Previous Page Table of Contents Next Page


Keynote Paper: Individual food intake survey methods

Anna Ferro-Luzzi
National Institute for Food and Nutrition Research
Rome, Italy

Executive summary

This paper describes the suitability of individual -level information on dietary intakes for estimating the state of food security of population groups at national or subnational levels, and illustrates the constraints, shortcomings and potential advantages of this methodological approach. It begins by specifying that consumption of foods in a quantity and quality sufficient to meet energy and nutrient requirements represents the core of the food security concept, and that any consideration relating to the subjective perception of deprivation or a condition of poverty comes second. The paper reviews the diverse methodological approaches available for assessing individual food intake and briefly describes their inherent and logistic constraints, as well as the need to adapt them in specific socio-economic and cultural contexts. An overview is provided of the stringent research that has gone into the improvement of the quality of data collected and the level of reliability of the data that can be achieved. The degree of flexibility afforded by the individual dietary survey methods, the availability of effective validation and standardization procedures, and the nature of the information obtained are features that are unique to this methodological approach. Also unique is that its error structure is far better understood than for any other method employed for assessing food security, and independent validation of the results through concurrent measurement of energy expenditure by the doubly labelled water approach is feasible. Finally, this is the only existing method that can reveal intrahousehold distribution of food. Neglecting to account for the potential of unequal intrahousehold allocation patterns may lead to faulty conclusions relative to the food security of the household and may mask the existence of at-risk subgroups in the community. The paper identifies the phenomenon of under-reporting as the main weakness of the method, but also considers that logistic considerations and the cost of the surveys are potential constraints. The paper concludes that the individual dietary survey method is a robust approach for assessing food security, but its cost and other considerations, such as logistics, the degree of collaboration required from subjects and particular cultural constraints, make its use difficult, especially in developing regions of the world. It also concludes that this method is of great value for the validation of other, more expedient methods for assessing food security.

Introduction

This paper describes various survey methods for measuring food intake at the individual level and the feasibility of using these methods for the assessment of food security. Food security exists when "all people, at all times, have physical, social and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life" (FAO, 1998). Thus, while the concept of generic deprivation and poverty, and the subjective perception of material insecurity are all legitimate features of the food insecurity concept, it appears self-evident that the core measure of food security is that the intake of energy and of nutrients must be commensurate with their requirements.

The assessment of dietary intake at the individual level raises several methodological issues: accuracy and representativity of the results as well as logistics and cost of conducting the surveys. It also raises interpretation issues, as the assessment of energy and nutrient adequacy or inadequacy may only be obtained on a probabilistic level, and therefore it is not possible to identify specific individuals with inadequate intake, as set out by the expert WHO consultation (FAO/WHO/UNU, 1985). Interpretation of survey results at the probabilistic level is, however, likely to provide a sufficient basis for decision-making for policy purposes. These aspects will be discussed in further detail below.

Scope, purpose and outcomes of individual dietary surveys

The overall objectives of individual dietary surveys may be several: to judge the adequacy of the diet to meet energy and nutrient requirements, to establish the presence of a link between a putative dietary risk factor and a given health outcome, or to monitor for clinical purposes the response of a patient to manipulations of the diet. Three types of information may be provided: the estimate of individual intakes, the ranking of individuals on the basis of their food and/or nutrient intakes within a group and the average intake of the group. Most commonly, individual dietary surveys are undertaken for the purpose of establishing the existence, strength, direction and level of associations between dietary exposure and health outcome in the individual. However, nationally representative individual dietary surveys have also been conducted for the purpose of providing an estimate of national nutritional status, but only in a limited number of countries and seldom in developing countries for reasons of cost, expediency and logistics. National or subregional diet and nutrition monitoring surveys have been conducted in Ireland (Irish Nutrition and Dietetic Institute, 1990; National Nutrition Surveillance Centre, 1993), in Great Britain (Gregory et al., 1990), in the Netherlands (Hulshof, 1993), in Finland (National Public Health Institute, 1998), in the United States (Federation of American Societies for Experimental Biology, Life Sciences Research Office, 1995; National Center for Health Statistics, 1994; Rizek and Pao, 1990; US Department of Agriculture, Nutrition Monitoring Division, 1986), in Spain (Gobierno Vasco, 1994), in Denmark (National Food Agency, 1990), in Italy (Turrini, Leclercq and D'Amicis, 1999), in India (Department of Women and Child Development, 1998), and in China (Ge, Zhai and Wang, 1997). (This is not an exhaustive list.) These surveys were all conducted at the individual level and employed diverse techniques including a seven-day dietary history, a two-to seven-day weighed records survey and single or repeated 24-hour recalls.

Individual -level dietary intake information is particularly indicated for studying special subgroups of the population, such as pregnant women, infants, the elderly or persons in special circumstances such as refugees or hospital patients (see, for example, Scaccini et al., 1992). The individual approach may also be required when it would be incorrect to assume that food is equally distributed within the household according to each member's needs or there are individual family members who may be at higher risk than others.

Individual surveys provide data on the amount of foods or food categories consumed by the selected individual over the period covered by the survey. As described in the next section, the period may be in some recent or less recent past or may be concurrent to the survey. The survey may also retrieve ancillary information, such as the daily episodes when specified foods have been consumed, whether they were home-produced or bought and, in such cases, also the commercial label, the way in which foods were cooked, the place where they were consumed, whether there was any edible wast- age and so on. Tables of food composition are used to calculate the energy and nutrient content of the consumed foods.

A Review Of Individual Dietary Survey Methods

Descriptions of the various methods used for the assessment of dietary intake can be found in the literature (Dwyer, 1994; Gibson, 1990; Willett, 1998), including comprehensive reviews (such as Bingham, 1987; Medlin and Skinner, 1988). In the 1990s, four successive international conferences were held to review and improve dietary assessment methods. The proceedings of first two conferences have been published in extenso (Buzzard and Willett, 1994; Willett and Sampson, 1997), the third conference published only the abstracts, and the proceedings of the fourth conference is expected to be published in the near future. This renewed interest in the dietary survey methodology grew out of a general discontent with the limitations of the dietary survey methods, and in particular with those aiming to estimate individual dietary intakes. The impetus has come in particular from epidemiologists who explore the association between dietary intakes and health outcomes, and who therefore need valid methods for assessing dietary intakes at the individual level.

Basically, the methods for assessing individual dietary intakes can be classified into two main categories: the retrospective reporting of intake from the recent or remote past and the prospective recording of consumption. In each of these two categories, several subcategories exist, as shown in Table 1. The type of information obtained by individual surveys varies by the method adopted: it can be the individual's actual "point" intake or the longer-term "true" intake, a mean of the group's intake or a simple ranking, e.g. by quintiles. Each of these methods has its strengths and limitations, and the choice of one or the other method must be made by balancing the expediency and the logistical simplicity of some of the methods with the robustness of the outcome variables obtained by the others, such as ranking or precise quantitative data, according to the specific needs of the study.

TABLE 1. TYPOLOGY OF INDIVIDUAL DIETARY SURVEY METHODS

All these methods have undergone various phases of improvement, simplification, optimization and customization for specific purposes and/or local circumstances. Each of these methods can be used in a variety of different ways depending on the purpose of the study, a fact that makes them very flexible instruments, applicable under the most diverse conditions. However, it is precisely this remarkable flexibility that, while representing one of the strengths, has also created one of the main drawbacks, as many adaptations have been developed and used without adequate validation. The processes of validation of the individual survey methods has revealed a wide variability in the quality of the data (such as repeatability and precision) depending upon the context and method of the survey, the eating habits of the subjects, the nature of the diets and also the specific nutrient under study (Rockett and Colditz, 1997).

It is worth while to note that a similarly stringent analysis of the errors and validation of the results have yet to be undertaken for the several other available approaches to estimate dietary intake, such as, for example, the household survey, the household budget survey (see paper by Smith in this series) and the food balance sheet method (see paper by Naiken in this series). The description and evaluation presented later on known errors of the individual dietary survey methodology must not be seen as singling out the different ways of measuring individual dietary intake as the worst case, but instead should serve to highlight the close scrutiny that this set of methods has undergone and to suggest that other methods of studying food security or undernutrition should undergo a similar critical analysis.

Dietary history method

The dietary history method is one of the oldest approaches for assessing individual diets (Burke, 1947), but it is used less frequently today. The dietary history is carried out in three steps. The first step is to establish the usual eating pattern and to determine which foods are habitually eaten at which meals. Both the frequency and the quantity of consumption may be obtained, the latter usually expressed in common household measures. Food models or pictures are sometimes used to improve the precision of the estimated quantities. The second step is a kind of cross -check during which the interviewer uses a comprehensive list of foods and fills in possible memory gaps of the interviewee. Again, quantities consumed are asked in household measures, and the frequency of consumption over the designated interval is obtained. The final step is the recording of all foods consumed, usually in household measures, over the next three days. The time interval covered by the dietary history can vary from a few weeks to a season or to a year or more, and the period may range from the immediate or recent past to the remote past. The interview is carried out face to face and may take between 30 and 45 minutes to complete without undue rushing (Mensink, Haftenberger and Thamm, 2001). However, more time is required on the part of the subject to record all foods eaten over the next three days.

Food frequency questionnaires

First developed by Wiehl in 1960 (Wiehl and Reed, 1960), the food frequency questionnaire (FFQ) has many similarities with the dietary history method. It is now routinely employed when financial resources and time are limited, and it is suitable for very large population samples. It is the tool to measure food intake most often used in epidemiological research. In its original design, the questionnaire was employed for rating or grading dietary items into four or five categories of consumption for the purpose of differentiating consumers and non-consumers of particular foods and to test for association with disease. Different variations of the FFQ have been developed, such as those that use qualitative food- frequency scores, (for example Kumanyika et al., 1997) or those that provide semiquantitative estimates of the usual diet. The questionnaire consists of a checklist with a number of individual food items and various levels of food aggregation. The number of food items can be as many as 190 or more (Katsouyanni et al., 1997) or, in the abbreviated formats, less than 22 items such as in the Hispanic Health and Nutrition Examination Survey (Rockett and Colditz, 1997). The length of the list of foods in the questionnaire influences the response: the longer the list, the higher the level of the reported consumption (Kaaks, Slimani and Riboli, 1997). The questions relate to the frequency (daily, weekly or seasonally) of consumption of foods or food groups over a given period of time in the recent or remote past. Questionnaires are administered by trained or lay personnel in face -to-face interviews or by telephone, but also may be self-administered through postal surveys. Questionnaires that are optically scannable simplify and facilitate translation of the information into computer-ready data.

FFQs have been used alone or in combination with a variety of memory aids such as food models, paper silhouettes, photographs, calibrated household measures and computers (computer-assisted self-interviewing for the quantification of portion sizes) (Kohlmeier et al., 1997).

24 -hour recall

The 24 -hour recall method is widely used to provide information on foods consumed by the individual over the previous 24 hours. The quantities consumed are retrieved in the course of an interview that, in the experience of the author, may last from 30 to 45 minutes or more, during which a first free report is obtained without prompts, followed by a prompted report for the purpose of filling in any possible gaps or omissions. The interviewer may use appropriate memory aids, such as photographs of prepared dishes with diverse and calibrated servings, food models, standardized household measures, paper silhouettes of portions or portable scales. Information on cooking methods, recipes, and labels of industrially prepared foods are also retrieved.

The 24 -hour recall method relies on the subject's capacity to remember what they have eaten. As memory declines rapidly beyond one day, the recall method usually retrieves information only on the previous day's consumption. Multiple 24 -hour recalls are better suited for the purpose of estimating food intake over a longer period. For information on children's diet, if the child is less than nine to ten years of age, the parents are usually interviewed. Older children are capable of providing a reasonably faithful picture of their diets, although some uncertainty persists on the quality of the information retrieved (Baranowski and Domel, 1994). There also might be problems with older subjects whose short-term memory tends to fail.

Weighed dietary records

This method involves recording the weights of all foods prior to their consumption. Subjects are provided with a calibrated scale of the appropriate capacity (2 to 5 kg) and accuracy (± 1 to 5 g), and a booklet in which to enter the weight of all foods and beverages consumed and of any edible or inedible leftovers. For meals consumed away from home, estimates of the weights are recorded by the subject and later checked by the dietitians with the help of household measures, models, photographs, etc. The modality of weighing foods varies. The actual foods eaten may be weighed, or weights of the individual ingredients may be taken from recipes or label information in the case of commercially prepared foods. In the "precise weighing" dietary survey, all ingredients of the cooked composite dishes are weighed raw, before cooking and after cooking. If foods are prepared for the entire family, this approach requires that the entire prepared dish be weighed and then the individual subject's portion weighed as well as any leftovers. The precise weighing method requires that the subject be provided with a scale of appropriate capacity (up to 10 kg). The weighing may be performed by the subjects themselves or, in populations with a high prevalence of illiteracy, by an observer. In the "estimated food weight record" method, the weight of the foods consumed is estimated by the subject rather than weighed.

The number of days of food weighing and recording usually varies between one and seven, rarely more. The length of the survey depends on the purpose of the investigation, the sample size, whether individual or group data are desired, the extent of intra-individual and between-subject variation of intake, the nutrient that is being studied and the desired precision of the estimate (Bingham, 1987).

The dietary record method is considered to be the most accurate of all available methods and, in the precise weighing modality, is taken to represent the "gold" standard. It is therefore often used as the reference method for validating more expedient methods. However, because of its intrusiveness in the life of the subjects, this approach may induce changes in their eating behaviour. There is a trade-off, therefore, between the accuracy of this method and the risk of systematic bias of the results.

Chemical analysis of diets

This method requires the subject to provide a duplicate of all foods consumed, including beverages, over the stipulated number of days. The duplicates are collected in an appropriate, refrigerated wide -neck container. There are different ways of collecting dietary duplicates: one modality is to collect the exact weighed replica of all the foods contextually to their consumption, a second modality is to collect an aliquot of each of the servings, and a third is to reconstruct a posteriori, on the basis of the records of a dietary survey, aliquots of all raw food ingredients that have been eaten by the subject. At the end of the survey, the container is brought to the laboratory, weighed, homogenized and chemically analysed for the nutrient of interest.

The dietary duplicate method is mentioned here only for the sake of completeness, as it is seldom used outside epidemiological studies owing to its inconvenience and cost. It was used in the Seven Countries Study, where there was the need not only to check the quality of the dietary surveys and their cross-country comparability, but also to obtain highly accurate and comparable information on the nutrient of interest, namely the fatty acid composition of the diet (Keys, 1966).

Strengths and weaknesses of the individual dietary survey approach

None of the methods used to assess dietary intakes is totally free from inherent errors, and none can provide a fully accurate measure (Beaton, Burema and Ritenbaugh, 1997; Bingham, 1987; Kaaks and Riboli, 1997). This consideration applies equally to individual as well as to household surveys of dietary intake, as it is reasonable to believe that most of the errors of the individual survey methods are shared by household surveys. In the case of individual surveys, much work has been done to analyse the errors, and appreciable progress has been made in developing statistical techniques for adjusting or otherwise coping with these errors. A similar evaluation has not yet been carried out for all other dietary survey approaches, and therefore the nature, direction and dimension of their errors are less well understood, thus representing a major limitation for food security monitoring and surveillance purposes.

The growing awareness of the limitations of dietary surveys in generaland more specifically of those that estimate individual dietary intakes, as in epidemiological research of dietary risk factors - has in recent years produced an ample body of research aiming to enhance the performance of the various methods. One result of this increased interest has been the development of alternative techniques for correcting or reducing the errors most commonly encountered and for calibrating the results. Much attention has also been paid to internal and external validation of the various survey tools, as has been strongly recommended as a means to enhance the quality of data obtained from dietary survey (Beaton, Burema and Ritenbaugh, 1997; Black et al., 1997; Buzzard and Willet, 1994; Kaaks, 1997; Kipnis, Carroll and Freedman, 1999; McKeown et al., 2001; Willett and Sampson, 1997).

Dietary survey errors can be random (reduced reliability) or systematic (bias). Random errors generate larger total variances, reduce the statistical power to detect an association between intake and disease, and may attenuate or inflate the estimates of the possible association (Tarasuk and Beaton, 1992). One possible consequence of random errors is the overestimation of the prevalence of inadequate intake owing to a larger lower tail of the distribution. Systematic errors represent a more serious challenge and a greater hazard than random errors, as they can alter the results, and very little can be done to correct for their effect. The structure of the errors differs according to the type of survey method - those relying on recall or the ability to provide reliable estimates of usual eating habits tend to be more prone to systematic errors, while other methods that rely on direct measurement are more often associated with random errors. A brief review of the errors of dietary surveys is presented in the next several sections. This review is largely based on the work done by Beaton, Burema and Ritenbaugh (1997) and Bingham (1987).

Table 2 lists the main sources of error associated with the various types of individual food intake surveys. Some of these errors are shared by all methods, for example the quality of the tables of food composition on which rests the ultimate precision for the estimation of energy and nutrient intake. Other errors are specific to one or another of the methods. A larger number of potential errors are associated with the FFQ and the dietary history method, while the weighed record method appears the least affected.

TABLE 2. SOURCES OF ERROR IN THE METHODS FOR ASSESSING INDIVIDUAL DIETARY INTAKES

Source of Error

Weighed Food Records

Estimated Food Weight Records

24-hour Recall

Dietary History and FFQs

Food tables/recipe books

+

+

+

+

Food coding

+

+

+

+

Wrong weight of foods

-

+

+

+

Reporting error

-

-

+

+

Variation of diet with time

+

+

+

-

Wrong frequency

-

-

-

+

Modified eating pattern

±

±

-

-

Response bias

±

±

±

±

Sampling bias

+

+

+

+

Adapted from Bingham (1987) with permission.

Random errors

One of the main sources of random errors is the day-to-day variability of intakes. Such variability affects the probability that the estimated individual intake represents the true habitual intake and has important implications for the ability to estimate individual food intake (Bingham, 1987). Within-individual variation may be reduced and estimates of intake may be made more reliable by increasing the number of food-recording days for each individual (Tarasuk and Beaton, 1992). Table 3 demonstrates how the average within-individual variation percent and the subsequent standard error percent vary by nutrient type, ranging from a minimum of 23 percent for energy to a maximum of 60 percent variation for vitamin C. The number of recording days necessary for a ± 10 percent precision is greater for the nutrients with higher within-individual variation percents and lowest for macronutrients (Bingham, 1987). A greater precision of the estimate is of major import for detecting with confidence the existence of differences between groups of individuals. The implications of a large within-individual variability are shown in Table 4, which demonstrates the carryover effect of random error generated by the within-person variability on the categorization of the distribution of intakes into tertiles. The data used to generate the specific example have been derived by Bingham (1987) from surveys on adults. As can be seen, the variability differs not only between nutrients but also between surveys and age groups, thus making it difficult to establish a general rule that is universally applicable.

TABLE 3. NUMBER OF FOOD RECORD DAYS NECESSARY FOR VARIOUS NUTRIENTS TO BE WITHIN ± 10 PERCENT OF AVERAGE

Item

Average within- person variation percentage

percentage standard error of a 7- day record average

Number of days for records to be within ± 10 % of average

Energy

23

9

5

Carbohydrate

25

9

6

Protein

27

10

7

Fat

31

12

10

Dietary fibre

31

12

10

Calcium

32

12

10

Iron

35

13

12

Thiamin

39

15

15

Riboflavin

44

17

19

Cholesterol

52

20

27

Vitamin C

60

23

36

Adapted from Bingham (1987) with permission.


TABLE 4. NUMBER OF FOOD RECORD DAYS NECESSARY TO CORRECTLY CLASSIFY 80 PERCENT OF MEN INTO THIRDS OF THE DISTRIBUTION (P<0.05)

Item

British Civil Servants

Random selection of British men

Random selection of Swedish men

Energy

7

5

7

Protein

6

5

7

Fat

9

9

7

Carbohydrate

4

3

3

Sugar

2

2

-

Dietary fibre

6

10

-

P:S ratio

11

-

-

Cholesterol

18

-

-

Alcohol

4

-

14

Vitamin C

-

6

14

Thiamin

-

6

15

Riboflavin

-

10

-

Calcium

-

4

5

Iron

-

12

9

Adapted from Bingham (1987) with permission.

Ways to cope with random errors are available, such as increasing the number of days or the number of subjects. The choice of one or the other approach rests on the logistics and costs of recruiting more individuals as opposed to extending the duration of the survey. For example, Beaton calculated that with a one - day survey to estimate the group average for energy intake, increasing the group size from 50 to 200 subjects would halve the 95 percent confidence range, from ± 10 percent to ± 5 percent (Beaton et al., 1979).

Other approaches for removing or reducing the day-to-day effect have also been developed with the use of sophisticated statistical methods (Aickin et al., 1994; Carriquiry et al., 1994; Subcommittee on Criteria for Dietary Evaluation, National Research Council, 1986). Methods to remove the effect of day-to-day variation were applied to data from a survey to estimate vitamin A intake in non-pregnant women, showing that the prevalence of inadequate intake was estimated at 16 percent by traditional approaches but was reduced dramatically when two adjustment methods (Carriquiry et al., 1994; Subcommittee on Criteria for Dietary Evaluation, National Research Council, 1986) were applied to reduce the day- to-day variation (to 10 percent and four percent respectively).

Systematic errors

Under-reporting is a well -known source of systematic error in individual dietary surveys (Black, Welch and Bingham, 2000; Høidrup et al., 2002; Mensink, Haftenberger and Thamm, 2001). The phenomenon of under-reporting is quite widespread, as shown in Table 5 (Black, 2000). Under-reporting has been described for large surveys in the United States (Briefel et al., 1997; Mertz et al., 1991) and elsewhere (Howat et al., 1994; Livingstone et al., 1990; Livingstone, Prentice and Coward, 1992; McDowell, Briefel and Alaimo, 1994). The degree of underreporting varies widely between surveys but also within single surveys by age, racial group, degree of overweight and other factors. Beaton estimated that there was a 25 -30 percent under-report in the United States Department of Agriculture (USDA) Continuing Survey of Food Intakes by Individuals (CSFII) - 85 (Beaton, Burema and Ritenbaugh, 1997). While, in the past, it was thought that obese persons and women were the most systematic under-reporters (Bandini et al., 1997; Beaton, Burema and Ritenbaugh, 1997; Bingham and Day, 1997; Black et al., 1993; Briefel et al., 1997), it has now been established that the phenomenon is much more widespread and affects all ages and both sexes, and that other subgroups of the population besides the obese are also affected, such as athletes. Table 6 demonstrates under-reporting of energy intake by sex and weight status, using data from the National Health and Nutrition Examination Survey (NHANES). Figure 1 shows the results of a recent meta-analysis on the validation of energy intake by means of the measure of energy expenditure (data from Hill and Davies, 2001). Under-reporting is mostly associated with the dietary history method, the FFQ and the 24 -hour recall. Methods have been proposed for identifying energy under-reporters on the basis of cutoff points for multiples of the BMR (Black, 2000; Goldberg et al., 1991). The problem, however, cannot be fully resolved by any of the available methods and might even be aggravated by them (Beaton, Burema and Ritenbaugh, 1997; Flegal and Larkin, 1990). In fact, memory may be not only quantitatively imperfect but also selective, with people tending to report lesser consumption of foods to which social stigmas are attached (e.g. high-fat or high- sugar foods). As a consequence, the systematic errors associated with under-reporting are usually compounded by a non-proportional scaling bias of various nutrients (Kaaks, Slimani and Riboli, 1997).

TABLE 5. OVER-AND UNDER-REPORTING OF ENERGY INTAKE IN FREE-LIVING ADULTS*

Reporting category

Men %

Women %

Total

Over-reporters

5

4

4

Under-reporters

28

38

34

Acceptable-reporters

67

59

62

*As evidenced by the concurrent measure of energy expenditure using the doubly labelled water method.
Adapted from Black (2000) with permission


TABLE 6. UNDER-REPORTING OF ENERGY INTAKE IN NHANES 1988 - 1991, BY SEX AND WEIGHT STATUS

Category

Energy intake/basal metabolic rate (estimated)

Low weight

Middle weight

Overweight

Men

1.75

1.54

1.28

Women

1.62

1.32

1.09

Adapted from Briefel, et al. (1997) with permission by the American Journal of Clinical Nutrition. © Am. J. Clin. Nutr. American Society for Clinical Nutrition.

FIGURE 1. RECENT EVIDENCE OF OVER- AND UNDER-REPORTING OF ENERGY INTAKE

Data source: Hill et al., 2001

Note: each column corresponds to the mean energy intake of one study compared with the energy expenditure measured concurrently by the doubly labelled water method.

No statistical method exists that can correct fully for reporting problems, and the only effective way is to apply extra care when using survey methods that lend themselves to introducing differential biases in nutrient intakes (Beaton, Burema and Ritenbaugh, 1997; Brown et al., 1994). Internal as well as external validation assessments have been advocated as an important means for ensuring better-quality data from dietary surveys (Beaton, Burema and Ritenbaugh, 1997).

The problem with tables of food composition

As mentioned earlier, a separate but common step in all dietary surveys is the calculation of energy and nutrient content of the foods consumed. This is performed by consulting tables of food composition. Independently from the quality of the dietary consumption data, this last step is most critical as it has the power of introducing a new set of errors, most likely of a systematic nature. Many countries have their own national food composition tables, but often they consist of little more than a compilation of literature data within some cases - added analytical data of few national staple foods. Often, they are severely outdated and do not keep track of the compositional changes of the food commodities taking place over the years. These changes concern not only industrially prepared foods but also the raw ingredients such as the fatty acid composition of vegetable oils (for example, the earlier high content of erucic acid content in rapeseed oil and its later substitution with monounsaturated fatty acids in canola oil, the decrease of trans-fatty acids of margarines) or the fat content of pork, poultry and fish. Such and other often unperceived changes in the composition of foods are continuously brought about by the development of plant breeding techniques, by the genetic selection of crops and by the evolution of animal husbandry and agricultural practices. The problem is further compounded by the evolution over time of the food commodities available to the consumer, as a large number of newly formulated foods are placed on the market every year by the food industry.

The problem of conversion of foods to their energy and nutrient composition becomes even more serious when the data are in the form of aggregated food categories (such as the data obtained by household budget surveys and food balance sheets). In this case, a further element of uncertainty is introduced by the need to assume a given proportion of the individual food commodities that enter each of the categories. Individual dietary surveys are not immune from this type of problem either. This applies to the FFQ, the 24 - hour recall and the dietary history methods, where prepared/cooked foods are recorded as consumed, and ingredients such as the fat or sugar content can only be back-calculated from standard recipes. The problem becomes more acute in those societies where domestic preparation of foods prevails. For example, in recent surveys conducted in Greece, only approximately one -third of the items listed in the FFQ were prepared foods (Ferro-Luzzi, James and Kafatos, 2002).

Those conducting dietary surveys are usually well aware of the errors potentially introduced by the lack of nationally relevant tables of food composition, especially but not solely in less advanced regions of the world (see Kigutha, 1997), but the problem is unlikely to have a simple solution given the costs and expertise required to develop or update a reliable dataset of food composition.

The problem of food composition tables is not specific to individual dietary surveys but is shared with all available approaches for estimating food security based on the assessment of dietary intakes. There is a plurality of datasets, mostly outdated and obtained from compilation of literature sources, that may not be nationally relevant. A necessary step for the purpose of global estimate of food security is the cross-country comparability of national datasets with the use of standard ways to classify foods, for example as simple commodities, categories of aggregated foods, or recipe books (Leclercq, Valsta and Turrini, 2001)

Strengths of the individual dietary survey approach

The unique strength of the individual dietary survey method is that its results can be validated. The validation of energy intake is performed by means of the doubly labelled water method, which measures energy expenditure. This is a major breakthrough, as the method is fairly robust, non-invasive and allows measurement of energy expenditure over a period of about seven to 12 days in free -living, unrestrained individuals, requiring a minimum of cooperation. Under conditions of body weight stability, energy expenditure can be equated to energy intake. The doubly labelled water method uses stable isotopes of oxygen and hydrogen, and exploits the equilibrium maintained in the body between the O 2 of body water and that of expired CO 2 by the action of the enzyme carbonic anhydrase. Validation of the survey results relative to the intake of several nutrients is also possible, although it is more invasive than that for energy, requiring the collection and biochemical analysis of body fluids (blood and/or urine).

TABLE 7. EXAMPLE OF THE EFFECT OF UNEQUAL INTRAHOUSEHOLD FOOD DISTRIBUTION ON THE ESTIMATE OF HOUSEHOLD FOOD SECURITY, COMPARING THE HOUSEHOLD APPROACH WITH THE INDIVIDUAL INTAKE APPROACH

Normative energy requirement

Hypothetical householda

Hypothetical householdb

Household member

Energy intake

Adequacy of energy intakec

Energy intake

Adequacy of energy intake

Age in years

Sex

kcal/day

kcal/day

%

kcal/day

%

30

M

3 500

3 000

86

3 500

100

30

FLd

3 500

2 000

57

3 500

100

10

F

2 000

2 000

100

1 300

65

5

F

1 800

1 800

100

1 000

56

1

M

1 000

1 000

100

500

50

Household total (kcal /day)


11 800

9 800


9 800


Adequacy of household energy intakea




83


83

Adequacy of household energy intakeb




89


74

a based on the conventional approach of using aggregated household consumption data.
b based on the individual intake approach of using actual intake data for each household member.
c energy intake/energy requirement.
d Lactating female.
Source: Ferro-Luzzi, Norgan and Paci (1981).

Another major advantage of the individual dietary survey approach is that it can provide information on the allocation of food within the household. Any other approach must rely on the assumption that food is equitably distributed within the household and that each member receives food in proportion to their energy and nutrient requirements. There is little information available on this topic and the literature is far from conclusive. However, although inconsistent evidence has been previously published (Abdullah and Wheeler, 1985; Anonymous, 1992; Chaudhury, 1988; Chinwe and Nnanyelugo, 1989; Ferro-Luzzi, Norgan and Paci, 1981; Gittelsohn, 1991; Haddad and Bouis, 1991; IFPRI, 1992), a recent review of the subject has confirmed that food may be unequally accessed by the various members of a household (Haddad et al., 1996). The contradictory evidence on the equity of food distribution within the household may well reflect diverse sociocultural contexts across the studies. It may also reflect the status attributed to individual members of the household based on gender, age, religion or social position (such as first or second wife in a polygamous household). Furthermore, it is reasonable to believe that the level of household food insecurity might play an additional important role in food distribution patterns - the greater the food insecurity, the more likely the inequitable allocation of scarce resources, possibly favouring the more vulnerable household members such as the younger children or instead the most valuable members in terms of their income/work productive potential. The gender of persons allocating food in the family might also play a role in redirecting insufficient family resources. The picture is further complicated by possible large fluctuations in the level of food security, for example owing to seasonal food shortages in rural communities of developing countries (Ferro-Luzzi et al., 2001), causing a periodic shift of the intrahousehold distributive allocation of food. The implications in terms of food security assessment of unequal distribution of food in the family are appreciable, as shown in Table 7. In this table, the effect of the diverse allocations of the same amount of food energy in a hypothetical household has been modelled to show that the conventional approach is not only erroneous but also insensitive to the prevalent pattern of inadequacy: it will overestimate the level of household food security if the younger members of the households are food insecure but underestimates it if the food insecure are the adults (Ferro-Luzzi, Norgan and Paci, 1981).

Summary

Figure 2 provides a graphic, non-quantified overview of the respective strengths and weaknesses of the various dietary intake survey methods. It shows that the quality of the information retrieved by dietary surveys depends to a large extent on the method adopted, and tends to be inversely proportional to the expediency and cost of the method. Thus, the precise weighed records and the chemical analysis of dietary duplicates have the highest costs and the lowest compliance, expedience and level of representativity. However, the information obtained from these types of surveys will have the highest degree of precision with comparably fewer risks of systematic and random errors. The FFQ and the household surveys are placed at the opposite end of the gradation. One of the main advantages of the individual dietary survey approach is that the error structure has been thoroughly investigated, and ways to deal with them have been developed. A second advantage is the possibility of independent validation by use of energy expenditure measurements. A third major advantage is the unique capacity of individual intake surveys to reveal uneven intrahousehold access to food. These types of surveys, besides being a primary source of data, also represent a valuable tool for validation of any other method and are thus often considered to be the gold standard.

FIGURE 2. COMPARISON OF METHODS FOR MEASURING DIETARY INTAKE FOR THE PURPOSE OF ASSESSING FOOD SECURITY*

The feasibility of individual food intake surveys in developing countries

Most dietary methods for estimating food intake at the individual level have been designed for operating in technologically advanced societies. The most expedient methods rely for their success on advanced technological inputs and support, such as access to telephone lines, special electronic computerized scales, bar-coding, the nutrition evaluation scale (Fong and Kretsch, 1990) and videotaping of meals (Brown et al., 1990; Kretsch and Fong, 1993). These are obviously too cumbersome or otherwise inconvenient for use in developing countries, or even in rural areas of some developed countries, because of limited technological support or low literacy.

The particular sociocultural and demographic characteristics of surveyed communities need to be taken into account when conducting dietary surveys, as what works well for one location may not be applicable to another. Adaptations of the methods for employment under diverse culture -specific settings should be validated against a reliable and independent standard (Jerome, 1997; Teufel, 1997). This has seldom been done, however. In 1997, out of the 84 studies included in the register of dietary assessment created for calibration- validation purposes, only 3 percent had been carried out in Asia and South America, and none in Africa (Thompson et al., 1997). To improve the feasibility of conducting food intake surveys in developing countries, adaptations must be made to the instruments and methodologies that have been developed for use in developed countries. For example, locally adapted FFQs, combined with region-and socio-economic-specific portion sizes and recipes, have been successfully administered by trained local personnel in rural Mexico (Romieu et al., 1997) and Mali (Torheim et al., 2002), and 24 -hour recall surveys were carried out in Russia on schoolchildren (Martinchik et al., 1997) and in China on a nationally representative population sample (Ge, Zhai and Wang, 1997). In Malawi and Ghana, the diets of women and preschool children were assessed by the 24 -hour recall (Ferguson et al., 1989; Ferguson et al., 1995; Ferguson, Gibson and Opare -Obisaw, 1994). The diets of Philippino, Thai and Gambian pregnant women were assessed by the precise weighing method (Durnin, 1987), and seasonal variations in dietary intakes were assessed by weighing among young adult women in a remote rural area of central Ethiopia (Ferro-Luzzi et al., 1990) and among adult men and women in New Guinea (Norgan, Ferro-Luzzi and Durnin, 1974). However, except for the China survey, most of these studies were conducted for research purposes on small groups of selected individuals, and not for food security assessment.

The difficulties and constraints encountered in conducting a dietary survey in developing countries come from many sources: the culturally specific ways of purchasing, storing, cooking and sharing food; the energy and nutrient composition of local foods and recipes (a large proportion of the foods are home -grown or gathered in the bush); the level of literacy in the society; culturally derived reluctance to allow strangers to handle foods destined for home consumption; and local taboos or other rules that cause embarrassment and shyness (Kigutha, 1997). In general, diets in these countries have little day-to-day variation, thus suggesting that a smaller number of survey days may be required to obtain accurate information of habitual diets. However, often this advantage is largely outweighed by difficulties in assessing individual intake in cultures where several household members may eat from a common plate, or several units of polygamous households may share the same pot.

It is important to have an idea of cultural or environmental factors that influence intra-individual and interindividual variability in food intake in area where food intake surveys are being conducted. For example, several rounds of surveys should be made in those regions of the world where seasonal variations in access to food still cause appreciable fluctuations in the quantity and quality of foods consumed (Bates, Prentice and Paul, 1994; Ferro-Luzzi et al., 1990, Ferro-Luzzi et al., 2001).

Principles for estimating adequacy of food intake and food security

The estimate of the adequacy of food intake - and, thereof, food security - is obtained from the comparison of the intake of energy and nutrients with their respective requirements. The procedure for energy differs from that for protein and other nutrients for reasons that will be given briefly below but are described in detail in the 1985 report of the Expert Consultation Meeting on Energy and Protein Requirement (FAO/WHO/UNU, 1985) and reviewed in 2002 (FAO/WHO/UNU, in preparation).

For protein and micronutrients, the requirement is represented by the safe level of intake, which is set at two standard deviations above the average physiological needs. The safe level thus takes into account the interindividual variation in requirement while remaining well below the level of intake capable of causing harmful effects. The concept of the safe level does not apply to energy because there is only one single point at which any given individual is in energy balance; that point represents the requirement (Figure 3). Any excess or deficit in energy intake relative to that point results in detrimental changes in energy equilibrium, with deposition as body fat if energy is in excess, or in a negative energy balance. A sustained negative energy balance sets into action a number of responses directed at reestablishing energy equilibrium at the new and lower level of energy intake. These responses have been discussed at great length in the literature (Ferro-Luzzi, 1986; Ferro-Luzzi et al., 1997; James, 1989; Shetty, 1993, 1999; Water- low, 1989) and are considered to be associated with detrimental functional and/or health outcomes. These adverse outcomes include low birth weight and growth retardation in children, and smaller attained body mass, loss of body weight, reduced physical activity and lower productivity in adults (Ferro-Luzzi and Branca, 1998; Ferro-Luzzi et al., 2001; Haddad and Bouis, 1991). A very modest metabolic adaptation, leading to increased efficiency in energy handling at the cellular level - and thus to some reduction of energy requirement - might also occur (Ferro-Luzzi, 1986), probably mediated by hormonal mechanisms (Shetty, 1999), but the size of the energy thus spared is small and has little impact on overall energy requirement.

FIGURE 3. ENERGY (AVERAGE) AND PROTEIN (AVERAGE +2 SD) REQUIREMENTS

In the context of food security assessment, the information needed is relative to the group rather than to the individual. The group distribution of the individuals' energy and nutrient requirements is assumed to be normal. Thus, the group requirement for energy is the average of the individuals' requirement, and for protein and other nutrients, it is the average +2 SD. The estimate of the group adequacy of energy and nutrient intake is made on the basis of a probabilistic approach, applicable whether the information on intake is at the individual or at the group/household or national level. This approach provides an estimate of the proportion of individuals that are likely to have inadequate intakes, but does not specify who those individuals are. These principles have been described in detail in the 1985 WHO report and have been retained in the forth- coming revision of energy requirement 2002 (FAO/WHO/UNU, in preparation). An example of the probability approach for protein, taken from the 1985 report, is given in Table 8.

Although food security includes also protein and other nutrients, energy is undoubtedly the main parameter and occupies a pivotal position of crucial importance. In fact, while it is recognized that the sufficiency of energy may not ensure that the other requirements are equally met, especially for micronutrients (vitamins and trace elements), it is safe to assume that a diet that is adequate in all nutrients is unlikely to be energy-deficient. For energy, the adequacy of intake is expressed on the basis of a comparison between the two distributions of intake and requirement. The difference between the two curves provides evidence of inadequacy or excess, and thus of food security. However, to obtain a quantitative estimate of the prevalence of inadequate or excessive intake, two types of information are needed, namely the strength of the association between the intake and requirement, and the variability of the energy requirement. The link between expenditure and intake may be assumed to be fairly robust and tight. However, while exposure to sustained insufficiency of dietary energy will lead eventually to curtailment of energy expenditure, the converse does not always hold: extended periods of intake in excess of expenditure can, and indeed do, occur without a concurrent increase in energy expenditure. As for the variance of energy requirement, an estimate might indeed be derived from detailed information on physical activity, but this is seldom available. Nevertheless, the BMR has an interindividual coefficient of variation of around ± 12.5 percent. On this basis, Beaton has estimated that, in a population group, the variance of total energy expenditure for activity levels ranging from 1.5 to 2.1 BMR multiples may approximate 23 percent, i.e. double that of BMR alone (Beaton, Burema and Ritenbaugh, 1997). Recent progress in the measure of energy expenditure by means of doubly labelled water has provided some support to this assumption (Butte, 2001; Ferro-Luzzi, 2001; Torun, 2001).

TABLE 8. PROBABILITY OF INADEQUACY OF OBSERVED PROTEIN INTAKES WHEN EXPRESSED AS GRAMS PER KILOGRAM PER DAY

Observed intake of an individual in relation to requirement

Proportion (%) of a group of similar individuals whose intake is inadequate

Probability that an individual's intake is inadequate

25% above average requirement

2.3

0.02

12.5% above average requirement

15.9

0.16

At average requirement

50.0

0.50

12.5% below average requirement

84.1

0.84

25% below requirement

97.7

0.98

Adapted from FAO/WHO/UNU (1985).

One of the questions that need to be addressed in the context of a global evaluation of food security is the cross-country comparability of the information obtained by the various approaches. In considering the individual survey approach, the extent to which the results are comparable is likely to be relatively high and might be enhanced further by detailed descriptions of the survey methodology, calibration procedures and food composition tables used. These considerations are particularly relevant for methods characterized by a large subjective component, such as the FFQ or dietary history where food quantities are estimated, but they apply also to other approaches such as the household surveys. In all these cases, if proper attention is paid to the calibration procedures and to the quality and comparability of food composition, especially for the average composition of food categories, then it is safe to assume that the prevalence of food insecure persons or households can be compared across countries.

With respect to the assessment of food security of large representative population groups at the subnational or national level, particularly in developing countries, the most accurate methods for assessing individual dietary intakes (weighed dietary survey, chemical analysis of dietary duplicates) automatically disqualify themselves because of the scale of the population to be surveyed and the costs and logistics of conducting these types of surveys. The most reasonable candidate for the assessment of food security in nations across the world might be the FFQ because of its relative cheapness, apparent simplicity and expediency. However, serious prior attention must be given to (1) developing culturally competent questionnaires; (2) developing locally pertinent food composition tables and recipe books; (3) validating questionnaires under the actual conditions of the survey; (4) developing methods to deal with issues such as eating from common plate and seasonal shifts in food intake and energy requirements; (5) considering the need for cross-country comparability of the information collected; and (6) calculating whether the financial costs of the operation are commensurate with the available resources.

In conclusion, the above considerations reveal that it would be difficult, if not impossible at present, to derive a measure of household food security from the assessment of food intake of one individual member. Obviously, if all the household members were surveyed, the information would be highly informative and could possibly represent a gold standard. But costs and other considerations make this solution highly improbable. An alternative approach might be to identify within each community or society the most "food insecure" category of household members and assess the food intake of only these individuals. Ideally, the sensitivity and specificity of using such a category of individuals as indicators of community events could be assessed to order to increase the reliability of the extrapolation. The validity of the choice of these indicator persons might be reinforced by the concurrent use of suitable anthropometric indicators, measures that are simple to perform and are considered reliable and relatively easy to interpret in adults (see Bailey and Ferro-Luzzi, 1995; paper by Shetty in this series). Once reliably identified, the use of a key member as indicator of the household level of food security has a number of advantages over the alternative approach that estimates food security on the basis of the average household food intake. With the latter approach, the overall intake of the household is compared with its overall requirement, possibly leading to an underestimation of food insecure households, as an uneven food allocation that privileges the adults might mask the borderline or deficient intake of the younger members (Ferro-Luzzi, Norgan and Paci, 1981).

Costs and other features of individual survey methods

Most of the individual survey methods are expensive to conduct (Byers et al., 1997), and even for the less expensive versions such as the food frequency method, the extra work needed to develop culturally competent survey methods (Teufel, 1997) and to validate the adapted methods under the local conditions (Jerome, 1997) has a non-indifferent cost.

Given the flexibility of the dietary survey tool and the many variations of the methods that exist, it is difficult to provide an indication of the actual costs of carrying out an individual dietary intake survey under many different local conditions. A range of US$3.50 (FFQ including optical scanning) to US$120 per person (a four-day food record) has been given recently for dietary surveys conducted in the USA (Drewnowski, 2001), but it is obvious that the cost of the survey will depend on the country and local context of the survey. In general, however, it is reasonable to expect that the cost will be indirectly proportional to the validity and precision of the method used, being lowest with the expedient approach of the FFQ, higher with 24 -hour recall, and highest in the weighing and chemical analysis approach (Figure 2). Food frequency surveys, used mostly for epidemiological purposes, are advantageous because a large number of subjects can be assessed at reduced costs, time and number of defaulters with respect to other methods. This type of survey requires less effort from the subject and limits intrusiveness, as it can be carried out by mail rather than by face-to-face interviews. However, the advantages of doing FFQ surveys in developed countries most likely will not be seen when conducted in non-western societies. There is also the problem of comparability of the results of surveys conducted in culturally different communities, as no information may exist on the performance of the survey tool either within the same country or, more so, between diverse countries and regions of the world.

When individual dietary surveys are conducted in developing countries, it is generally necessary to rely to a large extent on local personnel who are often specifically recruited and trained. Salaries and the travel costs required to reach communities at higher risk of food insecurity, such as rural and isolated communities, need to be taken into account when estimating the overall costs of the survey. Seasonal variation in food security of these communities is also likely to exist (Ferro-Luzzi and Branca, 1993), both for energy and for micronutrients, thus requiring methods such as repeated surveys over different seasons that are capable of capturing the extent and severity of the phenomenon (Ferro-Luzzi et al., 2001).

Overall conclusions

The present review of methods to assess individual dietary intakes has been conducted, keeping in mind the global context in which a food security information system is expected to function, with special emphasis on the feasibility of using these methods in developing countries. A synthesis of the strengths and weaknesses of the individual dietary survey methods is displayed in Table 9.

Several different methods exist for assessing individual dietary intakes, applicable under widely differing circumstances. The data collected allow analysis at a very fine level of disaggregation, providing a strong operational basis for policy interventions. Individual dietary survey methods have undergone the closest methodological scrutiny of any other dietary survey approach; therefore, their error structure is better documented, and they are amenable to further improvements. Validation and calibration methods are available to ensure cross-country comparability of results.

Of the individual intake approaches, the FFQ, because of its low cost, apparent simplicity and expediency, appears to be the most suitable approach to employ for the assessment of food security in nations across the world. However, this is the least robust method, and much work still needs to be done to adapt it to diverse geographic and cultural situations.

Limitations to the use of individual dietary survey method for the assessment of food security at a country level are also noted. The first limitation is represented by the scale of the population to be surveyed, and therefore by the cost and the logistics involved. Thus, although the individual dietary intake method has the potential as a means to measure national food security, cost considerations should be kept upfront. Given the wide range of cost per survey unit of the diverse methods, the FFQ is the most likely candidate, to be combined with appropriate sampling of the study individuals.

TABLE 9. COMPARATIVE STRENGTHS AND WEAKNESSES OF INDIVIDUAL DIETARY SURVEYS FOR MONITORING FOOD SECURITY

Strengths

Weaknesses

Ample methodological flexibility
High quality of the information collected
The error structure is well understood
Results can be ef fectively validated
Can reveal unequal intrahousehold food distribution
Can capture subgroups at risk of food insecurity (elderly, children, etc.)
Can identify target foods for target people
Usable for international standardization

Requires adaptation to cultural contexts
Logistically complicated and relatively costly
Requires trained staff
Requires collaboration from subjects
Risk of under-reporting

The issue of representativeness of individuals for the household or community has been discussed, and with proper sampling criteria, the information generated should provide a reasonably accurate picture of the prevalence of food insecurity in any given population. This approach might be particularly suitable when intrahousehold distribution of food has a large variance owing to inequitable food distribution.

Large -scale individual dietary intake surveys may not be the best way to monitor trends of food security over time because of the associated costs and logistical difficulties. Only a very small number of large -scale national surveys have been carried out in the international setting, and they are rarely undertaken at regular intervals for the purpose of monitoring changes in food security. Perhaps the most important role of individual intake survey methodology might be its use as a validation tool for the food security measurement approaches that are routinely used, such as the food balance sheet method and household budget surveys. Data from individual intake surveys could be used to validate the household approach, as both of these methods can provide disaggregated data by geographic and demographic characteristics, unlike the food balance sheet method.

Acknowledgements

The author wishes to thank E. Toti, and P. Ferranti for assistance in preparing this document.

References

Abdullah, M. & Wheeler, E.F. 1985. Seasonal variations, and the intra -household distribution of food in a Bangladeshi village. Am. J. Clin. Nutr., 41: 1305 -1313.

Aickin, M., Ritenbaugh, C., Woolf, M.S., Graver, E., Rees-McGee, S., McGee, D., Sampliner, R. & Alberts, D. 1994. A general method for adjusting dietary data to obtain stable estimates of usual intakes. Am. J. Clin. Nutr., 59: 303S (abstract).

Anonymous. 1992. Women on the short end of the loaf: intrahousehold food allocation among members of poor Nepalese families. Nutr. Rev., 50(8): 236 -237.

Bailey, K.V. & Ferro-Luzzi, A. 1995. Use of body mass index of adults in assessing individual and community nutritional status. Bull. WHO, 73: 673 -680.

Bandini, L.G., Cyr, H., Must, A. & Dietz, W.H. 1997. Validity of reported energy intake in preadolescent girls. Am. J. Clin. Nutr., 65 (Suppl. 4): 1138S -1141S.

Baranowski, T. & Domel, S.B. 1994. A cognitive model of children's reporting of food intake. Am. J. Clin. Nutr., 59 (Suppl. 1): 212S -217S.

Bates, C.J., Prentice, A.M. & Paul, A.A. 1994. Seasonal variations in vitamins A, C, riboflavin and folate intakes and status of pregnant and lactating women in a rural Gambian community: some possible implications. Eur. J. Clin. Nutr., 48: 660 -668.

Beaton, G.H., Milner, J., Corey, P., McGuire, V., Cousins, M., Stewart, E., de Ramos, M., Hewitt, D., Grambsch, P.V., Kassim, N. & Little, J.A. 1979. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am. J. Clin. Nutr., 32: 2546 -2559.

Beaton, G.H., Burema, J. & Ritenbaugh, C. 1997. Errors in the interpretation of dietary assessments. Am. J. Clin. Nutr., 65: 1100 -1107.

Bingham, S.A. 1987. The dietary assessment of individuals; methods, accuracy, new techniques and recommendations. Nutr. Abstr. Rev., 57: 705 -742.

Bingham, S.A. & Day, N.E. 1997. Using biochemical markers to assess validity of prospective dietary assessment methods and the effect of energy adjustment. Am. J. Clin. Nutr., 65 (Suppl. 4): 1130S-1137S.

Black, A.E. 2000. The sensitivity and specificity of the Goldberg cut-off for EI: BMR for identifying diet reports of poor validity. Eur. J. Clin. Nutr., 54: 395 -404.

Black, A.E., Prentice, A.M., Goldberg, G.R., Jebb, S.A., Bingham, S.A., Livingstone, M.B. & Coward, W.A. 1993. Measurements of total energy expenditure provide insights into the validity of dietary measurements of energy intake. J. Am. Diet. Assoc., 572 -579.

Black, A.E., Bingham, S.A., Johansson, G. & Coward, W.A. 1997. Validation of dietary intakes of protein and energy against 24 hour urinary N and DLW energy expenditure in middle-aged women, retired men and post-obese subjects: comparison with validation against presumed energy requirements. Eur. J. Clin. Nutr., 51: 405 -413.

Black, A.E., Welch, A.A. & Bingham, S.A. 2000. Validation of dietary intakes measured by diet history against 24 h urinary nitrogen excretion and energy expenditure measured by the doubly-labelled water method in middle-aged women. Br. J. Nutr., 83: 341 -354.

Briefel, R.R., Sempos, C.T., McDowell, M.A., Chien, S., Alaimo, K. 1997. Dietary methods research in the third National Health and Nutrition Examination Survey: underreporting of energy intake. Am. J. Clin. Nutr., 65: 1203 -1209.

Brown, C.C., Kipnis, V., Freedman, L.S., Hartman, A.M., Schatzkin, A. & Wacholder, S. 1994. Energy adjustment methods for nutritional epidemiology: the effect of categorization. Am. J. Epidemiol., 139: 323 -338.

Brown, J.E., Tharp, T.M., Dahlberg-Luby, E.M., Snowdon, D.A., Ostwald, S.K., Buzzard, I.M., Rysavy, S.M. & Wieser, S.M. 1990. Videotape dietary assessment: validity, reliability, and comparison of results with 24-hour dietary recalls from elderly women in a retirement home. J. Am. Diet. Assoc., 90: 1675 -1679.

Burke, B. 1947. The dietary history as a tool in research. J. Am. Diet. Assoc., 1041 -1046.

Butte, N. 2001. Energy requirements of infants. Energy Background Paper No. 4 (June 27 -July 5, 2001). Rome, FAO/WHO/UNU Working Groups on Energy and Protein in Human Nutrition.

Buzzard, I.M. & Willett, W.C., eds. 1994. First International Conference on Dietary Methods: Assessing Diets to Improve World Health. Am. J. Clin. Nutr., 59 (Suppl. 1): 1S -306S.

Byers, T., Serdula, M., Kuester, S., Mendlein, J., Ballew, C. & McPherson, R.S. 1997. Dietary surveillance for states and communities. Am. J. Clin. Nutr. 65: 1210 -1214.

Carriquiry, A.L., Jensen, H.H., Fuller, W.A., & Guenther, P. 1994. Methods for estimating usual intake distributions. Am. J. Clin. Nutr., 59: 305S (abstract).

Chaudhury, R.H. 1988. Adequacy of child dietary intake relative to that of other family members. Food Nutr. Bull., 10: 26 -34.

Chinwe, O.E. & Nnanyelugo, D.O. 1989. Intrafamiliar distribution of food and nutrients in a rural Nigerian population. Ecol. Food Nutr. 23: 109 -123.

Department of Women and Child Development. 1998. India nutrition profile. New Delhi, Ministry of Human Resource Development, Government of India.

Drewnowski, A. 2001. Diet image: a new perspective on the food-frequency questionnaire. Nutr. Rev., 59: 370 -374.

Durnin, J.V. 1987. Energy requirements of pregnancy. An integrated study in five countries: background and methods. Lancet, 14; 2(8568): 895 -896.

Dwyer, J.T. 1994. Dietary assessment. In M.E. Shils, J.A. Olson & M. Shike, eds. Modern nutrition in health and disease, pp. 842 -860. Philadelphia, PA, Lea & Febiger.

FAO/WHO/UNU. 1985. Energy and protein requirements. Report of a joint FAO/WHO/UN ad hoc expert consultation. WHO Technical Report Series, No. 724, Geneva.

FAO. 1998. Guidelines for national FIVIMS. Background and principles, by the Interagency Working Group. Rome (available at http://www.fao.org/docrep/003/x8346E/x8346e01.htm).

Federation of American Societies for Experimental Biology, Life Sciences Research Office. 1995. Third Report on Nutrition Monitoring in the United States. Report No. ES1-IV-3. Washington, DC, Federal Printing Office (executive summary available at www.cdc.gov/nchs/data/misc/tronm.pdf).

Ferguson, E.L., Gibson, R.S., Ounpuu, S. & Sabry, J.H. 1989. The validity of the 24 hour recall for estimating the energy and selected nutrient intakes of a group of rural Malawian preschool children. Ecol. Food. Nutr., 23: 273 -285.

Ferguson, E.L., Gibson, R.S. & Opare-Obisaw, C. 1994. The relative validity of the repeated 24 h recall for estimating energy and selected nutrients intakes of rural Ghanaian children. Eur. J. Clin. Nutr., 48: 241 -252.

Ferguson, E.L., Gadowsky, S.L., Huddle, J.M., Cullinan, T.R., Lehrfeld, J. & Gibson, R.S. 1995. An interactive 24-h recall technique for assessing the adequacy of mineral intakes of rural Malawian women: its advantages and limitations. Eur. J. Clin. Nutr., 49: 565 -578.

Ferro-Luzzi, A. 1986. Range of variation in energy expenditure and scope for regulation. In T.G. Taylor & N.K. Jenkins, eds. Proc. XIII International Congress of Nutrition, pp. 393 -399. London, John Libbey Press.

Ferro-Luzzi, A. 2001. Conceptual framework for estimating energy requirements. Energy Background Paper No. 3. Rome, FAO/WHO/ UNU Working Groups on Energy and Protein in Human Nutrition, June 27 -July 5, 2001.

Ferro-Luzzi, A., Norgan, N.G. & Paci, C. 1981. An evaluation of the distribution of protein and energy intakes in some new Guinean Households. Nutr. Rep. Int., 24: 153 -163.

Ferro-Luzzi, A., Scaccini, C., Taffese, S., Aberra, B. & Demeke, T. 1990. Seasonal energy deficiency in Ethiopian rural women. Eur. J. Clin. Nutr., 44: 7 -18.

Ferro-Luzzi, A. & Branca, F. 1993. Nutritional seasonality: the dimensions of the problem. In S. Ulijaszek & S.S. Strickland, eds. Seasonality and human ecology. 35th symposium volume of the Society for the Study of Human Biology, pp. 149 -165. Cambridge, UK, Cambridge University Press.

Ferro-Luzzi, A., Petracchi, C., Kuriyan, R. & Kurpad, A. 1997. Basal metabolism of weight- stable chronically undernourished men and women: lack of metabolic adaptation and ethnic differences. Am. J. Clin. Nutr., 66: 1086 -1093.

Ferro-Luzzi, A. & Branca, F. 1998. Coping with poverty: the biological impact of nutrition insecurity. In M. Livi -Bacci & G. De Santis, eds. Population and poverty in developing world, pp. 49 -60. Oxford, UK, Clarendon Press.

Ferro-Luzzi, A., Morris, S.S., Taffese, S., Demissé, T. & D'Amato, M. 2001. Seasonal undernutrition in rural Ethiopia: magnitude, correlates, and functional significance. Research Reports Series 118, pp. 1 -74. Washington, DC, International Food Policy Research Institute.

Ferro-Luzzi, A., James, W.P. & Kafatos, A. 2002. The high-fat Greek diet: a recipe for all? Eur. J. Clin. Nutr., 56: 796 -809.

Flegal, K.M. & Larkin, F.A. 1990. Partitioning macronutrient intakes estimates from a food frequency questionnaire. Am. J. Epidemiol., 131: 1046 -1058.

Fong, A.K. & Kretsch, M.J. 1990. Nutrition evaluation scale system reduces time and labor in recording quantitative dietary intake. J. Am. Diet. Assoc., 90: 664 -670.

Ge, K., Zhai, F. & Wang, Q. 1997. Effect of nationality on dietary pattern and meal behavior in China. Am. J. Clin. Nutr. 65: 1290 -1294.

Gibson, R.S. 1990. Food consumption of individuals. In R.S. Gibson, ed. Principles of nutritional assessment, pp. 37 -54. New York, Oxford University Press.

Gittelsohn, J. 1991. Opening the box: intrahousehold food allocation in rural Nepal. Soc. Sci. Med., 33: 1141 -1154.

Gobierno Vasco. 1994. Encuesta de nutricion de la comunidad autonoma del Pais Vasco. Vitoria, Spain, Departamento de Sanidad, Gobierno Vasco.

Goldberg, G.R., Black, A.E., Jebb, S.A., Cole, T.J., Murgatroyd, P.R., Coward, W.A. & Prentice, A.M. 1991. Critical evaluation of energy intake data using fundamental principles of energy physiology: 2. Evaluating the results of published surveys. Eur. J. Clin. Nutr., 45: 583 -599.

Gregory, J., Foster, K., Tyler, H. & Wiseman, M. 1990. The dietary and nutritional survey of British adults. London, HMSO Books.

Haddad, L.J. & Bouis, H.E. 1991. The impact of nutritional status on agricultural productivity: wage evidence from the Philippines. Oxford Bull. Econ. Stat., 53: 45 -68.

Haddad, L., Pena, C., Nishida, C., Quisumbing, A.R. & Slack, A. 1996. Food security and nutrition implications of intrahousehold bias: a review of literature. Discussion Paper No. 19. Washington, DC, International Food Policy Research Institute.

Hill, R.J. & Davies, P.S. 2001. The validity of self-reported energy intake as determined using the doubly labelled water technique. Br. J. Nutr., 85: 415 -430.

Høidrup, S., Andreasen, A.H., Osler, M., Pedersen, A.N., Jørgensen, L.M., Jørgensen, T., Schroll, M. & Heitmann, B.L. 2002. Assessment of habitual energy and macronutrient intake in adults. Comparison of a seven day food record with a dietary history interview. Eur. J. Clin. Nutr., 56: 105 -113.

Howat, P.M., Mohan, R., Champagne, C., Monlezun, C., Wosniak, P. & Bray, G.A. 1994. Validity and reliability of reported dietary intake data. J. Am. Diet. Assoc., 94: 169 -173.

Hulshof, K. 1993. Assessment of variety, clustering and adequacy of eating patterns. Dutch National Food Consumption Survey. Ph. D. thesis. Maastricht, The Netherlands, Rijksuniversiteit Limburg.

International Food Policy Research Institute. 1992. Understanding how resources are allocated within households. IFPRI Policy Briefs 8. Washington, DC.

Irish Nutrition and Dietetic Institute. 1990. Irish national nutrition survey 1990. Dublin, Fineprint.

James, W.P. 1989. Nutritional adaptation and variability. Eur. J. Clin. Nutr., 43: 405 -408.

Jerome, N.W. 1997. Culture -specific strategies for capturing local dietary intake patterns. Am. J. Clin. Nutr., 65: 1166 -1167.

Kaaks, R. & Riboli, E. 1997. Validation and calibration of dietary intake measurements in the EPIC Project: methodological considerations. Int. J. Epidemiol., 26 (Suppl. 1): S15 -S25.

Kaaks, R., Slimani, N. & Riboli, E. 1997. Pilot phase studies on the accuracy of dietary intake measurements in the EPIC project: overall evaluation of results. Int. J. Epidemiol., 26 (Suppl. 1): S26 -S36.

Kaaks, R.J. 1997. Biochemical markers as additional measurements in studies of the accuracy of dietary questionnaire measurements: conceptual issues. Am. J. Clin. Nutr., 65 (Suppl. 4): 1232S -1239S.

Katsouyanni, K., Rimm, E.B., Gnardellis, C., Trichopoulos, D., Polychronopoulos, E. & Trichopoulou, A.1997. Reproducibility and relative validity of an extensive semi - quantitative food frequency questionnaire using dietary records and biochemical markers among Greek schoolteachers. Int. J. Epidemiol., 26: S118 -S127.

Keys, A. 1966. Dietary survey methods in studies on cardiovascular epidemiology. Voeding, 27: 9 -28.

Kigutha, H.N. 1997. Assessment of dietary intake in rural communities in Africa: experiences from Kenya. Am. J. Clin. Nutr., 65: 1168 -1172.

Kipnis, V., Carroll, R.J., Freedman, L.S. & Li, L. 1999. Implications of a new dietary measurement error model for estimation of relative risk: application to four calibration studies. Am. J. Epidemiol., 150: 642 -651.

Kohlmeier, L., Mendez, M., McDuffie, J. & Miller, M. 1997. Computer-assisted self-interviewing: a multimedia approach to dietary assessment. Am. J. Clin. Nutr., 65: 1275 -1281.

Kretsch, M.J. & Fong, A.K.H. 1993. Validity and reproducibility of a new computerized dietary assessment method: effects of gender and educational level. Nutr. Res., 13: 133 -146.

Kumanyika, S.K., Tell, G.S., Shemanski, L., Martel, J. & Chinchilli, V.M. 1997. Dietary assessment using a picture -sort approach. Am. J. Clin. Nutr., 65: 1123 -1129.

Leclercq, C., Valsta, L.M. & Turrini, A. 2001. Food composition issues - implications for the development of food-based dietary guidelines. Public Health Nutr., 4: 677 -682.

Livingstone, M.B., Prentice, A.M., Strain, J.J., Coward, W.A., Black, A.E., Barker, M.E., McKenna, P.G. & Whitehead, R.G. 1990. Accuracy of weighed dietary records in studies of diet and health. Br. Med. J., 300: 708 -712.

Livingstone, M.B., Prentice, A.M. & Coward, W.A. 1992. Validation of estimates of energy intake by weighed dietary record and diet history in children and adolescents. Am. J. Clin. Nutr., 56: 29 -35.

McDowell, M.A., Briefel, R.B. & Alaimo, K. 1994. Energy and macronutrient intakes of persons ages 2 months and over in the United States: Third National Health and Nutrition Examination Survey, phase 1, 1988-91. Department of Health and Human Services Publications. Washington, DC, US Government Printing Office.

McKeown, N.M., Day, N.E., Welch, A.A., Runswick, S.A., Luben, R.N., Mulligan, A.A., McTaggart, A. & Bingham, S.A. 2001. Use of biological markers to validate self-reported dietary intake in a random sample of the European Prospective Investigation into Cancer United Kingdom Norfolk cohort. Am. J. Clin. Nutr., 74: 188 -196.

Martinchik, A.N., Baturin, A.K. & Helsing, E. 1997. Nutrition monitoring of Russian schoolchildren in a period of economic change: a World Health Organization multicenter survey, 1992 -1995. Am. J. Clin. Nutr., 65: 1215 -1219.

Medlin, C. & Skinner, J.D. 1988. Individual dietary intake methodology: a 50 -year review of progress. J. Am. Diet. Assoc., 88: 1250 -1257.

Mensink, G.B.M., Haftenberger, M. & Thamm, M. 2001. Validity of DISHES 98, a computerised dietary history interview: energy and macronutrient intake. Eur. J. Clin. Nutr., 55: 409 -417.

Mertz, W., Tsui, J.C., Judd, J.T., Reiser, S., Hallfrisch, J., Morris, E.R., Steele, P.D. & Lashley, E. 1991. What are people really eating? The relation between energy intake derived from estimated diet records and intake determined to maintain body weight. Am. J. Clin. Nutr., 54: 291 -295.

National Center for Health Statistics. 1994. Plan and operation of the Third National Health and Nutrition Examination Survey, 1988 -94. Department of Health and Human Services Publications. Washington, DC, US Government Printing Office.

National Food Agency. 1990. Food monitoring in Denmark. Nutrients and contaminants 1983 -1987. Soborg, Denmark, National Food Agency of Denmark.

National Nutrition Surveillance Centre. 1993. Nutritional surveillance in Ireland 1993. Series Report No. 1. 1993. Galway, Ireland, University College Galway, Centre Health Promotion Studies.

National Public Health Institute. 1998. The 1997 dietary survey of Finnish adults. Publication Series B 8/1998. Helsinki.

Norgan, N.G., Ferro-Luzzi, A. & Durnin, J.V. 1974. The energy and nutrient intake and the energy expenditure of 204 New Guinean adults. Philos. Trans. R. Soc. London B., 268: 309 -348.

Rizek, R.L. & Pao, E.M. 1990. Dietary intake methodology I. USDA surveys and supporting research. J. Nutr., 120: 1525 -1529.

Rockett, H.R. & Colditz, G.A. 1997. Assessing the diets of children and adolescents. Am. J. Clin. Nutr., 65 (Suppl. 4): 1116S -1122S.

Romieu, I., Hernandez-Avila, M., Rivera, J.A., Ruel, M.T. & Parra, S. 1997. Dietary studies in countries experiencing a health transition: Mexico and Central America. Am. J. Clin. Nutr., 65: 1159 -1165.

Scaccini, C., Sette, S., Mariotti, S., Verdecchia, A. & Ferro-Luzzi, A. 1992. Nutrient adequacy of dietary intakes of the elderly. Age Nutr., 3: 41 -47.

Shetty, P.S. 1993. Chronic undernutrition and metabolic adaptation. Proc. Nutr. Soc., 52: 267 -284.

Shetty, P.S. 1999. Adaptation to low energy intakes: the responses and limits to low intakes in infants, children and adults. Eur. J. Clin. Nutr., 53 (Suppl. 1): S14 -S33.

Subcommittee on Criteria for Dietary Evaluation, National Research Council. 1986. Nutrient adequacy: assessment using food consumption surveys. Design, methods and first results. Washington, DC, National Academy Press.

Tarasuk, V. & Beaton, G.H. 1992. Statistical estimation of dietary parameters: implications of patterns in within-subject variation - a case study of sampling strategies. Am. J. Clin. Nutr., 55: 22 -27.

Teufel, N.I. 1997. Development of culturally competent food-frequency questionnaires. Am. J. Clin. Nutr., 65 (Suppl. 4): 1173S -1178S.

Thompson, F.E., Moler, J.E., Freedman, L.S., Clifford, C.K., Stables, G.J. & Willet, W.C. 1997. Register of dietary assessment calibration- validation studies: a status report. Am. J. Clin. Nutr., 65: 1142 -1147.

Torheim, L.E., Barikmo, I., Hatloy, A., Diakité, M., Solvoll, K., Diarra, M.D. & Oshaug, A. 2002. Validation of a quantitative food-frequency questionnaire for use in Western Mali. Public Health Nutr., 4: 1267 -1277.

Torun, B. 2001. Energy requirements of children and adolescents. Rome, FAO/WHO/ UNU Working Groups on Energy and Protein in Human Nutrition.

Turrini, A., Leclercq, C. & D'Amicis, A. 1999. Patterns of food and nutrient intakes in Italy and their application to the development of food-based dietary guidelines. Br. J. Nutr., 81 (Suppl. 2): S83 -S89.

US Department of Agriculture, Nutrition Monitoring Division. 1986. Nationwide Food Consumption Survey: CSF II. Continuing survey of food intakes by individuals - 1986. Low- income women 19 -50 years and their children 1 -5 years, 4 days, 1986. Washington, DC.

Waterlow, J.C. 1989. Nutritional adaptation and variability. Eur. J. Clin. Nutr., 43: 203 -210.

Wiehl, D.G. & Reed, R. 1960. Development of new or improved dietary methods for epidemiological investigations. Am. J. Public Health, 50: 824 -828.

Willett, W.C. 1998. Nutritional epidemiology, 2nd edition. New York, Oxford University Press. Willett,

W.C. & Sampson L., eds. 1997. Second International Conference on Dietary Assessment Methods. Am. J. Clin. Nutr., 65: 1098S -1368S.

Discussion opener - individual food intake survey methods

Jeanne H.M. de Vries
Wageningen University
Wageningen, The Netherlands

General comments

The paper by Ferro-Luzzi provides a complete overview of assessment of dietary intake. However, to choose a method for assessing food security, in particular undernourishment, errors of the various methods have to be evaluated directly in relation with these aims. First, the type of information in statistical terms that is required to meet the aims should be determined. Subsequently, a method providing the required information should be chosen. To choose an appropriate method, the objectives of study, the target group, the reference period, the timetable and the need for cross -population comparability must be taken into account. Finally, the errors of the chosen method should be determined and, if possible, corrected for.

Assessment of undernourishment

In the assessment of food security and undernourishment, accurate information is needed in order to estimate the proportion of individuals with inadequate (or excessive) intake.

One way to obtain this information is by monitoring anthropometric measurements. In addition, individual dietary intake surveys can be used to assess the relationship between food intake and food requirement. The required information for estimating the proportion of people with inadequate intake clearly differs from that needed to assess diet -disease relationships. To know whether the acquired information is sufficiently valid for measuring undernourishment, the nature of random as well as systematic errors of a method have to be known.

Measurement error

For the required type of information, we have to deal with four types of error: random and systematic within-person error, and random and systematic between-person error. Random within-person error occurs, for example, when the method does not account for day-to-day variation of a person when determining habitual consumption. It also occurs when foods or portion sizes are incorrectly estimated in a non-systematic way. Increasing the number of replicate measurements may reduce random within-person error. Systematic within-person error may be caused when a person under-or overestimates their food intake, for instance when an important food is not included in the questionnaire. Repeated measurements do not improve this type of error. Random between- person errors are distributed randomly among individuals and should not bias the estimated mean intake. However, this type of error artificially reduces the precision of the estimate. Increasing the number of subjects or replicate measurements may improve precision, but not the validity of estimates for the percentage of undernourished subjects. Systematic between-person errors are not randomly distributed among individuals, e.g. socially desirable answers may be given by one group of persons and not others. As a consequence, neither the mean intake nor the percentage of undernourished persons is estimated correctly.

Type of dietary assessment method

The food frequency questionnaire is often regarded as the method of preference in large - scale studies because of cost aspects and practical advantages. However, it is often forgotten that it is laborious to develop a good questionnaire. Also, food frequency questionnaires should be culturally based, making it difficult to use them across different cultures. In the EPIC study (European Prospective Investigation into Cancer and Nutrition), systematic bias of the various food frequency questionnaires developed for use in seven countries had to be evaluated before intercohort analyses were allowed. Nevertheless, questionnaires may be useful for follow-up studies within populations. One advantage of the 24 -hour recall method is its suitability for large populations of different ethnicities because it is open-ended and therefore not culturally based. Furthermore, it has a relatively low respondent and interviewer burden. Increasing the number of 24 -hour recall periods and/or the number of subjects reduces random error. For health monitoring purposes, at least two non-con-secutive 24 -hour recalls are necessary, while more than two are needed for estimating undernourishment. To account for systematic errors such as undereating while recording food intake, the method should be combined with another measurement, such as a checklist of important foods or measurement of body weights.

If food composition tables are of a low quality, assessment of food patterns or the intake of a few important food groups, rather than nutrients, could be considered. These assessments could also be combined with measurement of body weight to assess whether energy intake is sufficient. Another problem of 24 -hour recalls is that they are not suitable to be applied in children below the age of ten years. In a study we performed on five -and six-year-old children, we asked the parents to report their children's intake during the preceding 24 hours. Food records and histories have major disadvantages, which make them less suitable to achieve the above -mentioned aims; for example, they are not applicable in illiterate populations, or there is a need for skilled interviewers. When household budget surveys are applied, trends of intake could be compared with those of individual surveys, in order to assess the nature of errors.

Correction of errors

Formulas that correct the effects of random measurement error are available for the distribution of intake and for measures of associations with other variables. Statistical corrections can be made by calibration or triad analysis with the use of accurate biomarkers of intake. Information for applying formulas and corrections can be obtained from dietary validity studies performed in large representative samples. Such studies are urgently needed for non-western countries.

Application of the method

It should be stressed that errors can be reduced by practical improvements in the methodology and that the use of computers can facilitate this. Major problems to deal with in surveys are selection bias, non-response, incomplete time sampling, inaccurate estimation of portion size and recall bias. The use of strictly standardized procedures for sampling, interviewer qualification and training and quality control is needed in order to prevent or minimize errors, especially in the case of cross-country comparisons. To assess portion sizes, a picture book is recommended.

The conclusion of this discussion is that for a valid assessment of dietary intake, the type of information required when measuring undernourishment differs from that needed when assessing diet -disease relationships, the most common use of these methods. Furthermore, validation studies in the target population are necessary to identify the nature of errors. Repeated 24 -hour recalls, performed according to standardized procedures and combined with other measurements, appear to be the preferred method.

Acknowledgement

This paper was written in collaboration with Wija A. van Staveren.

Discussion opener - individual food intake survey methods

Lauren Lissner
Gothenburg University
Gothenburg, Sweden

Researchers working in areas of food abundance as well as food scarcity have all experienced doubts when seeking the "truth" about dietary exposures. Dr. Ferro-Luzzi's keynote paper gives a comprehensive overview of different methods that can be used for assessing individual diets and how data collected using these methods might be applied to the assessment of food security in high-risk regions. The paper emphasized that dietary assessment is fraught with uncertainties regarding measurement error and bias. A number of well -known examples are given, including studies involving obese subjects, results of which have forced many investigators to question the validity of food intake survey methods in well -nourished populations. In this "discussion opener", I will give a few additional examples of biases present in dietary assessment studies, including general under-reporting and selective under-reporting, which can have profound consequences in both descriptive and analytic studies.

The problem of how to measure habitual food intake in overweight individuals has long been an enigma in nutritional research (Lissner et al., 1989). The existence of obesity-specific under-reporting was rather controversial until the advent of the doubly labelled water method gave credence to much previously anecdotal evidence of such a bias (Prentice et al., 1986). Studies involving urinary nitrogen as a biomarker for total protein intake have also been consistent with this conclusion (Heitmann and Lissner, 1995; Heitmann, Lissner and Osler, 2000). Specifically, under-reporting is a feature of most dietary surveys, but seems to be exaggerated in overweight individuals, for reasons that have been explored using a variety of research designs.

A special group of reporting biases relating to perceptions of the acceptable or correct answer are inherent in dietary surveys in general, but may be particularly relevant in samples including many overweight individuals. The term "social desirability bias" has been widely adopted to imply a tendency that varies among individuals to supply answers that place the interviewee in a favourable light (Hebert, 1995). This concept has been broken down further into intentional reporting error ("impression management") and unconscious error ("self-deception") (Roth, Snyder and Pace, 1986). In dietary surveys, additional factors (e.g. memory, effort required) clearly compound the normative biases described above. In the case of obesity, it is important to keep in mind that certain errors can occur even without any intentional or unintentional deception. For instance, the act of keeping food records, a time -honoured behavioural tool in weight loss programmes, may cause overweight subjects to underconsume, and therefore appear to under-report, even if the data reflect the truth at the time the records are being kept. Of particular interest in the context of this meeting, it has been demonstrated that social desirability can produce under-reporting independent of obesity (Taren et al., 1999). Finally, two closed metabolic studies have observed that obesity- specific under-reporting did not occur in an artificial setting where subjects were aware of being observed (Gripeteg and Lissner, 2000; Poppitt et al., 1998). These conclusions have been partially confirmed in a natural setting using the "bogus pipeline paradigm" in which investigators found that reporting accuracy in former dieters was improved when they believed that researchers could verify their food records with doubly labelled water (Muhlheim et al., 1998).

Whether under-reporting is intentional or unintentional, it is also unclear which food types, if any, are selectively under-reported in a predictable way. Let us consider the situation in a survey in which subjects reported their usual diets while collecting urine samples for nitrogen assessment. There was proportionately less under-reporting of protein compared with other energy sources combined, providing evidence that foods high in fat and/or carbohydrates were being under-reported relatively more often (Heitmann and Lissner, 1995; Heitmann, Lissner and Osler, 2000). These and other findings of food specific under-reporting in subjects under direct observation (Gripeteg and Lissner, 2000; Poppitt et al., 1998) suggest that a duel bias may be present in nutritional surveys: general under-reporting, particularly in obese subjects, compounded by food-specific under-reporting. The types of foods that are most under- reported appear to vary from study to study: systematic omission of snacks and alcohol has been reported in one investigation (Poppitt et al., 1998), while significant under-reporting of butter as well as fruits and vegetables was reported by others (Gripeteg and Lissner, 2000). Taken together, these under-reported foods might be characterized as being easy to forget, rather than having consistently negative health connotations.

Systematic or non-random errors of the type investigated in the above -mentioned studies have important implications in studies of diet and disease. One example comes from a study that did not directly involve obesity per se, but rather a large sample of women who were asked to report their usual diets twice, first when they were free of breast cancer and some years later after a subset had received diagnoses (Giovannucci et al., 1993a, b). These results are consistent with the idea that knowledge of disease affects how individuals recall their past intakes, particularly for substances such as fat or alcohol that have specific social desirability attributes. The fact that the diet - disease association assessed retrospectively was slightly attenuated in the case of remote alcohol recalls but became larger in the case of dietary fat should alert nutritionists to the very complicated and unpredictable character of these biases.

In conclusion, normative biases have been observed in well -nourished populations and in overweight groups, and are surely also present in conditions of undernutrition ranging from anorexia to famine. In addition to errors involving the whole diet, it appears that certain types of foods are specifically under-reported and possibly over-reported, which would result in non-proportional scaling biases for various nutrients. It is interesting to speculate on types of reporting biases that may be present in situations of food scarcity, and how they might affect results of dietary surveys aiming to document these conditions. A tendency to overestimate intake when true intake is very low would obviously result in an underestimate in the magnitude of the population's food deprivation. However, it is possible to imagine a scenario of under-reporting even when true intake is already low, which would have the opposite impact when estimating the magnitude of the problem. Furthermore, selective over-reporting of socially desirable foods and/or omission of undesirable foods is likely to occur and may have a varying impact on accuracy of reported consumption of Western products, indigenous products, donated products etc. Finally, in transitional areas with sharply increasing prevalences of overweight and obesity, increasing dietary under-reporting is likely to accompany this trend in future nutritional surveillance efforts.

References

Giovannucci, E., Stampfer, M.J., Colditz, G.A., Manson, J.E., Rosner, B.A., Longnecker, M.P., Speizer, F.E. & Willett, W.C. 1993a. Recall and selection bias in reporting past alcohol consumption among breast cancer cases. Cancer Causes Control, 4: 441 -448.

Giovannucci, E., Stampfer, M.J., Colditz, G.A., Manson, J.E., Rosner, B.A., Longnecker, M.P., Speizer, F.E. & Willett, W.C. 1993b. A comparison of prospective and retrospective assessments of diet in the study of breast cancer. Am. J. Epidemiol., 137: 502 -511.

Gripeteg, L. & Lissner, L. Reported versus observed food intakes of men residing in a metabolic chamber (abstract). Fourth International Conference on Dietary Assessment Methods, September 17 -20, 2000, Tuscon, AZ.

Hebert, J.R., Clemow, L., Pbert, L., Ockene, I.S. & Ockene, J.K. 1995. Social desirability bias in dietary self-report may compromise the validity of dietary intake measures. Int. J. Epidemiol., 24: 389 -398.

Heitmann, B. & Lissner, L. 1995. Dietary under- reporting by obese individuals: Is it specific or nonspecific? Br. Med. J., 311: 986 -989.

Heitmann, B.L., Lissner, L. & Osler, M. 2000. Do we eat less fat, or just report so? Int. J. Obes., 24: 435 -442.

Lissner, L., Habicht, J.-P., Strupp, B., Levitsky, D.A., Haas, J.D. & Roe, D.A. 1989. Body composition and energy intake: do overweight women overeat and underreport? Am. J. Clin. Nutr., 49: 320 -325.

Muhlheim, L.S., Allison, D.B., Heshka, S. & Heymsfield, S.B. 1998. Do unsuccessful dieters intentionally underreport food intake? Int. J. Eat. Disord., 24: 259 -266.

Poppitt, S.D., Swann, D., Black, A.E. & Prentice, A.M. 1998. Assessment of selective under-reporting of food intake by both obese and non-obese women in a metabolic facility. Int. J. Obes., 22: 303 -311.

Prentice, A.M., Black, A.E., Coward, W.A., Davies, H.L., Goldberg, G.R., Murgatroyd, P.R., Ashford, J., Sawyer, M. & Whitehead, R.G. 1986. High levels of energy expenditure in obese women. Br. Med. J., 292: 983 -987.

Roth, D.L., Snyder, C.R. & Pace, L.M. 1986. Dimensions of favorable self-presentation. J. Pers. Soc. Psych., 51: 867 -874.

Taren, D.L., Tobar, M., Hill, A., Howell, W., Shisslak, C., Bell, I. & Ritenbaugh, C. 1999. The association of energy intake bias with psychological scores of women. Eur. J. Clin. Nutr., 53: 570-578.

Discussion group report - individual food intake surveys methods

Marie-Claude Dop
FAO, ESNA
Rome, Italy

Ms de Vries and Ms Lissner complemented Anna Ferro-Luzzi's comprehensive overview of individual dietary assessment methods with an in-depth discussion of issues of validity of the methods (owing to errors inherent in the measurement of food consumption and/or owing to characteristics of the subjects), and of cost and feasibility.

Ms de Vries stated that the choice of a method for assessing food security and undernourishment should be based on several criteria - the type of information the method is capable of providing, cross-country comparability, costs involved, and types of measurement error and ways of correcting them. Individual food intake survey methods can reasonably provide the mean and the distribution of energy intake and thus give an estimate of the prevalence of undernourishment in a population, although anthropometric monitoring is a reliable alternative. Individual intake surveys, in addition, could be useful to assess food patterns and to provide estimates of intake of particular foods, of interest when food composition tables are incomplete or inaccurate.

The three main sources of error encountered in these types of surveys are reporting errors, coding errors and errors in food composition tables. There are two types of errors, random and systematic, both of which can have within-or between-subject components and can affect the estimates of prevalence of undernourishment. While random errors may sometimes be reduced by increasing the number of subjects or days of measurement, systematic errors can rarely be corrected for.

Two methods are likely candidates for assessing the prevalence of undernourishment: repeated 24 -hour recalls and food frequency questionnaires. The repeated 24 -hour recall is open-ended and is not culturally based, and so it is suitable for populations of different ethnicities and allows cross-country comparisons to be made. The food frequency questionnaire, on the contrary, uses a closed list of foods and may not be comparable across countries. The time to explain and administer the methods is about the same, but coding is much longer with the 24 -hour recall, while development is much more laborious and time -consuming with the food frequen-cy questionnaire. Respondent burden is low with both methods. Thus, the most appropriate method in terms of validity and feasibility would be the repeated 24 -hour recall combined with anthropometry. Nevertheless, validation/calibration studies would be needed. Strict standardization of procedures would be useful to minimize errors.

Ms Lissner also addressed problems of errors in individual intake surveys. She gave examples of biases observed in the context of studies conducted in western countries, in particular among obese subjects, that highlight the complexity of errors encountered in dietary assessment.

Studies using biomarkers such as the doubly labelled water method and 24 -hour urinary nitrogen have proven that under-reporting is common in dietary assessment surveys -both general under-reporting in whole populations and selective under-reporting by obese subjects.

The so -called "normative biases" are related to the social desirability of intakes, as subjects tend to report socially acceptable intakes. This type of bias can have intentional and unintentional components. Some under-reporting is general and some is related to obesity.

Social desirability biases are compounded with other sources of error, such as memory or recording errors. For instance, during record- keeping, some subjects will underconsume and thus appear to be under-reporting. The subjects' knowledge of being observed reduces under-reporting but does not eliminate it completely. In retrospective clinical studies, there is a high risk of information bias, i.e. subjects' knowledge of the diagnosis will distort their memory of food intake.

An important issue is to identify foods and/or nutrients that are selectively under- reported. Some studies have shown that fat and carbohydrate intakes are under-reported to a larger extent than protein. Other studies have shown that the obese under-report specific foods, e.g. snack foods and alcohol. Nevertheless under-reporting of foods varies across studies. It may be unintentional in the case of certain foods. For example, foods that are easy to forget, such as snack foods, are often omitted.

Normative biases are probably also encountered in undernourished populations. Both over-reporting of socially desirable foods and omission of undesirable foods are possible, but studies are needed to assess whether these biases exist. In countries whose population is undergoing a nutrition transition, obesity-related under-reporting is likely to be encountered. Biases need to be assessed and understood so that their impact on assessment of food intake can be estimated.

In the discussion that followed, the group concluded that individual food intake methods have a much stronger basis than most of the other methods discussed at the Symposium, i.e. the FAO methodology, household income and expenditure surveys, and qualitative indicators of hunger. Individual food intake methods have been scrutinized much more intensively than other methods. Many validity and reliability studies have been conducted, and so the error structure is well known, while the validity of the other methods still needs to be assessed. Moreover, these types of surveys are feasible. Their costs are not greater than household income and expenditure surveys that are presently being conducted in many developing countries.

After this discussion of validity issues, the group examined the question of whether individual food intake methods were useful to assess the prevalence of undernourishment, defined as energy insufficiency, or whether they might be useful to measure other relevant information on consumption. The discussion group's opinion was that individual food intake methods were useful for assessing acute energy insufficiency, for instance in situations of famine, but were not accurate or precise enough to assess chronic energy insufficiency. Small daily energy deficits, e.g. deficits of 100 -200 kcal /person/day, can lead in the long- term to energy insufficiency. Deficits of this magnitude cannot be detected by individual food intake methodology. Moreover, there is some uncertainty in the estimation of energy requirements of populations because physical activity levels cannot be assessed precisely at the population level. Therefore, the energy balance derived from estimates of dietary intake and requirement would not be precise or valid enough to serve as a basis for the assessment of prevalence of undernourishment.

The group felt that individual dietary intake was a more appropriate method for assessing nutrient insufficiencies and other aspects of the diet that are important with regard to undernourishment and health. The concept of undernourishment referred to was broader than mere energy insufficiency. These other aspects of the diet include:

diet quality;
food patterns;
intakes of individual foods and food groups;
intakes of macronutrients;
diet composition (e.g. percent of energy from fat);
intakes of micronutrients.

The participants agreed that two 24 -hour recalls would provide the mean and distribution of food and nutrient intakes in the population after removing the within-subject variation of intake. Assessment of this type would be useful at the national level, but also at the subnational level, in order to identify groups at risk of dietary deficiencies and to target interventions.

The group then discussed which other methods among those discussed at the Symposium would be of interest to assess undernourishment in its broader definition. Although all methods provide relevant information, it was stated that simple anthropometry (height and weight) was the most useful and cost-effective measurement. The group recommended that anthropometry be integrated not only into individual food intake surveys or household income and expenditure surveys, but also into all economic or health surveys and surveillance programmes. Moreover, these programmes would benefit from the information supplied by anthropometric measurements. Apart from surveys at the national level, anthropometry should be integrated into surveillance programmes using sentinel sites or, for example, into country-level FIVIMS programmes.

The group gave some practical recommendations regarding cost-effective and feasible data collection efforts at the country level, taking into account the resources that countries could devote to assessment. If resources are very limited, only anthropometry should be measured. With more resources, countries could choose to do one 24 -hour recall, or one recall in the total sample and a second recall on a subsample to allow for the estimation and removal of within-subject variability. If resources permit, biochemical indicators should be added, for instance determination of haemoglobin or haematocrit and serum retinol where relevant.

Some participants thought that existing sources of information could be used to assess dietary quality and related aspects. For instance, food balance sheets could provide information on micronutrient availability at the national level.

First, the group concluded that individual food intake methods were not able to assess energy deficiency in populations in a sufficiently valid and precise way. Second, the group stated that anthropometry was the most important and cost-effective indicator of undernourishment. Third, the group agreed that a broader definition of undernourishment was more relevant than the assessment of energy insufficiency per se, and that this definition should encompass other nutrients and diet quality, for which individual food intake methods could provide useful and valid information.


Previous Page Top of Page Next Page