, the very large amounts of data that are A modification of the HBS, known as the list recall involved. method, includes quantities of items purchased, which strengthens the information. The information FAO food balance sheets and similar sources of available from HBSs, however, also needs to be con- information are primarily useful for formulating sidered in the context of their limitations. agricultural and health policy, for monitoring changes in national food supplies over time, and as a basis for ● Information provided by HBSs differs from country forecasting food consumption patterns. They can also to country both in the number of food items be used to make intercountry comparisons of food recorded and in the type of information collected. and nutrient supplies, provided that potential differ- ences in data coverage and accuracy are taken into ● Most surveys do not include expenditure informa- account. tion on food consumed outside the home. Household-based surveys ● Most surveys do not collect information on foods acquired by means other than purchase. For Household-based surveys determine the foods and example, food obtained as gifts, produced by the beverages available for consumption at family, house- household itself, or harvested from the wild. hold, or institutional levels. Some surveys such as household expenditure or household budget surveys ● Most surveys do not collect information on domes- determine the amount of money spent on food for a tic wastage, i.e., food given to pets, spoiled food and given period, while others, such as the food account, plate waste, or food provided for guests. food inventory, and food record methods, attempt to describe the food available and/or consumed by a ● It is often difficult to estimate the nutrient content household or institution. of the food available to the household because data are reported only at food group and not individual Household food expenditure surveys food level. Household food expenditure surveys determine the amount of money spent on food by a household over ● Differences in food coding systems between coun- a given period. Household food expenditure data can tries make it difficult to compare data between provide useful information for nutritionists on food countries. expenditure patterns of different types of households, but without quantitative information this cannot be Three conclusions emerge from this list. First, the translated into nutrient patterns. data obtained from HBSs are not necessarily compa- rable between countries. Second, most HBSs do not Household budget surveys (HBSs) are conducted collect all of the information needed to provide an at regular intervals in many European countries and accurate assessment of the total food supply available many also collect data on food quantities as well as at household level. Third, provided that the HBS cost. HBSs have several advantages: methodology remains consistent, HBSs can provide a great deal of valuable information about food pat- ● they are usually conducted at regular intervals of terns over time, in different sociodemographic groups, between 1 and 5 years and in different parts of the country, and how these relate to social, economic, and technological changes ● they are conducted on representative samples of in the food supply. households Household food account method ● the food supply information collected can be In the food account method, the household member classified by sociodemographic characteristics, responsible for the household’s food keeps a record geographical location, and season. of the types and amounts of all food entering the household including purchases, gifts, foods produced The large amount of nutrition-related information by the household itself such as from vegetable and collected by these surveys offers the potential to assess fruit gardens, foods obtained from the wild, or from the nutritional patterns of different population other sources. Amounts are usually recorded in retail

244 Introduction to Human Nutrition units (if applicable) or in household measures. be introduced by inaccurate memory or expression of Information may also be collected on brand names answers. and costs. The recording period is usually 1 week but may be as long as 4 weeks. Household food inventory method The food inventory method uses direct observation This method is used to obtain food selection pat- to describe all foods in the household on the day of terns from populations or subgroups within a popu- the survey. The investigator records the types and lation. It has the advantage of being fairly cost-effective amounts of foods present in a household, whether and is particularly useful for collecting data from raw, processed, or cooked, at the time of the study. large samples. It may also be repeated at different Information may also be collected on how and where times of the year to identify seasonal variations in food is stored. food procurement. A food inventory may be combined with the food The food account method does not measure food account to determine the changes of food stocks consumption, wastage, or other uses, nor does it during the survey period. It may also be used together account for foods consumed outside the home. It with a food procurement questionnaire to describe assumes that household food stocks stay constant the acquisition of foods in the household. This throughout the recording period, which may not nec- method is time-consuming for the investigator and essarily be the case. For example, food purchases may very intrusive for the respondent, but is useful when be done once a month and therefore stocks may be foods are procured by means other than purchase and depleted in the days preceding the purchase. It also when levels of food security in vulnerable households does not reflect the distribution of food within the need to be assessed. household and therefore cannot be used to determine food consumption by individuals within the house- Household food record hold. Since the method relies on the respondents All foods available for consumption by the household being literate and cooperative, bias may be introduced are weighed or estimated by household measures in populations with high levels of illiteracy. The prior to serving. Detailed information such as brand fact of having to record the acquisition may lead to names, ingredients, and preparation methods are respondents changing their procurement patterns also recorded over a specific period, usually 1 week. either to simplify recording or to impress the This method provides detailed information on the investigator. food consumption patterns of the household, but it is very time-consuming and intrusive and relies Household food procurement heavily on the cooperation of the household. As for the other household methods, it does not provide questionnaire/interview information on distribution of food within the house- A food procurement questionnaire or interview may hold or on individual consumption. When details of be used as an alternative method to the food account the household composition are given, estimates of method. In this method, the respondent indicates, individual intakes may be calculated. The method from a list of foods, which are used, where these are also does not determine foods eaten away from the obtained, the frequency of purchase, and the quanti- home nor does it take into account food eaten by ties acquired for a given period. The uses of the food guests to the home. procurement method are similar to those of the food account: to describe food acquisition patterns of 10.3 Direct measures of food intake populations or subpopulations. In contrast to the food account method, it does not require the respon- Information on food intake can be obtained directly dent to be literate as it may be administered as an from consumers in a number of different ways. Direct interview and it does not influence purchasing or measures are usually used to obtain food intake data other procurement patterns. from individuals but may also be used to obtain data from households. For example, in societies where it is The food procurement questionnaire/interview usual for members of the household to eat out of the does not provide information on actual food con- sumption or distribution within the household. As the method relies on recalled information, errors may

Measuring Food Intake 245 same pot this may be the only practical approach Portion size (g) × Frequency × because it does not disrupt the normal pattern of food Nutrient content per g intake. Unlike the indirect measures of food intake, direct measures provide sufficient information on and summed for all foods eaten by each individual food consumption to convert the food intake into during the study period. energy and nutrient intakes. Direct measurement of food intake can be divided Irrespective of the method used, the process of into two basic approaches: obtaining food intake information and converting this to energy and nutrient data is the same. The 1 reports of foods consumed on specified days: menu procedure for measuring food and nutrient intake records, weighed records, estimated records, and 24 involves five steps (Figure 10.1). hour recalls 1 obtaining a report of all the foods consumed by 2 reports of food intake over a period in the past, each individual which are used to construct typical food patterns: food frequency questionnaires and diet histories. 2 identifying these foods in sufficient detail to choose an appropriate item in the food tables Records are usually limited to fairly short periods, usually not longer than 7 consecutive days, while 3 quantifying the portion sizes recalls may relate to a single period of 24 hours or 4 measuring or estimating the frequency with which occasionally 48 hours. Diet histories and food fre- quency questionnaires relate to longer periods and each food is eaten their purpose is to obtain an assessment of habitual 5 calculation of nutrient intake from food tables. intake over the period in question and not a detailed day-to-day recall of what was eaten during that time. To convert the information on food intake into When records are kept portions may be either weighed nutrient intake, the nutrient content of each food eaten is calculated from food tables as: Basic procedure Type of dietary assessment Step 1 Report RECORD RECALL HISTORY FFQ Step 2 Identify food Researcher’s choice Subject’s choice from food tables from questionnaire Step 3 Quantify Weighed/ Estimated Average portion estimated or small/medium/large Step 4 Frequency As reported Subject's Subject’s choice estimate from questionnaire Step 5 Calculate intake Food code Food code x Portion weight x Portion weight x Nutrient content/g x Frequency x Nutrient content/g Figure 10.1 The five basic steps in a dietary assessment and the variations according to different methods. FFQ, food frequency questionnaire.

246 Introduction to Human Nutrition or estimated in terms of standard household mea- Box 10.2 sures. In recalls and questionnaires quantities always have to be estimated. Dimensions, photographs of Accuracy: the extent to which an estimate approximates the true foods, food models, and, sometimes, actual foods may value. be used to assist in this process. Each of these Bias: the extent by which an estimate differs from the true value. approaches has specific advantages and disadvantages Coefficient of variation: the standard deviation of a set of obser- and no single method of measuring food intake can vations expressed as a percentage of their mean. be regarded as the ideal method for all situations. Habitual intake: an estimate of the long-term average intake of an individual. Until recently, weighed intake recorded over a 7 day Precision: the extent to which a method is repeatable. Usually period was taken as the reference method against expressed in terms of the coefficient of variation, i.e., the standard which less detailed methods were compared. It has, deviation of the results of repeated analyses of the same para- however, been realized that this method has its limita- meter expressed as a percentage of their mean. tions and that it is not only desirable but also neces- Random errors: errors that are randomly distributed about the sary to use physiological and biochemical measures true value. Random errors increase the variability of a set of obser- to determine whether any method of measuring vations but do not affect their mean. food intake is actually measuring what it sets out to Repeatability (reproducibility): a method is repeatable or repro- measure. This will be discussed in more detail in ducible when it gives the same result on repeated measurement. Section 10.6. Systematic errors: errors that are not randomly distributed about the true value. Systematic errors can increase or decrease the vari- Basic concepts ability of a set of observations and also affect the estimate of their mean. The effect on the mean is referred to as bias. Before describing the most commonly used direct Validity: a method is valid if it measures what it is intended to dietary assessment methods it is appropriate to measure, i.e., the true value: introduce four fundamental concepts relevant to the process of dietary assessment and evaluation. A brief • Absolute validity or accuracy: terms used to describe the definition of terms related to these concepts is given extent to which a method measures the true value. It is not pos- in Box 10.2. Terms are listed in the box in alphabetical sible to determine the absolute accuracy of a dietary method by order for ease of reference. comparison with another dietary method. Habitual intake • Face or content validity: a method which gives results that are The objective of virtually all dietary assessments is to consistent with other data related to the parameter they are obtain an estimate of the habitual or average long- intended to measure. term intake for the group or the individual of interest. Habitual intake represents what is “usual” in the long • Relative (comparative) validity: the extent to which a test term and not simply at a specific moment in time. It method performs in relation to a criterion or reference method, is this level of intake that is relevant for maintenance i.e., the dietary method judged to provide the “best” available of energy balance and nutrient status, and for the measure of the true value. assessment of relationships between nutrient intake and health in the long term. Habitual intake, however, • Validity at group level: a method is valid at the group level if is difficult to measure because food intake varies it provides an unbiased estimate of mean intake for the group. widely from day to day and, to a lesser extent, from week to week and month to month. Figure 10.2 illus- • Validity at individual level: a method is valid at the individual trates the energy intake of one individual who main- level if it is able to rank respondents correctly (usually evaluated tained a weighed record every sixth day for 1 year. The in terms of the same proportion in the same quintile or tertile horizontal line indicates the overall average intake of the distribution compared with the “true” or more usually a over the year in MJ per day. The open circles show the “reference” value). intake on individual days and the solid circles the average intake over 7 day periods. It is obvious that Variance: statistical term to describe the variation that occurs in a intake on a single day does not provide a reliable set of observations. It is equal to the square of the standard devia- estimate of habitual intake and that even average tion of the individual observations: • Between-person variance: the variance arising from differ- ences between individuals. • Within-person variance: the variance arising from differences within individuals. intake over 7 days can differ by as much as 20% from the overall mean. The nature of error There is not, and probably never will be, a method that can estimate dietary intake without error. This

Measuring Food Intake 247 15 MJ/day 10 Figure 10.2 Energy intake of one indi- 5 vidual from weighed records obtained for 1 day every sixth day over 1 year. —, 10 20 30 40 50 60 overall mean; , weekly mean; , Days intake on individual days. does not mean that we should stop collecting dietary Precision/repeatability data but rather that dietary data need independent In the laboratory the precision of a method is given by validation. Methods need to be developed to assess the coefficient of variation (CV) of repeated determi- the error structure of dietary datasets so that it can be nations on the same sample made under the same taken into account in analyzing and evaluating the conditions (see Box 10.2). In the context of dietary data. Basically, there are two types of error: random studies we determine whether the same method gives error and systematic error. the same answer when repeated in the same individu- als, and the terms repeatability and reproducibility are Random error increases the variance of the esti- commonly used to describe the precision of a method. mates obtained and consequently reduces their preci- It is important to note that it is possible for a method sion (see below). The effects of random error can be to have high precision (good repeatability) yet not reduced by increasing the number of observations. provide an accurate (valid) estimate of intake. Day-to-day variation in food intake in individuals is one example of a random error that can be reduced Accuracy/validity by increasing the number of days of observation An accurate method is one that measures what the (Figure 10.2). method intends to measure, i.e., the “truth.” In the context of dietary studies the truth represents In contrast, the effects of systematic error cannot the actual intake over the period of the study. For be reduced by increasing the number of observations. example, a valid diet record is a complete and accurate Systematic error arises from errors that are not ran- record of all the food and drink consumed over the domly distributed in the group or in the data from a period that the record is kept. To be a valid record of given individual. Inappropriate nutrient data for habitual intake it also needs to reflect what would some food items will not affect the food intake data have been consumed had the individual not been for all individuals in the same way. For example, inap- keeping a record. If the process of recording influ- propriate nutrient data will have a greater effect on enced what was eaten then the record is not a valid nutrient intake data of individuals who consume the record of habitual intake, although it may be a true food in large amounts than on the data of those who record of actual intake over the period. Similarly, a consume only small amounts of the food. Systematic valid 24 hour recall is a complete and accurate account error leads to bias in the estimates of intake obtained.

248 Introduction to Human Nutrition of all food and drink consumed during the specified to provide clear instructions to the respondents not period if it reflects all foods and drinks consumed in only on how to weigh foods but also on how to the amounts that they were actually consumed. It may describe and record foods and recipes. When respon- not, however, be a valid reflection of habitual intake dents are responsible for weighing, the investigator if the items consumed were not typical of the indivi- needs to make regular visits to the respondent during dual’s usual intake. the recording period to ensure that the equipment is being used correctly and that information is recorded Determining the validity of a dietary technique has accurately and in sufficient detail. been impossible in the absence of external markers of intake, except for the 24 hour recall or for individual Weighing can be carried out in two different ways: meals. 1 The ingredients used in the preparation of each Methods for measuring intake on meal or snack, as well as the individual portions of specified days prepared food, must be weighed. Any food waste occurring during preparation and serving or food Menu records not consumed is also weighed. Menu records are the simplest way of recording infor- mation on food intake. They only require the respon- 2 All food and beverage items are weighed, in the dent to write down descriptions of the food and drink form in which they are consumed, immediately consumed at each meal and snack throughout the day before they are eaten, and any previously weighed for the specified days without quantifying the por- food that is not consumed is also weighed. tions. A menu record is useful when information on food patterns rather than intake is required over a The first approach is sometimes referred to as the longer period or when respondents have difficulty in precise weighing technique and is usually carried out providing quantitative information. For example, by trained fieldworkers rather than the respondents elderly people may have difficulty in reading the divi- themselves. It is thus very labor intensive, time- sions on household scales or in measuring out food consuming, and expensive to carry out. It is most portions. To derive information on nutrient intake appropriate when the food composition tables avail- from menu records investigators also need to obtain able contain few data on cooked and mixed dishes or information on portion sizes of commonly eaten if exposure to contaminants is being assessed. It foods. Information on portion sizes may be derived should be noted, however, that the precise weighing from existing data or collected in a preliminary study. technique does not allow for nutrient losses in cooking. To take these into account information on Menu records work well when the diet is relatively cooking losses for the most commonly used cooking consistent and does not contain a great variety of methods must also be available. foods. The method can be used to distinguish differ- ences in the frequency of use of specific foods over The second procedure, which is more widely used, time, to determine whether quantitative short-term involves weighing all food eaten in the form in which intake records are likely to be representative of habit- it is consumed. It is sometimes referred to as the ual intake and as a way of assessing compliance with weighed inventory method. Using this method the special diets. nutrient content of the diet can be determined either by chemical analysis of duplicate portions of indi- Weighed records vidual foods or aliquots of the total food consumed Weighed records require the respondent, or a field- or, most often, from tables of food composition. worker, to weigh each item of food or drink at the Scales used for weighing food need to be robust and time it is consumed and to write down a description able to weigh up to 2 kg, accurate to at most 5 g and of this item and its weight in a booklet specially preferably to 1–2 g. Digital scales are preferred as designed for this purpose (sometimes referred to as a these are more accurate and easier to read than spring- food diary). Weighed records are usually kept for 3, 4, balance scales. Record books must have clear instruc- 5, or 7 consecutive days. To obtain accurate informa- tions, be easy to use, and of a convenient size. They tion it is necessary to use trained fieldworkers to should contain guidelines for weighing and examples collect the data or to demonstrate the procedures and illustrating the level of detail required. Figure 10.3 shows an extract from the instructions and record

Measuring Food Intake 249 Seven day food diary instructions 1. Please use this booklet to write down everything you eat or drink for the following seven days. As you will see, each day is marked into sections, beginning with first thing in the morning and ending with bedtime. For each part of the day write down everything that you eat or drink, how much you eat or drink, and a description if necessary. If you do not eat or drink anything during that part of the day, draw a line through the section. 2. You have been provided with: a scale to weigh food, a measuring jug to measure liquids, and a set of measuring spoons to measure small amounts of foods and liquids. 3. Write down everything at the time you eat or drink it. Do not try to remember what you have eaten at the end of the day. 4. Before eating or drinking, the prepared food or drink must be weighed or measured and written in the record book. If you do not consume all the food or drink, what is left must also be weighed or measured and recorded in the book. 5. Please prepare foods and drinks as you always do. Also eat and drink in the same way as normal: eat the foods and drink in the amounts and at the times that you always eat and drink. Try not to change the way you eat and drink at all. 6. We need to know ALL the food and drink you take during these 7 days. So if you eat away from home (e.g., at work, with friends, at a cafe or restaurant) please take your measuring equipment with you so you can still measure your food. Also do not forget to measure food bought at take-aways. 7. Please write down the recipes of homemade dishes such as stews, soups, cakes, biscuits, or puddings. Also say how many people can eat from them or how many biscuits or cakes you get from the recipe. 8. On the next page is a list of popular foods and drinks. Next to each item is the sort of thing we need to know so that we can tell how it is made and how much you had. This list does not contain all foods, so if a food that you have eaten is missing, try to find a food that is similar to it. Please tell us as much about the food as you can. 9. Please tell us the amount and type of oil or fat that you use for cooking, frying, or baking. Figure 10.3 Extract of instructions and record sheets from a 7 day weighed food record (MacIntyre, 1998)

250 Introduction to Human Nutrition 10. Most packet and tinned foods, like Simba chips, Niknaks, corned meat, tinned pilchards, have weights printed on them. Tins, bottles, and boxes of cold drinks and alcoholic drinks also have weights printed on them. Please use these to show us how much you ate or drank. When possible, please keep the empty packets, bottles, or tins. Please note: we need to know the amount you eat or drink. So, if you do not eat the whole packet or tin of food, or drink the whole bottle of cold drink, please measure the amount you eat or drink. 11. At the end of each day there is a list of snacks and drinks that can easily be forgotten. Please write any extra items in here if you have not already written them down in some part of the day. 12. The research assistant will visit you during the record days to help you if you have any questions or problems. She will collect the equipment and record book after the 7 days. All the information you give us is strictly confidential. It will only be used for research purposes. Only your subject number appears on the record book. Nobody will be able to identify you from the record book. EXAMPLE Breakfast Food/drink Description and Amount Amount Amount Code preparation served left eaten Mealie meal porridge Iwiza. Soft, 1 cup meal 300g Milk and 3 cups water Bread Margarine Fresh, full cream Clover 300ml Tea Milk Brown 1 ¥ 60g Sugar Rama, soft 10 ml Figure 10.3 Continued Glenn tea bags 1 cup Fresh full cream 25 ml White 2 heaped teaspoons

Measuring Food Intake 251 Date __/___/ ____ Day no. 1 Day of week__________ Early morning — Before breakfast Food/drink Description and preparation Amount Amount Amount Code served left eaten Food/drink Breakfast Amount Amount Amount Code Description and preparation served left eaten Mid-morning — Between breakfast and lunch time Food/drink Description and preparation Amount Amount Amount Code served left eaten Food/drink Lunch time Amount Amount Amount Code Description and preparation served left eaten Mid-afternoon — Between lunch and dinner Food/drink Description and preparation Amount Amount Amount Code served left eaten Figure 10.3 Continued

252 Introduction to Human Nutrition Food/drink Dinner time Amount Amount Amount Code Description and preparation served left eaten Late night — up to last thing at night Food/drink Description and preparation Amount Amount Amount Code served left eaten Between meals, snacks and drinks if not already written in before Food/drink Description and preparation Amount Amount Amount Code served left eaten Sweets and chocolates Biscuits or cakes Simba chips Peanuts Other snacks Cold drinks Beer Tea Coffee Milk Figure 10.3 Continued

Measuring Food Intake 253 sheets of a 7 day weighed record used in a dietary using appropriate conversion data when available, or intake study in South Africa. by obtaining the necessary information when not. For example, the investigator can determine the volume of The strengths of the weighed record are that it pro- the measures commonly used in a given household vides the most accurate measurement of portion and then convert these to weights by weighing food sizes, as food is recorded as it is consumed, it does not portions of appropriate size or using information rely on memory, and it gives an indication of food about the density (g/ml) of different kinds of foods. A habits such as the number and times of meals and record book for this kind of study is similar to that for snacks. Weighed records kept for 3 or more days and a weighed record study. In some situations a precoded including a weekend day are usually considered to record form that lists the commonly eaten foods in represent habitual intake. terms of typical portion sizes may be appropriate, but an open record form is generally preferred. Limitations are that weighed records are time-con- suming and require a high level of motivation and Since there is no need for weighing scales to be commitment from both the investigator/fieldworkers provided the record forms can be distributed by mail and respondents. Respondents may change their food rather than by interviewers. This is convenient if a habits to simplify measuring and recording or may large number of respondents located over a large geo- not measure and record food items accurately. graphical area is involved. In this situation the follow- Samples of respondents who keep weighed records up interview, after completion of the record, could may not be representative of the population for three be conducted by telephone. In situations in which reasons: respondents may not be familiar with measuring foods, the investigator needs to train and provide 1 because of the high respondent burden, respon- clear instructions to the respondents and to check dents must be volunteers and thus random that respondents are performing measurements and sampling cannot be used recording correctly during the record period. 2 respondents are limited to those who are literate The strengths and limitations of estimated records and who are willing to participate are similar to those of the weighed record, but the method has a lower respondent burden and thus a 3 those who volunteer may have a specific interest in higher degree of cooperation. Loss of accuracy may food intake, e.g., being very health conscious, and occur during the conversion of household measures thus may not be representative of the population. to weights, especially if the investigator is not familiar with the utensils used in the household. Metabolic studies carried out to determine absorp- tion and retention of specific nutrients from mea- Weighed records are used in countries where surements of intake and excretion are a specialized kitchen scales are a common household item and application of the weighed food intake record. In quantities in recipes are given by weight, e.g., the UK. metabolic studies all foods consumed by the respon- Estimated records are favored in countries where it is dents are usually either preweighed or weighed by the customary for recipe books to give quantities by stan- investigators at the time of consumption. The foods dard spoons and cups, e.g., the USA and Canada. consumed are usually also analyzed for the nutrient The dietary literature frequently uses the phrase “diet constituents of interest or prepared from previously record” without specifying how portions were quanti- analyzed ingredients. fied. In these instances, estimated records are most likely to have been used. Estimated records This method of recording food intake is essentially Recalled intake similar to weighed records except that the amounts of Information on dietary intake over a specified period food and beverages consumed are assessed by volume can also be obtained by asking individuals to recollect rather than by weight, i.e., they are described in terms the types and amounts of food they have eaten. This of cups, teaspoons, or other commonly used house- approach therefore does not influence the type of hold measures, dimensions, or units. Food photo- food actually consumed in the way that a food record graphs, models, or household utensils may be used to may do. However, it is open to misrepresentation of assist quantification. These descriptive terms have then to be converted to weights by the investigator,

254 Introduction to Human Nutrition the dietary pattern, with respondents either reporting period. In order to obtain sufficient information to a “good” dietary pattern in order to project a good quantitatively analyze food intakes from a 24 hour self-image or reporting a “poor” dietary pattern in the recall, a skilled interviewer will use several “passes” or hope of receiving hand-outs or other assistance. stages in questioning the respondent. This procedure Response rates in short-term recall studies tend to has become known as the multiple-pass 24 hour range from 65% to 95% and depend largely on how, recall. This is an interviewing technique consisting of under what conditions, and from whom the infor- three to five steps which take the respondent through mation is obtained. A recall may consist of a face-to- the previous day’s food consumption at different face or telephone interview or of a self-completed levels of detail. All multiple-pass 24 hour recalls com- questionnaire. mence with the respondent simply listing all foods and beverages consumed during the previous 24 The 24 hour recall is probably the most widely used hours. The content and number of further steps differ method of obtaining information on food intake from study to study. The US Department of Agricul- from individuals. It is often used in national surveys ture (USDA) has developed a five-step multiple pass because it has a relatively high response rate and can method comprising the following passes (steps) provide the detailed information required by regula- (Conway et al., 2003). tory authorities for representative samples of differ- ent population subgroups. Pass 1 Quick list: the respondent lists all food and beverages consumed during the preceding 24 hours The 24 hour recall is an attempt to reconstruct in any order without any prompting or interrup- quantitatively the amount of food consumed either tions from the interviewer. in the previous 24 hours or on the previous day. This period is considered to provide the most reliable Pass 2 Forgotten foods list: the interviewer asks about recall of information. With longer periods memory categories of foods, such as snacks and sweets, becomes an increasing limitation. Incomplete recalls which are frequently forgotten. are more likely with self-completed records unless these records are subsequently checked with the Pass 3 Time and occasion: the interviewer asks for respondent by the investigator. An example of a 24 details of the times and names of the eating occa- hour recall sheet is shown in Figure 10.4. sions at which foods were consumed. Traditionally, the food intake has been reviewed Pass 4 Detail: the interviewer asks for details, such as chronologically, i.e., starting from the time the respon- descriptions and preparation methods, and dent wakes up and going through the day until the amounts of foods consumed. following morning. Recalling daily activities often assists the respondent to remember food intakes. Pass 5 Review: the interviewer goes through the Problems encountered in estimating the amounts of information probing for any foods which may have foods consumed are similar to those encountered been omitted. with estimated records. Recalls conducted by means of a face-to-face interview often use aids such as pho- A simplified version of the multiple-pass 24 hour tographs, food models, and household utensils to recall consists of three steps: help the respondent to describe how much food was eaten. In telephone recalls respondents may be pro- Pass 1 the respondents provide a list of all foods eaten vided with pictures or other two-dimensional aids on the previous day using any recall strategy they prior to the interview to help them to describe the desire, not necessarily chronological. amounts consumed. There is, however, very little information on how effective these aids are. For this Pass 2 the interviewer obtains more detailed infor- type of study a standardized interview protocol, which mation by probing for amounts consumed, descrip- is based on a thorough knowledge of local food habits tions of mixed dishes and preparation methods, and commonly used foods, is essential when more additions to foods such as cream in coffee, and than one interviewer is involved. giving respondents an opportunity to recall food items that were initially forgotten. In its simplest form, the 24 hour recall consists of foods and the amounts consumed over a 24 hour Pass 3 in a third pass the interviewer reviews the list of foods to stimulate reports of more foods and eating occasions.

Measuring Food Intake 255 Place Food/drink and preparation method Amount Amount Code (home, (HHM) (g) (office use work, Did you eat or drink anything when you got up yesterday? (office use only) friends, What did you have? only) etc.) Did you eat or drink anything during the morning (before about midday/lunch time)? What did you have? Did you eat or drink anything in the middle of the day (lunch time)? What did you have? Did you eat or drink anything during the afternoon (between lunch and dinner time)? What did you have? Did you eat or drink anything at dinner time? What did you have? Did you eat or drink anything during the night (after dinner and before you went to sleep)? What did you have? Is there anything else that you ate or drank yesterday that you haven’t told me already? Figure 10.4 Example of a 24-hour recall record sheet (Reproduced from Kruger, 2003, with permission of the author). HHM, household measure.

256 Introduction to Human Nutrition The multiple-pass approach is thought to assist recall habitual dietary intake if only a single 24 hour recall more effectively than chronological cues and thus is conducted. This may be overcome to some extent provide more accurate and complete information. by conducting repeated 24 hour recalls. Another dif- This approach, however, is more time-consuming ficulty is that the 24 hour recall relies on the respon- than the traditional 24 hour recall and may irritate dent to accurately recall and report the types and respondents by seemingly asking about the food amounts of foods consumed. There is a tendency for intake over and over again. Irrespective of the respondents to overestimate low intakes and under- approach used, it is essential that all interviewers are estimate high intakes. This is known as the flat slope thoroughly trained, that the approach is tested in the syndrome. Respondents may omit certain foods that target population prior to the study, and that the same are considered “bad” or include foods not consumed procedure is used by all interviewers with all respon- but considered “good” (phantom foods) in order to dents throughout the study. impress the interviewer. A major drawback of the 24 hour recall is that it Of the methods so far described weighed records provides information for only a single day and there- should contain the least error as they report all food fore does not take account of day-to-day variation in consumed on specified days with weighed portions. the diet. In large cross-sectional studies in which the Estimating the size of portions increases error and, if aim is to determine average intakes of a group of menu records are quantified with average portions, individuals, a single 24 hour recall may be sufficient. then the error at the individual level is further When the diets of individuals are assessed or when increased. If food that has already been eaten has to sample sizes are small, repeated 24 hour recalls are be recalled then poor memory can introduce an addi- required. This method is known as multiple 24 hour tional source of error. All methods that report intake recalls. The number of recalls depends on the aim of on specified days are also subject, in individuals, to the study, the nutrients of interest, and the degree the error associated with day-to-day variation in of precision needed. For example, when diets intake, but this error can be reduced by increasing the consist of a limited variety of foods two 24 hour number of days studied. recalls may be sufficient whereas four or more recalls may be required when diets are complex. Recalls Methods for measuring intake over may also be repeated during different seasons to the longer term take account of seasonal variations. (Note that mul- tiple 24 hour recalls must not be confused with the Food frequency questionnaires multiple-pass 24 hour recall technique. The multiple- Food frequency questionnaires consist of a list of pass 24 hour recall refers to an interviewing tech- foods and options to indicate how frequently each nique, whereas the multiple 24 hour recall method food is consumed. Respondents indicate the frequency refers to repeated 24 hour recalls conducted per of consumption during a specified period by marking respondent.) the appropriate option column. The food lists may contain only a few food items or may contain up to The strengths of the 24 hour recall method are that 200 foods. The type and number of foods included is it has a low respondent burden in comparison with determined by the purpose of the study and the target food records and thus compliance is high, it does not population. For example, a food frequency question- require respondents to be literate, it does not alter naire designed to determine calcium intake would usual food intake, and it is relatively quick and inex- contain only foods which provide calcium, while a pensive to administer and therefore is cost-effective questionnaire to measure overall dietary adequacy when large numbers of respondents are involved. It is would need to contain all foods known to be con- most successful in populations with limited dietary sumed by the target population. Likewise, a food fre- variety and when respondents are able to accurately quency questionnaire designed to assess dietary recall and express the types and amounts of foods intakes of a homogeneous target population with a consumed and when interviewers are skilled in the diet of limited variety will be shorter than one interview technique. designed to assess food intakes of a heterogeneous population with a variety of food intake patterns. As already stated a major drawback of the 24 hour There are several types of food frequency question- recall is that it does not give an accurate reflection of

Measuring Food Intake 257 Box 10.3 ● 1–2 times per week ● 1–2 times per month Food frequency questionnaire (FFQ) (simple/nonquantitative): ● occasionally respondents report usual consumption of foods and beverages ● never. from a set list of items for a specific period. Portion sizes are not determined. This type of response format requires only that the Semiquantitative food frequency questionnaire: a FFQ which appropriate columns be marked and is most suitable includes estimation of portion sizes as small, medium, or large. A for self-administered questionnaires. When appropri- reference portion (usually medium) may be provided as a guide. ately designed, such questionnaires can be optically Quantitative food frequency questionnaire: more precise esti- scanned, which saves time on data entry and checking mates of portion sizes are given by reference to portion size mea- procedures (Figure 10.5). surement aids (PSMAs) such as food models and photographs of known weight, household measures, or by direct weighing. Closed response options, however, treat the fre- List-based food frequency questionnaire: food items are listed quency of consumption as a categorical variable and according to groups or categories of similar foods or foods usually assume that frequency of consumption is constant eaten together. throughout the recall period. The choice of categories Meal-based food frequency questionnaire: foods are asked may bias the results: too few categories may underes- about according to meals or the time of day at which they are timate frequencies whereas too many may overesti- consumed. mate frequencies. Respondents may have difficulty in Culture-sensitive food frequency questionnaire: a FFQ that matching their food intake to the available categories. takes account of the food values, beliefs, and behaviors of a For example, when food is purchased on a monthly specific population or cultural group. basis food items such as fresh fruit and vegetables may be consumed every day while the stocks last, but, once naires, which are defined in Box 10.3. The type of used up, will not be consumed until purchased in the food frequency questionnaire used depends on the following month. purpose of the study, the target population, and the required level of accuracy of food portion An alternative method for recording responses is to estimation. provide columns headed as the number of times per day, per week and per month, seldom, and never. The period of recall depends on the study objec- From this, the average frequency of consumption and tives. In the past, most food frequency questionnaires the amount of food consumed per day can be calcu- used the preceding year or 6 months as the reference lated. Table 10.3 shows an extract from such a food period. Theoretically, this should eliminate the effects frequency questionnaire. The advantage of this of season. In practice, however, respondents tend to response format is that it allows the respondent to answer according to what is in season or available at describe the frequency of consumption in detail. the time of the study. For example, intake of oranges Responses such as consumption of a food twice a was found to be higher when interviews were carried day for 6 days of the week can be recorded. The out during the citrus season than at other times of the disadvantages are that clear instructions must be year. Information may be more reliable when recall given, making this method more appropriate for period is shorter. Recent recommendations are that interviewer-administered questionnaires than self- recall periods should not be longer than 1 month. If administered questionnaires, and the interview annual intakes are required, the food frequency ques- takes longer and requires more writing and calcula- tionnaire must be repeated in different seasons. It is tions than the closed format, making more room for very important that the respondent understands what errors. the recall period is and that only this period should be considered when giving frequencies of intake. Most food frequency questionnaires obtain infor- mation only on the frequency of consumption of a The frequency of consumption is usually indicated food over a given period and not on the context in by options such as: which the foods were eaten, i.e., on meal patterns. Meal-based food frequency questionnaires have been ● more than once a day used on the basis that it may be easier for respondents ● daily to provide the information in the context of meals. ● 3–4 times per week

258 Introduction to Human Nutrition N 1 2 3 or 1 2 3 or 4 5 or 6 Less 1–3 Please completely fill one oval in every line. E Please MARK LIKE THIS: V time times more time times times times than times E times once R per month per week per day Flavoured milk drink (cocoa, Milo™ etc.) Nuts Peanut butter or peanut paste Corn chips, potato crisps, Twisties™ etc. Jam, marmalade, honey or syrups Vegemite™, Marmite™ or Promite™ Dairy products, meats and fish Cheese Ice cream Yoghurt Beef Veal Chicken Lamb Pork Bacon Figure 10.5 Typical layout of a food frequency questionnaire suitable for optical scanning (reproduced with permission of Anti Cancer Council of Victoria, Melbourne, Australia). The information on meal patterns obtained from factor. Since the cost of administration and respon- such questionnaires is, however, more limited than dent burden are relatively low, they are suitable for that which can be obtained from a dietary history. use when sample sizes are large, particularly if a postal Additional questions, if included, can provide some method is used. information on cooking methods. The success of a food frequency questionnaire Some food frequency questionnaires also attempt depends on how closely the food list and portion size to quantify the frequency information by obtaining descriptions reflect the food patterns of the target data on portion size. Information on the quantity population. This is sometimes referred to as being of each food consumed may be obtained by asking culture sensitive. Much time and care must be put respondents to indicate on the questionnaire whether into the development of a food frequency question- their usual portions are small, medium, or large naire in order to ensure that it provides an accurate relative to those eaten by others, by asking subjects to reflection of the dietary intakes of a population. Pre- describe their usual portion size in terms of a stan- liminary studies, using 24 hour recalls, food records, dard portion size described on the questionnaire or indirect methods may be needed to obtain infor- (semiquantitative food frequency questionnaire), or mation on food items, frequency of consumption, by reference to a picture atlas of food portions, food and portion sizes in the target population. Since food models, or actual food portions (quantitative food frequency questionnaires are usually developed for frequency questionnaire). When portion sizes are use in specific target populations, a food frequency used, it is important that these reflect the consump- questionnaire developed for use in one population tion patterns of the population. may not be appropriate for use with another popula- tion with different food intake patterns. It is also Food frequency questionnaires are mainly used in extremely important that the questionnaire be tested studies designed to look for associations between for repeatability and validity in the target population food intake and disease or risk of disease, particularly before being used, even if it has been tested in a when specific foods rather than the level of consump- different population. tion of a nutrient are thought to be the important

Measuring Food Intake 259 Table 10.3 Extract from a food frequency questionnaire using open ended response options (MacIntyre, 1998) Office use only Food Description Amount Times eaten Code Amount/day Per day Per week Per month Seldom Maize meal porridge Stiff porridge 4256 4255 Soft porridge 9999 Sour porridge 9012 4043 Do you pour milk on your porridge? YES NO A014 If NO, go to next question on sugar. 9816 If YES, what type of milk: fresh, sour, milk powder (Nespray), skim milk powder (name)? 4040 A018 If yes, how much milk? 4001 Do you pour sugar on your porridge/cereal? YES NO 4002 If no, go to ‘samp’ 4057 If yes, how much sugar? 6508 6521 Samp (only) 6509 Samp and beans Tshidzimba (samp, beans, and peanuts) Rice Macaroni/spaghetti Bread White bread Brown bread Fat cakes Other types of bread Do you spread anything on the bread? YES NO If no, go to next section (chicken, meat, fish) IF YES, What do you spread and how much? Margarine Hard Soft Peanut butter Other types of spread Detailed reviews of the design and evaluation of ● an interview to obtain usual diet food frequency methods for use in epidemiological ● a cross-check of this information by food group studies are given by Willett (1998) and Cade et al. ● a 3 day record of food consumed in household (2002). measures. Diet history The principal objective of the diet history is to obtain The 3 day record component is now seldom used as detailed information on the habitual intake of an a regular component of a diet history. Its purpose individual. As first proposed by Burke in the 1940s, originally was as a way of checking the data obtained the method had several components (Burke 1947): from the diet history interview.A diet history is usually obtained by an experienced nutritionist by means of an open interview followed by some kind of cross-

260 Introduction to Human Nutrition check using a standard list of commonly consumed In contrast, the errors that are associated with spe- foods. The interview usually begins with a review of cific methods are generally much more dependent on the food that was eaten in a specific time-frame (e.g., the nature of the method and the abilities of the yesterday) or on a typical day, and then moves on to respondents, and therefore less easy to control. Errors explore the variations in food intake that occur for of this type include: each meal over a given period. Information on the usual size of food portions is obtained with the aid of ● estimation of portion size food models or photographs in the same way as for a ● recall or memory error 24 hour recall. The time-frame for a diet history can ● day-to-day variation in intake range from the previous month to the previous year. ● effect of survey method on food intake. In practical terms it is easier for respondents to Figure 10.6 illustrates the points in the dietary assess- reconstruct the immediate past, but the past year is ment process at which different kinds of error often used to capture seasonal variation. Whatever the operate. time-frame used, it is important for it to be clearly specified. In the literature the term “diet history” Sampling bias is sometimes used loosely to describe any form of diet recall, including the 24 hour recall and self- Sampling bias arises when the sample studied is not administered food frequency questionnaires, as well truly representative of the population of interest. The as interviewer-administered recalls of habitual or importance of minimizing sampling bias depends on longer term intake. This broader use can be confusing the purpose of the dietary study. Except in method- and is best avoided. The dependence of the diet ological studies, volunteers are not generally appro- history on both respondent and interviewer skills priate respondents because frequently the objective is may make the results obtained less comparable to study a representative sample of a particular group between individuals than those obtained from other in order to extrapolate the results to the population methods, and for this reason it is often considered group from which the sample was drawn. For this more appropriate to categorize diet history data (e.g., purpose it is clearly important that as many as possi- as high, medium, low) rather than to treat them as ble of the sample group originally selected participate intakes expressed in terms of absolute units per day. in the study. The diet history is favored in Scandinavia and the The proportion of the sample that agrees to par- Netherlands, where a structured interview may be ticipate in the study can vary considerably even with used. The structured interview is more standardized the same method. It depends not only on the group but may miss elements specific to the individual or being studied but also on the circumstances of the bore the respondents with irrelevant questions. The study. In general, response rates tend to be greater in open-ended interview allows for tailoring to the indi- studies that use methods such as the 24 hour recall vidual, but risks missing important items. and food frequency questionnaires (which make fewer demands on respondents) and lower in studies 10.4 Sources of error in dietary studies such as 7 day weighed records (which require much more cooperation, effort, and time from the respon- The major sources of error in dietary studies have dents). It is always important to try to maximize been reviewed in detail by Bingham (1987). Four pos- response rates, for example by increasing respondent sible sources of error occur to some degree with all motivation, providing specific assistance if required, dietary methods, but can be minimized by careful and by allowing respondents as much flexibility as study design and execution: possible in participation within the context of the study objectives. ● sampling bias ● response bias Response bias ● inappropriate coding of foods ● use of food composition tables in place of chemical Response bias arises when respondents provide incomplete or inappropriate responses. The extent of analysis. this problem is not easy to assess, but can be investi- gated by making measurements that are independent

Measuring Food Intake 261 Behavior and interactions Altered food Poor memory Inability to “summarize” dietary patterns Messages from researcher choice False perception of own diet Attitudes to food Socioeconomic status FFQ Image management Report RECORD RECALL HISTORY Inadequate Identify food Researcher’s choice Subject’s choice communication/ from food tables from questionnaire descriptions Quantify Weighed/ Estimated Average portion Poor quantification: Weighing errors estimated or small/medium/large – Questionnaire choices Poor estimations – Subject’s perceptions Underreporting Frequency As reported Subject’s Subject’s choice Overreporting estimate from questionnaire Food table values Calculate Food code Food code differ from actual intake x Portion weight x Portion weight composition x Nutrient content/g x Frequency x Nutrient content/g Figure 10.6 The process of dietary assessment showing the different sources of error (shown in italics) and the stages at which they operate in different dietary methods. FFQ, food frequency questionnaire. of dietary intake, both during and after dietary study procedures do not necessarily “standardize” respon- periods. Measurements suitable for this purpose dent response. To date, relatively little work has will be discussed further in the context of validity. focused on the respondent response aspect of dietary Response bias can probably best be minimized by assessment. providing the respondents with clear and well- presented instructions, adequate support, and Since all dietary methods engage the cognitive appropriate incentives. Such incentives may include processes of respondents, an appreciation of the relevant dietary feedback where this is appropriate or properties of human cognition and its limitations is monetary or other rewards provided that these are fundamental to improving the accuracy of dietary within ethical principles. assessments. Recently, research into the cognitive aspects of dietary assessment has been undertaken in In dietary studies that involve more than one inter- an attempt to increase the understanding of how viewer, the training of interviewers and the use of respondents process dietary intake data (Domel, standard procedures for interviewing is one way 1997; Thompson et al., 2002; Vuckovic et al., 2002; of reducing unnecessary random variation (error) Matt et al., 2006). that might otherwise arise because different inter- viewers conduct interviews in different ways. The use Some of the important issues in this area that are of standard procedures, however, can also introduce relevant to improving the quality of dietary data systematic error; for example, if one interviewer is include identification of: assigned to interview all respondents in areas of low socioeconomic status and another to interview all ● factors that improve communication between respondents in areas of high socioeconomic status. It respondent and investigator is important to recognize also that standard interview ● the most effective cues for recall over different periods

262 Introduction to Human Nutrition ● factors that influence retention of dietary informa- of this section is simply to review briefly the kinds of tion over time error that can arise as a consequence of using food composition tables to calculate nutrient intake, com- ● the ways in which individuals conceptualize foods pared with chemical analysis of the diet, and which and food quantities. can lead to both random and systematic errors. Coding Systematic error can result from: Coding refers to the allocation of a specific code ● the way in which results are calculated or to each food item. Since the nutritional content of a expressed food varies with different processing and preparation methods, it is vital that the correct codes be assigned ● the analytical method used to each food item. Coding errors arise when the ● the processing and preparation methods in common food that has been consumed is not described in suf- ficient detail to enable unambiguous allocation, by use. the investigator, to a food category in a food composi- tion table or database. Food frequency questionnaires Food composition tables for different countries often are often precoded to reduce the time needed for use different ways of expressing results and different coding and the possibility of coding errors (see Table analytical methods. The ways in which food items are 10.3). Making it easy for respondents to describe processed or prepared are also likely to differ and for foods with the level of detail required is therefore these reasons different sources will not necessarily an important consideration in study design. This is provide comparable data for the same foods. increasingly difficult, particularly in industrialized Systematic differences, which may not necessarily be countries where the food supply now consists of errors (e.g., when foods are prepared differently in thousands of different manufactured foods, the names different countries), often only become evident when of which are often no longer a good guide to their different food composition tables are used to evaluate nutrient content. the same diets. Coding errors are also likely to arise when more Random error arises from the fact that most foods than one person is involved in coding and there is vary in their composition as a result of changes in no agreed procedure and/or comprehensive coding composition associated with the conditions of pro- manual. Coding errors arising exclusively from duction, processing, storage preparation, and con- inadequate description of foods have resulted in sumption. The random error associated with the use coefficients of variation ranging from 3% to 17% for of food composition databases generally decreases as different nutrients. Note that a standard procedure the size of the sample group increases. This may not for coding foods, while minimizing differences be true, however, in institutional settings where every- between coders (random error), can also introduce one is likely to be consuming food from the same bias if the coding decisions that are made are not source. based on up-to-date knowledge of the local food supply and food preparation methods. Gross errors To compare calculated and analyzed data without associated with weights of foods can be checked, the complication of other sources of error it is neces- before analysis, by means of computer routines that sary that the diets are analyzed by collecting a dupli- identify values outside a prescribed range and by cate of what has been eaten at the same time as the using data-checking techniques such as duplicate data diet record. At group level it has been observed that entry. mean intakes calculated from the food tables are gen- erally within approximately 10% of the mean ana- Use of food composition tables lyzed value for energy and macronutrients, but not for micronutrients. However, a large proportion of Most dietary studies use food composition tables or individuals have values that fall outside this range. databases rather than chemical analysis to derive the nutrient content of the foods consumed. Chapter 2 In general, calculated and analyzed values for nutri- describes in detail the way in which data on food ents agree more closely: composition are derived and compiled. The purpose ● for groups than for individuals ● for macronutrients than for micronutrients ● when data for locally analyzed foods are used.

Measuring Food Intake 263 Estimation of portion size ● commercial or home-made food models: ● food pictures or drawings of different portion Estimation of portion size has long been recognized sizes as an important source of error in dietary studies ● photographs of foods in different portion sizes (Young et al., 1953) with coefficients of variation of ● abstract shapes of cardboard, wooden or plastic the differences between estimates and weights of food blocks, wedges, circles, balls, and mounds in portions regularly reported to be around 50% for various sizes foods and 20% for nutrients (Nelson and Bingham, ● household utensils and containers such as cups, 1997). However, despite the fact that individuals spoons, jugs, glasses, bowls, and plates in various are known to vary widely in their ability to estimate sizes portion size, relatively few studies have attempted to ● containers and packets of bought foods, e.g., quantify the size of this error or to “calibrate” their sweet wrappers, potato crisp packets, cold drink respondents in this respect. The influence of some cans and bottles, yoghurt and ice cream cups, factors on the determination of portion size is milk cartons. summarized in Box 10.4. Each PSMA has advantages and disadvantages. The In attempts to assist respondents to describe type of PSMA chosen will depend, among others, portion sizes, a number of visual aids, known as on the type of study, the target population, whether portion size measurement aids (PSMAs), have been interviewers go from house to house or respondents developed. These include: go to a research centre, available resources, and the availability of appropriate PSMAs. Probably the most ● weighed portions of actual foods effective method is a combination of PSMAs such as ● allowing respondents to serve out portions of food food photographs and household utensils. Irrespec- tive of the type of PSMA used, it is essential that and direct weighing or measuring of the serving respondents are able to identify and relate to the PSMA, that PSMAs be tested in the target population Box 10.4 prior to their use, and that PSMAs are used con- sistently throughout the study. Food characteristics • No consistent associations observed with type of food, although Recall errors foods of indeterminate shape are more often associated with Factors that have been studied in relation to the accu- larger and liquid foods with smaller errors. racy of dietary recall include food consumption • Large portion sizes appear to be more difficult to estimate accu- patterns, weight status, gender, and age. Many other rately than small portion sizes. characteristics, such as intelligence, mood, attention, and salience of the information, however, have also Visual aids been demonstrated to influence performance tests of • Household measures may be associated with considerable general recall, but have not been studied in the context of dietary recall. errors. • Food models produce more reliable results than household mea- Short-term memory Like the ability to estimate portion size, the ability to sures, but because only one size is usually available they may remember what was eaten varies with the individual. “bias” respondents to report portion sizes similar to those Studies that have compared the abilities of different shown. groups to remember what they have eaten conclude • Graduated food models and two-dimensional pictures may be that women are generally better than men and that as helpful as three-dimensional models for estimating portion younger adults are better than older adults. In short- size. term recalls of intake (e.g., 24 hour recalls) indivi- • The range of reference portion sizes available may influence the duals more often tend to omit an item or items that estimates. they have consumed than to include ones that they • The use of multiple photographs results in more accurate esti- mates. Respondents • Respondents of all ages have been reported to have difficulty with portion size estimates. • Women have sometimes, but not invariably, been reported to be better able to estimate portion size than men, but this may simply reflect the fact that they tend to handle food more often than men.

264 Introduction to Human Nutrition have not consumed. For this reason, 24 hour recall the recall and the reference period. The accuracy of studies often provide estimates of food intake that are frequency estimates also deteriorates with time. While lower than food records obtained over the same individuals appear to report more frequently eaten period. The size of the error incurred by the omission foods with greater frequency than less frequently eaten of one or more food items clearly depends on what is foods, there are differences between individuals in the omitted and not only on the proportion of food way that they report the same frequency of consump- omitted. For example, the effect, on 24 hour energy tion. Ranking of individuals on the basis of the usual intake, of omitting a cup of black coffee, a glass of frequency of intake is thus likely to lead to misclassifi- milk, or a bar of chocolate is quite different. cation unless the extent of the differences between individuals is known and can be taken into account. It The omission of food items in 24 hour recall studies is difficult to see how such misclassification can be can be reduced by appropriate probing by the inter- reduced unless it is possible to classify individuals, in viewer in relation to meals, between-meal snacks, and some way, in terms of their ability to provide reliable other activities on the previous day, but even when information on habitual long-term intake. respondents have previously weighed their food the average energy intake may still be underestimated by Day-to-day variation in intake as much as 20%. We have already seen that individuals vary consider- Long-term memory ably in their intake of nutrients from day to day (see The diet history and most food frequency ques- Figure 10.2). In addition, the extent of day-to-day tionnaires set out to measure the habitual intake of variation differs between nutrients. The implication an individual over a period of weeks or months. of the first observation is that short-term intake data Individuals are not asked to recall their food intake on (e.g., 24 hour recall data) are unlikely to provide a specific days, but to construct a picture of their “usual” reliable estimate of habitual intake for most individu- food consumption pattern over a specified reference als. The implication of the second observation is that period. To provide reliable information individuals the length of time for which dietary data need to be thus need to be able to remember the range of foods collected, in order to estimate habitual intake with that they usually consume, to judge the frequency of any given level of confidence, varies with the nutrient consumption on a long-term basis, and to be able to of interest. estimate correctly the average amount that is usually consumed. These are complex cognitive tasks. Table 10.4 expresses the impact of this variation in terms of the number of days of dietary information As in the case of 24 hour recalls, no attempt is needed to classify 80% of individuals into the correct usually made to assess how well individuals are able third of the distribution. It is clear from this table that to perform these various tasks. From the limited not only 24 hour recalls but also 7 day records are amount of data available from comparative studies likely to be inadequate to classify 80% of individuals between diet histories and long-term diet records, it correctly into the appropriate third of the distribution appears that the two methods do not give concordant for most micronutrients. This is an important reason, results in individuals. Food frequency questionnaires although not the only reason, why short-term records are subject to the same difficulties, and have the added are only rarely used for epidemiological studies, in problem that estimates of portion size are based on preference to food frequency questionnaires, despite standard measures or, in the case of mailed question- the loss of detail and precision inevitably associated naires, are made in the absence of visual aids such as with the use of the latter. food models or photographs. Effect on usual diet When respondents are asked to report their intake over a period of weeks they rely largely on generic Recall methods clearly cannot change what has already knowledge of their diet and tend to report items that been eaten, but what has been eaten can be misre- they are likely to have eaten or items that they rou- ported either consciously or unconsciously. When tinely eat, rather than items that they specifically individuals are asked to keep records, however, they remember having eaten during the reference period. may also alter their normal habits as a consequence This tendency increases with the time interval between of the recording process. One obvious reason for

Measuring Food Intake 265 Table 10.4 Number of days of records required to enable 80% of men defined, since this will determine the kind of informa- to be assigned into their correct third of the intake distribution tion and the length of time for which it needs to be collected from each individual. Often, the purpose of Nutrient British civil Random sample Random sample the study also determines the level of precision that is servants of British men of Swedish men required to meet the study objectives and therefore Energy the sample size. These two considerations are the Protein 7 57 most important ones in determining the method to Fat 6 57 be used, because both the method and the size of the Carbohydrate 9 97 sample have implications for the human and financial Sugar 4 33 resources needed for the study. Dietary fiber 2 2– P:S ratio 6 10 – Purpose of the study Cholesterol 11 –– Alcohol 18 –– When dietary data are collected to describe the diet Vitamin C 4 – 14 of a group for comparison with that of another group Thiamin – 6 14 or groups, it is possible to use either a short-term Riboflavin – 6 15 method such as a 24 hour recall or record, or a longer- Calcium – 10 – term method such as food records obtained over Iron – 45 several days, a diet history, or a questionnaire about – 12 9 habitual intake. The final choice will depend on factors such as the importance of a representative P:S ratio, ratio of polyunsaturated to saturated fatty acids in the study sample, the resources available, and the level diet. (Reproduced from Margetts BM, Nelson M. Design Concepts in of precision required. Usually, the most efficient Nutritional Epidemiology. Oxford: Oxford University Press, 1991, with approach is to measure the diet of as many individu- permission from Oxford University Press.) als as possible for 1 day. doing so would be to simplify the process of record- However, if the purpose of the dietary study is to ing. Other reasons may include a desire to eat less in determine the proportion of individuals in the group order to lose weight or to be seen to conform with who are at risk of dietary inadequacy or excess, rela- dietary recommendations. If this is what happens in tive to some standard of reference, then a single day practice, then what is measured in short-term dietary of information on each individual is no longer ade- records may be actual intake or desired intake, but not quate because it is necessary to have a reliable esti- usual intake. mate of the distribution of habitual intake in the group. As Figure 10.2 shows, a single day of intake is Many studies have now demonstrated that there is generally not a reliable measure of an individual’s a tendency, in most population subgroups, for short- habitual intake. term dietary records to provide estimates of energy intake that are on average around 16% lower than To determine the distribution of habitual food would be expected on the basis of measured and/or intake in a group, at least 2 days (preferably not estimated levels of energy expenditure. consecutive) of information from each individual or a representative subsample of individuals from the These studies will be discussed further in the group of interest are needed. If several days of intake section on precision and validity. The fact that for are available they can be used to derive a mean intake some groups measurements of energy intake and for each individual and from this the distribution of energy expenditure agree quite closely indicates that average intakes for the group. Alternatively, statistical it is possible to achieve recording without a concomi- techniques can be used to adjust 1 day intake data, for tant change in diet when there is full cooperation the day-to-day variation that occurs in individuals, to from respondents, and highlights the importance of provide a better estimate of the underlying distribu- efforts to achieve such cooperation. tion of habitual intake for the group than is given by the 1 day data (Dodd et al., 2006). While the use of 10.5 Choosing a dietary appropriate statistical techniques can improve esti- assessment method mates of the proportion of individuals at risk of It is not possible to decide which dietary method to use until the purpose of the study has been clearly

266 Introduction to Human Nutrition deficiency by adjusting for within-person variation, resources available it is better either to abandon the they do not enable at-risk individuals to be study or to redefine the question than to collect inad- identified. equate data. When the purpose of the study is to assess the diet 10.6 Repeatability and validity of specific individuals it is necessary to obtain dietary information over at least a week and preferably longer. This section looks at ways in which it is possible to This is best done by obtaining either multiple 24 hour assess the repeatability and validity of dietary recalls or 24 hour food records over an extended methods. period. The minimum number of days needed to obtain an estimate of nutrient intake with a specified Repeatability level of confidence differs for different nutrients. Information on energy intake, which tends to show Assessing the repeatability (also referred to as the less day-to-day variation than other nutrients, can be reproducibility) of a laboratory method is relatively obtained over a shorter period (days) than informa- straightforward because, with care, it is possible to tion on a nutrient for which day-to-day intake is reproduce both what is measured and the conditions much more variable, such as vitamin A (weeks). of measurement. This is almost always impossible in the case of a dietary intake measurement. Individuals Precision do not eat exactly the same quantities or the same foods on different days or weeks. In studies of groups, precision is primarily a function of sample size, while in studies of individuals it is a All measures of repeatability obtained by applying function of the number of days of information avail- the same method to the same individuals on more able. Precision increases with sample size and with the than one occasion include not only measurement number of days for which information is collected, error but also real day-to-day or week-to-week vari- but so does the cost of the study. Precision therefore ability in intake. needs to be defined in relation to the purpose of the study. While at first sight it might appear easier to measure the repeatability of recall methods such as the 24 hour Usually, what is required of the nutritionist is to be recall and diet histories, this process also introduces able to provide the statistician with an estimate of the additional sources of variation since the interviews level of difference that it is important to be able to have to be conducted at different times and possibly detect (in nutritional, not statistical terms) and an by different interviewers. Measures of repeatability estimate of the variance or standard deviation for the for all dietary methods will thus tend to give an over- measurement(s) in question. For example, when estimate of the extent of measurement error because looking for differences in energy intake between two they will always include an element of variation due groups, would a difference of 500 kJ or 1500 kJ be to real differences in what is being measured and in regarded as biologically significant? the conditions under which it is being measured. Since the variance of a dietary measurement Usually, the repeatability of a dietary method is depends not only on the real variation within or determined by repeating the same method on the between respondents but also on the error of the mea- same individuals on two separate occasions, that is, surement, the precision of a dietary estimate can be by a test–retest study. The interval between tests improved not only by increasing sample size but also depends on the time-frame of the dietary method by reducing measurement error. being assessed, but should generally be short enough to avoid the effects of seasonal or other changes in Resources food habits and long enough to avoid the possibility of the first interview or recording period influencing It is inevitable that the resources available, both finan- the second one. cial and human, also influence the choice of method. They should not, however, be the primary consider- The difference between the results obtained on ation. The method used should be determined by the the two occasions can be expressed in a number of question to be answered. If the method or methods different ways. Table 10.5, which was compiled from needed to answer the question are beyond the data reported in the literature, shows various mea-

Measuring Food Intake 267 Table 10.5 Measures of repeatability for energy intake obtained for Validity a 3 day food record, a dietary history, and a food frequency question- naire (FFQ) Demonstrating that a dietary method measures what it is intended to measure is even more difficult than Measure of 3 day food Dietary demonstrating that a method is repeatable, because repeatability in effect it “requires that the truth be known.” record history FFQ Mean 954 This is almost always impossible unless it is possi- difference (kJ/day) 156 105 12.5 ble to observe, surreptitiously, what is consumed over 28.5 short periods such as 24 hours or at most a few days. Mean difference (% of 1.6 1.1 Observation is usually only feasible in institutional overall mean intake) settings or in situations specially set up to allow unob- 16.5 18.6 trusive observation of what people eat. Coefficient of variation of the differences ±3266 ±1819 ±4294 For methods that are designed to obtain informa- within individuals tion on habitual longer-term intake, such as the diet (%) – 0.86a 0.70a history or food frequency questionnaires, unobtru- 56 – 60* sive observation is impossible. This is a problem that Coefficient of has been faced by all investigators of dietary assess- repeatability (kJ) ment methods and until relatively recently was usually “solved” by assessing one dietary method in relation Correlation coefficient to another dietary method, usually a 7 day weighed Individuals classified in dietary record, which was considered to be the best available or criterion measure. Comparison with the same quartile or another dietary method provides at best only a rela- tertile* on both tive form of validity and at worst information that is occasions (%) unrelated to validity but reflects either real differences or similar errors between the methods. For example, a Intraclass correlation. comparison of data from a single 24 hour recall or a diet history with data from a 7 day weighed record sures of repeatability for energy intake obtained with for the same individuals does not compare the same different dietary methods repeated after an interval information because the time periods are not con- of time. current. However, because of the lack of a suitable external standard against which true validity could be The different measures of repeatability provide judged before the 1980s it was usually assumed that different information. The correlation coefficient is most dietary intake data, and weighed records in par- widely quoted but is not a good measure of repeat- ticular, provided valid data. Usually, a method was ability since a good correlation may be obtained even judged acceptable if the mean intake, as measured by if one set of measurements has been systematically both methods, did not differ significantly and if cor- biased and has a different mean from the other set. relations for nutrient intake in individuals exceeded The mean difference is not a good measure of repeat- 0.5. The magnitude of the coefficient of variation ability in individuals since it depends primarily on of the differences within individuals was generally whether the differences are random or systematic. ignored. Measures that reflect the differences between repeated measurements within individuals are to be preferred. Table 10.6 shows data from three studies that The coefficient of variation of the differences within provide additional information on agreement. All individuals and the coefficient of repeatability (which three studies compared data from a food frequency is simply twice the standard deviation of the differ- questionnaire with multiple days of food intake ences and represents the 95% confidence limits of records. When different methods are compared the agreement) give much better measures of their mag- mean differences tend to be higher (there is greater nitude. They are also more readily interpreted in bias) than those found in repeatability studies. practical terms than either a correlation coefficient or However, the range of values obtained for other the percentage of individuals classified in the same quintile, quartile, or tertile. If the standard deviation of the difference within individuals is of the order of 20–30% of mean intake, one is unlikely to describe the method as precise or repeatable even if the mean difference at group level is only 1%.

268 Introduction to Human Nutrition Table 10.6 Measures of relative validity for energy intake obtained from a food frequency questionnaire (FFQ) compared with multiple days of records in three different studies Measure of validity 7 days of weighed 12 days of weighed 15 days of records in records records household measures Mean difference (kJ/day) (FFQ – record) Mean difference (% of overall mean intake) 800 926 351 Coefficient of variation of the differences within individuals (%) 6 11 4 Coefficient of repeatability (kJ) 32.7 26.6 16.9 Correlation coefficient ±4542 Individuals classified in the same quintile or tertile* on ±8248 0.69 ±2950 – 42 0.71 both occasions (%) 43* 44 measures of agreement is generally similar to that measures are also subject to errors of measurement obtained in repeatability studies. Agreement at the and classification, but these errors are not related individual level is also not high, with coefficients of to the errors inherent in dietary intake assessment variation for differences in individuals ranging from methodologies. 17% to 33% in these studies and less than 50% of respondents classified in the same quintile of intake. The three most widely used measures to assess the Note that even good agreement between two dietary validity of dietary intake data are urinary nitrogen to methods does not necessarily indicate validity, but validate protein intake, energy expenditure as mea- may merely indicate similar errors. sured by the doubly labeled water (DLW) method to validate energy intake in weight-stable individuals, Biological measures to validate energy and and the ratio of energy intake to basal metabolic rate nutrient intake to identify “plausible” records of food intake. It is now clearly recognized that to assess the validity Urinary nitrogen of any dietary method, including weighed records, it One of the first to suggest an external measure as a is necessary to compare the dietary data with one or means of validating dietary intake data was Isaakson more objective measures that reflect but are indepen- (1980), who proposed urinary nitrogen as an inde- dent of food intake. At the group level such measures pendent measure of protein intake according to the include food supply or food expenditure data, and at equation: the individual level biochemical or physiological mea- sures that reflect energy and nutrient intake. The Reported protein intake (g) = latter are often referred to as biological or biochemical (24 hour urinary N + 2) × 6.25 (g) markers and include energy expenditure, urinary breakdown products of protein, sodium, and potas- Like the 24 hour recall, a single 24 hour urine collec- sium, plasma levels of vitamins, tissue levels of miner- tion does not necessarily reflect what is “usual.” als, and the fatty acid composition of subcutaneous However, it appears that urinary nitrogen excretion is adipose tissue. less variable from day to day than dietary protein intake, and that while 16 days of food intake are Biological marker – any biochemical index in an needed to assess habitual protein intake only eight 24 easily accessible biological sample that in health hour urine collections are needed to assess nitrogen gives a predictive response to a given dietary excretion with the same level of confidence. component. Although fewer 24 hour urine collections may (Bingham, 1987). be needed they are, in general, no more acceptable to respondents than 24 hour food records and also Biological markers are assumed to be objective, require access to laboratory facilities. Nevertheless, i.e., they do not rely on memory, or the respondents’ they can provide a practical independent assessment, ability to express themselves, and are free of biases not only of protein but also of potassium and sodium introduced by the presence of the interviewers. These intake.

Measuring Food Intake 269 Very few validation studies have attempted to eval- The main advantage of the DLW method is that it uate how well the different methods rank individuals. makes minimal demands on the respondents and It has been observed that the correlation between does not in any way interfere with their normal daily urinary nitrogen and dietary nitrogen measured by activities and therefore their habitual level of energy diet records was better (0.65 and 0.79) than between expenditure. Its main disadvantage is that the cost of urinary nitrogen and dietary nitrogen measured by the DLW required for each estimate is exceedingly food frequency questionnaires (0.15 and 0.24). high and the method also requires access to sophisti- cated laboratory equipment for mass spectrometric Doubly labeled water method analysis. It is, therefore, not available for use on a The DLW technique allows the measurement of routine basis for the validation of dietary intake energy expenditure in free-living respondents over data. several days with minimal inconvenience to the respondent and with a high level of accuracy and Ratio of energy intake to basal metabolic rate precision. Under controlled conditions the DLW Because of the limitations of the DLW method, method gives a small overestimate of 2–3% compared another approach that is used compares the energy with whole body calorimetry, and under field condi- intake (EI) reported from published studies with the tions bias is not expected to exceed 5%. presumed requirements for energy expenditure, both intake and expenditure being expressed as multiples The DLW method requires that the respondent of the basal metabolic rate (BMR). The relevant drinks a small measured dose of water enriched with equation is: naturally occurring stable isotopes of deuterium and 18O. The two isotopes disperse throughout the body, EI:BMR = EE:BMR (PAL) and are metabolized and then gradually lost from the body. Since the deuterium labels the body water pool, where PAL is the physical activity level. To determine and the 18O labels both the water and the bicarbonate whether reported energy intake is a “plausible” pools, the difference between the disappearance rates measure of actual diet during the measurement period of deuterium and 18O can be used to calculate carbon (i.e., represents either the habitual diet or is a low/ dioxide production. The level of both isotopes is high energy intake obtained simply by chance) an determined, using mass spectrometry, in a small equation was developed by Goldberg and colleagues sample of urine collected each day for between 5 and (1991) to calculate the 95% confidence limits of 28 days. Energy expenditure is calculated from carbon agreement between EI:BMR and PAL. This equation dioxide production using calorimetric equations. allowed for variation in EI, BMR, and PAL and also Further details of the DLW technique and the main for the length of the dietary assessment period and factors influencing its accuracy and precision are pro- study sample size. vided in Chapter 3. For a group, if mean reported EI:BMR is below the Using the DLW technique several investigators lower 95% confidence limit (cut-off) for the given have compared self-reported dietary energy intake study period and sample size, then there is definitely with energy expenditure based on the equation: bias to the underestimation of energy intake. Energy expenditure (EE) = Energy intake (EI) However, the identification of individual under- ± Change in the body energy store reporters is much more difficult, since reported EI can deviate quite markedly from energy expenditure (EE) Differences between measured energy intake and before it falls outside the limitations of the methods. expenditure varied from –44% to +28% depending Figures 10.7 and 10.8 illustrate the limitations of both on the population subgroup studied. This finding techniques. Figure 10.7 shows the energy intake and confirms the need to include one or more indepen- DLW EE of 264 women. The solid lines indicate EI: dent measures of validity in all dietary studies EE of 0.76 and 1.24. These are the 95% confidence to ascertain the level applicable to the particular limits of agreement between EI and EE, allowing for group under study, since it is not readily predicted day-to-day variation in food intake and within-subject on the basis of gender, age, or body mass index variation on repeat DLW measurements. Only women (BMI). with an EI:EE ratio above 1.24 can be confidently

270 Introduction to Human Nutrition 20 UR Women 3.0 UR AR 2.5 AR OR 2.0 OR 15 EI (MJ) EI:BMRest 10 1.5 1.0 5 0.5 0 0 0 5 10 15 20 1.0 1.5 2.0 2.5 3.0 DLW EE (MJ) PAL (EE:BMRmeas) Figure 10.7 Reported energy intake (EI) against energy expenditure Figure 10.8 Energy intake–estimated basal metabolic rate (EI:BMRest) (EE) measured by doubly labeled water (DLW) in 264 women aged against physical activity level [PAL; energy expenditure–measured 18–90 years. The solid lines represent the 95% confidence limits of BMR (EE:BMRmeas)]. Respondents are designated as acceptable report- the expected agreement between EI and EE (±24%). UR, underreport- ers (AR), overreporters (OR), or underreporters (UR) by the direct com- ers; AR, acceptable reporters; OR, overreporters (after Black, 2000, parison of EI:EE. The horizontal lines indicate the lower Goldberg reproduced with permission). cut-off for PAL = 1.55 and 1.95, 7 day records, and n = 1 (after Black, 2000, reproduced with permission). identified as overreporters, and those below 0.76 as application to other populations must be done with underreporters. In Figure 10.8, the same data are caution. expressed as EI:BMR and EE:BMR, and each respon- ● The ratio does not take differences in physical activ- dent is represented by the same symbol as in Figure ity into account (Black, 2000). 10.7. The line at EI:BMR = 1.05 indicates the Gold- ● Using a single cut-off point to identify under- berg cut-off for n = 1 and PAL = 1.55, which has been reporters has been found to have poor sensitivity widely used to identify low energy reporters (LERs). for underreporting (fails to identify underreport- These data demonstrate that only about 50% of ers), especially at high levels of energy intake (Black, underreporters (as defined by EI:EE = 0.76) are iden- 2000). tified as LERs. The second line at EI:BMR of 1.35 is ● Cut-off values differ among studies. Thus it is dif- the Goldberg cut-off for n = 1 and PAL = 1.95. This ficult to select an appropriate value and to compare cut-off identifies more of the underreporters, but also studies. includes some of the acceptable reporters. To improve ● Cut-off values apply only to individuals in energy on identification of underreporters it is necessary balance. They cannot be applied to growing chil- to have information on physical activity to enable dren or to adults trying to lose weight (Gibson, respondents to be classified into different levels of 2005). activity and to calculate cut-offs appropriate for each activity level. Characteristics of low energy reporters A number of studies have examined the characteris- While the ability to separate the food and nutrient tics of low energy reporters (LERs). Associations intake data of those with and without plausible energy between low energy reporting and a large number of intakes is a very important step in the evaluation of factors including high body weight, high BMI, obesity, dietary intake data, it does have limitations. dieting, and awareness of body image were found in every study that looked at these measures.Associations ● The equations for the estimation of BMR have been derived for Western populations and their

Measuring Food Intake 271 with other sociodemographic factors of the type nor- Allowing for the effects of underreporting mally included in nutrition studies, such as gender, age, education, socioeconomic status, and smoking, Although techniques for handling biased dietary data were inconsistent. have been developed, most are complex However, the following suggestions serve to promote critical 10.7 Evaluation of food intake data examination of data and wariness in drawing conclusions. Recognizing the impact of underreporting If the proportion of individuals who report implau- As indicated in Section 10.5, dietary studies are often sibly low intakes of energy differs between population conducted in order to compare food and nutrient subgroups of interest, then any comparisons between intake between different groups in the population, to them that do not take this into account will be biased. determine the proportion of individuals at risk of One way to draw attention to the possibility of bias dietary inadequacy or excess, or to determine the between groups is to report not only the mean or habitual intake of individuals. median energy intake of the groups being compared but also the EI:BMR ratio. If differences are evident In each case it is important first to assess the valid- then the groups should be compared both with and ity of the data. For most investigators the use of the without LERs included. One problem that arises is Goldberg cut-offs is currently the most practical that by subdividing the groups the sample size is option to indicate whether, and to what extent, the reduced and imprecision increased, so that a differ- results are likely to be biased. However, to use the ence of biological significance may be missed, not Goldberg cutoffs effectively dietary studies need to because it does not exist but because the sample size include: is too small to detect it statistically. ● measurements of weight and height, to be able to When dietary inadequacy or excess is the question estimate BMR from equations of interest, it is again important to consider LERs separately. Energy intake is highly correlated ● questions on activity level to provide guidance on with the intake of many nutrients and, consequently, suitable PALs for evaluation of both mean and indi- intake of nutrients is also likely to be underestimated vidual data. in underreporters and more likely to indicate inadequacy relative to recommendations for nutrient While the characteristics of “true” underreporters intake. An alternative approach is to compare (as opposed to LERs identified by a single EI:BMR nutrient intake per unit energy for both groups. If cut-off) remain to be confirmed, the associations this differs between LERs and the rest of the popula- consistently observed between high BMI, weight con- tion it provides evidence that the reporting of food sciousness, and low energy reporting suggest that, in intake is also likely to be selective. The nutrients for addition, questions on self-perception of body shape, which significant differences are observed can also dieting, and dietary restraint may also help in identi- provide clues as to the types of food likely to be fying true underreporters. involved. It cannot be overemphasized that it is always Studies that have examined macronutrient intake important to examine all dietary intake data critically between respondents above and below a given value because false conclusions generate false hypotheses of EI:BMR have generally found that the percentage that may take years to be disproved. of energy derived from protein was higher and that from fat lower in LERs than in non-LERs. Results for A classic case was the luxus konsumption hypoth- carbohydrate have been more variable, but, when esis, namely that lean individuals are energy prodigal separated into starch and sugars, energy from starch and obese individuals energy efficient. This hypothe- tended to be higher and energy from sugars lower in sis was generated by studies apparently showing that LERs. Nutrient density also tends to be higher for obese persons did not consume more energy than most nutrients in LERs than in non-LERs, providing their lean controls. Subsequently, DLW studies dem- further indication of differences in food patterns onstrated beyond doubt that obese persons recruited between the two groups. for studies of obesity grossly underreported their food intake.

272 Introduction to Human Nutrition Table 10.7 Nutrient intakes in low energy reporting (LER) and non-LER The second limitation is that all estimates of adults in the 1995 Australian National Nutrition Survey dietary adequacy/inadequacy obtained by compari- son with reference values for nutrient requirements Nutrient LER Non-LER Total depend on how the estimate is derived (see (n = 1291) (n = 9451) (n = 10 851) Chapter 7). Energy (MJ) Protein (g/MJ) 4.62 10.19 9.24 However, irrespective of the approach that is used Fat (g/MJ) 11.0 9.8 9.9 to assess dietary adequacy, unless the extent of under- Starch (g/MJ) 9.0 9.0 reporting is known and taken into account, the pro- Sugars (g/MJ) 8.1 14.9 15.0 portion of individuals at risk of inadequacy will be Vitamin (μg/MJ) 15.9 12.4 12.4 overestimated. While it may become possible to dis- Riboflavin (mg/MJ) 12.9 122 127 tinguish more reliably in population-based studies Folate (μg/MJ) 159 0.22 0.23 valid from invalid reports of dietary intake, this still Vitamin C (mg/MJ) 0.26 28 29 does not enable population-based estimates of inad- Calcium (mg/MJ) 37 13 13 equacy to be made unless those who provide valid Iron (mg/MJ) 19 90 92 intakes are also representative of the population as a 106 1.5 1.5 whole. All the evidence available to date suggests that 1.8 this is highly unlikely. Data used from the Australian Bureau of Statistics (abs.gov.au) (ABS, When the principal objective of a dietary survey is 1998). to identify the proportion of the population who may have inadequate intakes of energy and nutrients, it is Table 10.7 illustrates these general trends with data essential that the dietary intake information is inter- from the 1995 Australian National Nutrition Survey. preted in the light of appropriate biological measures Twelve per cent of men and 21% of women in this of nutritional status. survey were identified as LERs. The median energy intake in non-LERs was approximately 6% higher in 10.9 Assessing food intake men and 10% higher in women than for all men and women, and vitamin and mineral intake approxi- Nutritionists usually analyze dietary intake data by mately 5–10% higher in non-LER men and 6–15% converting the information on food intake into nutri- higher in non-LER women. Differences of this order ent intake using relevant food composition databases. of magnitude are important in the context of the This approach simplifies the process of analysis and assessment of dietary adequacy. enables the resulting data to be compared with energy and nutrient requirements (see Chapter 7). Describing Relatively few studies have reported on differences food intake in terms of foods rather than nutrients in foods eaten, but there appears to be a general ten- presents two practical difficulties that do not exist dency for LERs to report more foods such as meat, when food intake is analyzed in terms of nutrients. fish, vegetables, salads, and fruit, and fewer cakes, First, the variety of foods consumed is much greater biscuits, sugar, confectionery, and fats. than the range of nutrients for which food composi- tion data are available. Second, while essentially all 10.8 Assessment of dietary adequacy individuals in a group contribute to nutrient intake data, not all individuals contribute food data for all Methods for evaluating dietary adequacy are described foods, i.e., not all individuals are “consumers” of the in Chapter 7. This section simply draws attention to same foods. the limitations of these methods. There are, however, several uses for which informa- The first limitation is that the evaluation of nutri- tion on food intake is more relevant or for which ent intake can provide only an estimate of the risk of information on food intake is needed in conjunction nutrient inadequacy for a population or an indivi- with data on nutrient intake. For example, food regu- dual. None of the methods can identify the specific latory authorities and agencies concerned with food individuals who have a nutrient deficiency. Individu- safety and nutritional surveillance require data on the als with a nutrient deficiency or excess can be identi- fied only on the basis of biochemical or clinical measures of nutritional status.

Measuring Food Intake 273 availability and intake of foods in addition to infor- but may also be converted to nutrients to provide mation on nutrient intake. Similarly, nutritional epi- information on the nutrient contribution of individ- demiologists are also interested in the relationship ual foods or groups of foods. of different foods and dietary patterns with specific health outcomes. The use of dietary data in the context Tracking changes in the food sources of nutrients of epidemiological studies is covered in the textbook and nonnutrients is particularly important in the Public Health Nutrition (Gibney et al. 2004). context of technological developments in food pro- duction and manufacture that result in the addition The analysis and presentation of food intake data of nutrients to foods, in the development of foods for depends on the objectives of the study. When the specific functional purposes, and in the genetic modi- purpose is to examine intakes of specific foods, intakes fication of foods. A specific example of the need for of foods may be expressed as means, medians, or individual food, rather than nutrient, intake data is frequency distributions of intakes, as the number or provided by exposure assessments to dietary non- percentage of respondents consuming specific foods, nutrients such as food additives, pesticide residues, or as the percentage contribution of food items to the and other possible food contaminants. total food intake, energy intake, or intake of nutrients of interest. Since not all members of a sample consume 10.10 Food safety assessments a given food, it is always important to indicate whether the total sample size or only the number of respon- Safety assessments for food additives are expressed in dents consuming the food has been used in statistical terms of the acceptable daily intake (ADI) estimated calculations. on the basis of lifetime exposure. While it is clearly not possible to collect food consumption data over Although intakes of individual food items may be the lifetimes of individuals, it is important that the reported, food intake data are usually reduced to dietary data used for the purpose of estimating accept- more manageable proportions by grouping foods into able levels of intake over a lifetime reflect, as far as is appropriate categories. While this can be done in possible, the habitual level of intake of the foods being different ways, for example in terms of composition, assessed. biological origin, or cultural use, the process is relatively straightforward within a given culture or For the purpose of food safety assessments only the country. It is more difficult, however, to develop a intake of “consumers” is of interest. It follows, there- classification that can be used consistently across dif- fore, that the dietary data need to be adequate to ferent countries or food cultures. National food clas- obtain both an accurate estimate of the proportion of sification systems tend to differ not only because the the population who are consumers and of the average type and range of foods differs but also because the habitual intake of consumers. Because the frequency same foods are used in different ways. For the purpose of consumption varies between foods (some foods are of comparing food intake patterns between countries eaten by most people on most days, but many other or regions, it is, therefore, necessary to develop a food foods are eaten less frequently), the duration of the classification or coding system that allows food data dietary recording period influences both the esti- from individual regions or countries to be assigned in mate of the proportion of consumers and the average a consistent way. intake of consumers. Intake data for 1 day will inevi- tably underestimate the true number of consumers The United Nations University Food and Nutrition for most foods and overestimate the average habitual Program for an International Network of Food Data intake of those consumers because not all foods are Systems (INFOODS) was developed for the purpose eaten every day. However, it appears that 75% or more of supporting work on the classification and naming of household menu items are normally consumed conventions for individual foods and food groups within a 14 day period and that a 14 day diary pro- (see Chapter 2). vides a good estimate of the habitual intake of most foods by consumers. Indirect information on food consumption, such as that provided by FAO food balance sheets and by Most studies of the food intake of individuals, data from household budget and similar surveys, is however, do not last for 14 days because of the usually presented in terms of foods or food groups,

274 Introduction to Human Nutrition increased cost and nonresponse associated with such supplements (both nutrient and nonnutrient) and a long study period. For the purpose of food safety drugs (both social and medicinal). assessment an approach that combines a 3 day food intake record with a food frequency questionnaire has Acknowledgment the potential to give estimates for the intake of con- sumers that are similar to those obtained from 14 day This chapter has been revised and updated by Una E records. MacIntyre based on the original chapter by Ingrid HE Rutishauser and Alison E Black. 10.11 Perspectives on the future References It is unlikely that either the measurement or the eval- uation of food intake will become less complex in Australian Bureau of Statistics. Australian National Nutrition future. If anything, the reverse is likely to be true given Survey. ABS, Canberra, 1998. the increasing diversity in the food supply and the increasing recognition of the need to be able to assess Bingham SA. The dietary assessment of individuals; methods, accurately not only the intake of foods and nutrients accuracy, new techniques and recommendations. Nutr Abstr Rev but also the intake of nonnutrient constituents of (Series A) 1987; 57: 705–742. foods and dietary supplements. While the existence of errors in association with measurements of food Black AE. The sensitivity and specificity of the Goldberg cut-off for intake is now widely appreciated, much work still EI:BMR for identifying dietary reports of poor validity. Eur J remains to be done in this area. Clin Nutr 2000; 54: 395–404. Other aspects of food intake measurement that also Burke BS. The dietary history as a tool in research. J Am Diet Assoc require further development in the immediate future 1947; 23: 1041–1046. are likely to include the following. Cade J, Thompson R, Burley V, Warm D. Development, validation As all direct methods of food intake measurement and utilisation of food-frequency questionnaires: a review. Publ involve interaction between investigators and indi- Health Nutr 2002; 5: 567–587. viduals and our understanding of the cognitive aspects of these interactions is still limited, more work is Conway JM, Ingwersen LA, Vinyard BT, Moshfegh AJ. Effectiveness needed to improve the communications aspect of of the US Department of Agriculture 5-step multiple-pass dietary assessment. method in assessing food intake in obese and nonobese women. Am J Clin Nutr 2003; 77: 1171–1178. As the food supply becomes more complex indi- viduals will no longer be able to describe the foods Dodd KW, Guenther PM, Freedman LS, et al. Statistical methods they have eaten in adequate detail unless technologi- for estimating usual intake of nutrients and foods: a review of cal developments such as the use of barcodes and the theory. J Am Diet Assoc 2006; 106: 1640–1650. similar systems of food identification become an inte- gral part of dietary assessment. Domel SB. Self-reports of diet: how children remember what they have eaten. Am J Clin Nutr 1997: 65 (suppl): 1148S– As the number of food constituents of interest, in 1152S. relation to health, increases it is important that appro- priate physiological and biochemical markers are also Gibney MJ, Margetts BM, Kearney JM, Arab L. Public Health developed for these constituents, as well as for the Nutrition. Blackwell Publishing, Oxford, 2004; 67–75. nutrient constituents of foods. Gibson RS. Principles of Nutritional Assessment, 2nd edn. Oxford Finally, since food intake data serve no useful University Press, Oxford, 2005. purpose unless they can be appropriately evaluated it is essential that dietary studies include sufficient Goldberg GR, Black AE, Jebb SA et al. Critical evaluation of energy ancillary information to allow this to occur. This intake data using fundamental principles of energy physiology. means routinely collecting information not only on I. Derivation of cur-off limits to identify under-recording. Eur J age, gender, body size, and physiological status, but Clin Nutr 1991; 41: 569–581. also on key aspects of lifestyle such as physical activity and the consumption of nonfood items such as Isaksson B. Urinary nitrogen output as a validity test in dietary surveys. Am J Clin Nutr 1980; 33: 4–5. Kruger R. The determinants of overweight among 10–15 year old school children in the North West Province. University of Potchefstroom for CHE. Unpublished PhD Thesis, 2003. MacIntyre UE. Dietary intakes of Africans in transition in the North West Province. University of Potchefstroom for CHE. Unpublished PhD thesis, 1998. Matt GE, Rock CL, Johnson-Kozlov M. Using recall cues to improve measurement of dietary intakes with a food frequency question- naire in an ethnically diverse population: an exploratory study. J Am Diet Assoc 2006; 106: 1209–1217. Nelson M, Bingham SA. Assessment of food consumption and nutrient intake. In: Margetts BM, Nelson M, eds. Design Concepts in Nutritional Epidemiology, 2nd edn Oxford University Press, Oxford, 1997. Thompson FE, Subar AF, Brown CC, et al. Cognitive research enhances accuracy of food frequency questionnaire reports:

Measuring Food Intake 275 results of an experimental validation study. J Am Diet Assoc Cypel YS, Guenther PM, Petot GP. Validity of portion-size 2002; 102: 212–218, 223–225. measurement aids: a review. J Am Dielet Assoc 1997; 97: 289– Vuckovic N, Ritenbaugh C, Taren DL, Tobar M. A qualitative study 292. of participants’ experiences with dietary assessment. J Am Diet Assoc 2002; 100: 1023–1028. Food and Agriculture Organization.Food Balance Sheets: Application Willett W. Food-frequency methods. In: Willett W, ed. Nutritional and Uses. Available online at http://fao.org (accessed 18 July Epidemiology, 2nd edn. Monographs in Epidemiology and 2007). Biostatistics. Oxford University Press, New York, 1998: 74. Young CM, Chalmers FW, Church HN, Murphy GC, Tucker RE. Food and Agriculture Organization. Food Balance Sheets and Food Subjects’ estimation of food intake and calculated nutritive value Consumption Surveys: Comparison. Available online at http:// of the diet. J Am Diet Assoc 1953; 29: 1216–1220. fao.org (accessed 18 July 2007). Further reading Thomson FE, Byers T. Dietary intake resource manual. J Nutr 1994; 124 (Suppl): 2245S–2316S. Venter CS, MacIntyre UE, Vorster HH. The development and testing of a food portion photograph book for use in an African population. J Hum Nutr Dietet 2000; 13: 205–218. Black AE. Critical evaluation of energy intake using the Goldberg cut-off for EI:BMR. A practical guide to its calculation, use and limitations. Int J Obes 2000; 24: 1119–1130.

11 Food Composition Hettie C Schönfeldt and Joanne M Holden Key messages • Factors such as sampling, variability and analytical methods involved must be considered when developing such tables. • Reliable good-quality composition data of foods for human con- sumption are critical resources for a variety of applications. • Inadequacies of food composition tables can be minimized by calculating nutrient losses and gains during food processing and • These data are required for a spectrum of users ranging from preparation. international to national, regional, household, and individual levels. • New activities in food composition include: • future composition tables could include bioavailability and the • In general, data obtained on food intake by individuals, or groups glycemic index of individuals, are used to estimate the consumption of nutrients • harmonizing food composition tables regionally and to establish nutritional requirements and health guidelines. • focusing on biodiversity within species • investigation of the composition of specific traditional and • The determination of the consumption of nutrients can be ethnic foods achieved either by analyzing the foods consumed directly (by far • bioactives in foods and their effect on health and well-being the most accurate, but also the most costly method) or by using • food composition data and their role in nutrition and health food composition tables/databases. claims. • The food described in the food composition table should be rec- ognizably similar to that being consumed by the individual or group. 11.1 Introduction used today as there are limited up-to-date tables avail- able. Worldwide there are currently over 150 food Although the amount, quality and availability of food composition tables or nutrient databases, or their composition data vary among countries and regions, electronic/magnetic equivalents, in use. Many tables in general most developing countries still do not have are based on the data from the United States adequate and reliable data. This is despite the fact that Department of Agriculture’s (USDA) National the components of specific foods have been published Nutrient Database for Standard Reference, SR, for over 150 years. Over time food composition data available on the Nutrient Data Laboratory’s web site: have assumed more scientific, academic, and political www.ars.usda.gov/nutrientdata. A comprehensive list importance owing to their utility. Refer to Table 11.1 of the food composition tables available can be for practical examples of the uses of food composi- obtained from the Food and Agriculture Organization tion data. It was only in 1961 that a regional food of the United Nations (FAO) homepage on the World composition table was developed and published for Wide Web (http://www.fao.org/infoods/directory). Latin America, followed by a food composition table EuroFIR, the European Food Information Resource for Africa (1968), the Near East (1970), and Asia Consortium, is a partnership between 40 universities, (1972). The data in these tables were based on a very research institutes, and small to-medium-sized enter- limited number of samples, a limited number of prises from 25 countries in Europe. EuroFIR aims to nutrients and, in today’s terms, outdated analytical develop and integrate a comprehensive cohort and methodologies. However, these tables are still being validated network of databanks of food composition © 2009 HC Schönfeldt and JM Holden.

Food Composition 277 Table 11.1 Examples of the uses of food composition data Level Examples International Role of food in the provision of nutrients and/or the estimation of adequacy of the dietary intake of National population groups Regional (influenced by meal Investigation of relationships between diet, health, and nutritional status, e.g., epidemiologists correlate patterns and food patterns of disease with dietary components preferences) Evaluation of nutrition education programs Household Nutrition intervention and food fortification programs such as in food assistance programs; foods are Individual distributed or enriched to address the specific nutritional needs of populations, e.g., iodine or vitamin A In food trade nutritional labeling Monitoring at governmental level, the availability of foods produced and estimating the individual intake for specific dietary requirements, e.g., protein and energy Food balance sheet data are used to provide data on food available nationally for the whole population and are useful in monitoring trends in food consumption over time Researchers work to improve the food supply by selecting or developing new strains or cultivars, improving cultivation, harvesting, preservation, and preparation Estimation of adequacy of the dietary intake of groups within populations Investigation of relationships between diet, health, and nutritional status Evaluation of nutrition education programs Food and nutrition training Nutrition education and health promotion Nutrition intervention and food fortification programs Food and nutrition regulation and food safety Nutrition labeling of foods Institutions such as hospitals, schools, dormitories, and troops/armies (ration scale) formulate nutritionally balanced diets to the individuals in their care Food industries regulate the quality of their foods by routinely analyzing the components in their products Food industries change and improve their products to appeal to new customers by improving nutrient content or sensory appeal through the change in ingredients Product development Household food surveys provide data on household food consumption Household budget surveys Household food economics Dietary intake of the individual is assessed to understand present health and to monitor changes in dietary intake Impact of interpretation of choice and preference via data composition Individual energy expenditure is the only true measurement of energy need, e.g., in the management of a sportsman’s diet or in obesity Personal dietary needs and goals with associated likes and dislikes can be assessed on an individual basis Individual nutritional balance studies Therapeutic or restricting diets with specific nutrient contents, e.g., management of diabetes and hypertension, can only be described on an individual basis Individual shoppers scan the ingredient list and nutrient content on the labels of packaged foods Sports nutrition data for Europe. This network, although comprehen- Biodiversity for Food and Nutrition led by the FAO sive, has at present limited access to the broader nutri- and Biodiversity International focused on genetic tion society. diversity within species and of underutilized, unculti- vated, and indigenous foods. The investigation has There is still a continued need to carry out food highlighted the need for composition data of foods, analyses as the number of foods consumed all over the not only at species level, but also at subspecies level. world, especially unique foods, is still several times The limited amount of composition data for under- greater than the number for which analytical data utilized, uncultivated, and indigenous foods playing exist. The recent (2008) Cross Cutting Initiative on

278 Introduction to Human Nutrition important roles in the consumption patterns in under- foods that form a major part of the food supply and developed and developing countries increases this that are major contributors to the diet in the forms need for food composition analysis. Food analyses are most commonly obtained or consumed, and as many also needed under the following circumstances: as possible of the less frequently consumed foods. For instance, in the USA the number of foods contribut- ● when the data in existing tables are based on a ing to quartiles of critical nutrient intakes was identi- single or very limited number of samples fied as the following: 9 foods contribute to 25% of food intake, 34 foods to 50%, 104 to 75%, and 454 to ● when the content of a nutrient or other food com- approximately 100%. ponent is not available in an existing food table Databases can be compiled directly, where the ● when there is no information available on which compiler initiates sampling and analyses to obtain the foods are important sources of a nutrient or another data, or indirectly by drawing on the following sources food component of interest of data, in order of preference: ● when there is no information on the loss or gain of ● original analytical values nutrients in foods during preparation by the ● imputed values derived from analytical values from methods being used by the population under investigation a similar food, e.g., values for “boiled” used for “steamed” ● when it is necessary to check the comparability of ● calculated values derived from recipes, calculated the various food composition tables being used in from the ingredients and corrected for preparation a multicenter study factors ● borrowed values (refers to using data originally ● when the method available to determine a particu- generated or gathered by someone else) from other lar nutrient is considerably improved tables and databases. ● when scientific evidence is found correlating newly Today, database compilers normally draw on a com- recognized food components to health bination of the direct and indirect methods. ● when new foods are produced or existing foods are Description of foods reformulated. The food described in the food composition table 11.2 Foods should be recognizably similar to that being con- sumed by the individual or group. The precise descrip- Food composition tables normally consist of a list of tion of foods is a difficult task and much is required selected foods with data on the content of selected to ensure that foods are described adequately. The nutrients in each food. For a food composition table to introductory material (description and explanation) be of value in estimating nutrient content, a significant in a printed table may be almost as important as the portion of the foods consumed by the group or indi- data values. By using several words to describe a food, vidual being studied, as well as the nutrients of interest, called an extended or multifaceted description, the should be present in the table. To a large extent this chance of misinterpreting the data is reduced. As relationship is critical in determining the quality of the internationalization of food composition data con- information obtained by using the tables, assuming tinues, linguistic aspects of defining foods, with one that the data in the tables are of a desirable quality. definition meaning different things in different cul- tures and even from place to place within countries, Criteria for inclusion in tables are highlighted. For instance, sorbet or sherbet is made by beating whisked egg whites into the partly The identification of potential contributions of foods frozen mixture such as in apple sorbet and lemon to the diet of the population group being studied is sorbet. However, the term sorbet is preferred to unquestionably the first step in identifying and select- sherbet, since the latter can also refer to a flavored, ing which foods should and should not be included sweet, sparkling powder or drink, or a drink of sweet in the production of a database. However, common diluted fruit juice. The name tortilla is also applied to sense dictates that it is unreasonable to expect that all foods consumed by all individuals at all times be included in a specific food composition table at any one time. Therefore, most tables aim to include all