3

M e a s u r i n g food c o n s u m p t i o n o f i n d i v i d u a l s

This chapter describes methods commonly Such methods can be used to assess the usual

used for measuring the food consumption intake of foods or specific classes of foods.

of individuals. Subsequent chapters discuss With modification, they can also provide data

the factors associated with the reproducibility on usual nutrient intakes.

and validity of each of these methods (Chap­ The measurement of food consumption

ters 5- 7) and the calculation and evaluation at the individual level is costly and time

of nutrient intakes (Chapters 4 and 8). consuming. Hence, such studies should be

planned with care. Consideration should be

given to the cost-effective collection of addi­

3.1 Methods for measuring food tional data from the same subjects at the same

consumption of individuals time; such additional information may sig­

nificantly enhance the interpretation of the

Two groups of methods are used to measure dietary data. At a minimum, socioecon­

the food consumption of individuals. The first omic and health-related information, simple

group, known as quantitative daily consump­ anthropometric measures, possibly a physical

tion methods, consists of recalls or records activity questionnaire, and biological samples

designed to measure the quantity of the indiv­ for the determination of important biomark­

idual foods consumed over a 1-d period. By ers (Chapter 7) should be collected when

increasing the number of measurement days, time and resources permit (Buzzard and Siev­

quantitative estimates of the usual intakes of ert, 1994).

individuals can be obtained, using the same The accurate assessment of the food intake

methods. The number, selection, and spac­ of infants is a particularly difficult problem,

ing of the days depend on the food intake, the especially when infants are receiving both

nutrients of interest, the day-to-day variation breast milk and complementary foods (Pi­

in nutrient intake, and the level of preci­ woz et al., 1995). WHO ( 1 9 9 8 ) has pub­

sion required. Determination of the usual lished guidelines that can be used to evaluate

intake is particularly critical when relation­ nutrient intakes of breastfed infants receiving

ships between diet and biological parameters complementary foods.

or chronic disease are assessed. Estimates of

usual intakes are also needed to estimate the

3.1.1 Twenty-four-hour recall method
prevalence of inadequate intakes.

The second group of methods includes the In the 24-h recall method, subjects and their

dietary history and the food frequency ques­ parents or caretakers are asked by the nutri­

tionnaire. Both obtain retrospective infor­ tionist, who has been trained in interviewing

mation on the patterns of food use during techniques, to recall the subject's exact food

a longer, less precisely defined time period. intake during the previous 24-h period or

41
42 Principles of Nutritional Assessment

Name: Date:

Street address: Day of the week:

Town/city:

LAB. USE ONLY

Place Time Description of food or drink. Amount Day/Meal Food Amount

eaten Give brand name if applicable code code code

Additional questions:

Was intake unusual in any way? Yes( . . . ) N o ( . . . )

If yes, in what way?

Do you take vitamin or mineral supplements? Yes ( . . . ) No ( . . . )

If yes, how many per day? ( . . . ) per week? ( . . . )

If yes, what kind? (give brand if possible)

Multivitamin Iron Ascorbic acid

Other (list)

Table 3 . 1 : Data sheet for a 24-h record. Modified from Weiner and Lourie (1969) with permission.

preceding day. Thus the method assesses tions of specified foods are outlined in Gibson

the actual intake of individuals. However, a and Ferguson (1999).

single 24-h recall is not sufficient to describe In the third pass, estimates of the amount of

an individual's usual intake of food and nutri­ each food and beverage item consumed are

ents; multiple 24-h recalls on the same indiv­ obtained, generally in household measures,

idual over several days are required to achieve and entered on the data sheet (Table 3 . 1 )

this objective (Section 3 . 1 . 2 ) . Nevertheless, or computer-based data-entry form. Photo­

multiple single-day recalls on different indi­ graphs, a set of measuring cups, spoons, and

viduals can give a valid measure of the intake rulers, local household utensils (calibrated

of a group or population (Section 3 . 3 . 1 ) . for use), or food models of various types

A four stage, multiple-pass interviewing (Section 5.2.6) can be used as memory aids

technique is often used; details are given in or to assist the respondent in assessing por­

Gibson and Ferguson (1999). Briefly, in the tion sizes of food items consumed (Gibson

first pass, a complete list of all foods and bev­ and Ferguson, 1999). Information on the in­

erages consumed during the preceding day is gredients of mixed dishes consumed by the

obtained, followed, in the second pass, by a respondents must also be collected at this

detailed description of each food and bever­ time. Finally, in the fourth pass, the recall

age consumed, including cooking methods is reviewed to ensure that all items, includ­

and brand names (if possible). Standardized ing use of vitamin and mineral supplements,

probe questions should be used to elicit spe­ have been recorded correctly. (Methods for

cific details for each food item. For example, coding the completed 24-h recalls and poten­

for milk products, probe questions should in­ tial sources of coding errors are discussed in

clude the kind of dairy product, brand name Section 5 . 2 . 7 . )

(if appropriate), and percentage fat (as butter­ Any 24-h interview protocol must be stan­

fat or milk fat). Further examples of probes dardized and pretested prior to use. Standard­

that can be used to obtain detailed descrip- ization is particularly important in large-scale
Measuring food consumption of individuals 43

national surveys and for comparisons across taken into account (Section 6.2.4). The re­

countries (Slimani et al., 1999). Pretesting spondent burden is small for a single 24-h

should be undertaken in an area near the study recall, so that compliance is generally high.

site, using respondents similar to those who The method is quick and relatively inexpen­

will participate in the actual study. Some­ sive, and it can be used equally well with both

times, the pretesting can be carried out on literate and illiterate subjects.

the field staff if they are comparable to the A 24-h recall has been used in some nation­

participants (Gibson and Ferguson, 1999). al nutrition surveys, including the New Zea­

Adherence to the interview protocol and land National Nutrition Survey (MOH,1997),

accuracy of food coding by the interview­ the U . S . National Health and Nutrition Exam­

ers should be checked periodically during ination Survey (NHANES) (NCHS, 1994),

the survey, and the interviewers must be re­ and the Continuing Survey of Food Intakes

trained if necessary to minimize interviewer by Individuals (CSFII) (USDA, 1998). Since

bias (Section 5 . 2 . 3 ) . Detailed suggestions on 2002, the CSFII has included a computerized

how to conduct the interview can be found multiple-pass recall with a number of built­

in Sanjur ( 1 9 8 2 ) and Hughes (1986), who in cues to specifically improve the recall of

stressed that leading questions and judgmen­ easily overlooked foods such as nonalcoholic

tal comments should be avoided. An indirect and alcoholic beverages, sweets, snacks, and

approach employing open-ended questions is breads. The CSFII is now integrated with

recommended. This enables respondents to NHANES.

freely express their feelings so that answers A modification of the 24-h recall- termed

are not biased. an interactive 24-h recall-has been devel­

In general, recall interviews can be con­ oped to collect information on rural pop­

ducted on children aged � 8 y (Young, 1 9 8 1 ; ulations in developing countries (Ferguson

Livingstone and Robson, 2000) and on most et al., 1995). The modifications are listed in

adults, except for persons with poor memo­ Box 3.1 and discussed in more detail in a

ries (e.g., some elderly). Children aged from manual containing practical guidelines and

4-8 y should be interviewed along with their procedures (Gibson and Ferguson, 1999).

primary caretaker, usually the mother. It may All recall interviews should be conducted

be necessary to interview several people if the in the respondent's home whenever possible,

children are at school or play in the homes of because the familiar environment encourages

friends, to ensure that foods eaten away from participation, improves the recall of foods

home are reported. For this younger age consumed, and facilitates calibration of local

group, questions should always be directed

toward the child (Sobo et al., 2000).

Very often when conducting recalls, es­ To improve the recall of food items in developing

countries, investigators can:

pecially on children, the interviewing pro­

ceeds as a consensus recall, with family mem­ Provide group training on portion size estima­

bers helping the respondent to remember the tion before the actual recall.

amounts consumed. This consensus approach Supply picture charts on the day before the re­

was shown to increase the accuracy of dietary call for use as a checklist on the day the food is

actually consumed, and for comparison with the

recalls of U.S. children (Eck et al., 1989).
recall to reduce memory lapses.
When 24-h recalls are used to character­

ize the average usual intake of a population Provide bowls and plates for use on the recall

days to help the respondents visualize the amount

group, the subjects should be representative
of food consumed.
of the population under study. In addition,
Weigh the portion sizes of salted replicas of the
the survey should be conducted in such a way
actual foods consumed by the respondent.
that all days of the week are equally repre­

sented. In this way, any day-of-the-week Box 3.1: Interactive 24-h recall modifications sug­

effects on food or nutrient intakes will be gested for rural populations in developing countries.
44 Principles of Nutritional Assessment

household utensils by the interviewer. In the foods and beverages (including brand names)

end, the success of the 24-h recall depends on and their method of preparation and cooking

the subject's memory, the ability of the sub­ are also recorded. For mixed dishes such as

ject to convey accurate estimates of portion spaghetti bolognese, the amount of each raw

sizes consumed, the degree of motivation of ingredient used in the recipe, the final weight

the respondent, and the persistence of the in­ of the mixed dish, and the amount consumed

terviewer (Acheson et al., 1980). by the subject should be recorded, wherever

possible. The information is recorded on a

form similar to that shown in Table 3 . 1 , ex­

3.1.2 Repeated 24-h recalls
cept that household measures are used for

Twenty-four-hour recalls can be repeated dur­ food amounts. Usually, the subject, par­

ing different seasons of the year to estimate ent, or caretaker completes the food record,

the average food intake of individuals over a although in less-developed countries a local

longer time period (i.e., usual food intake). field investigator may perform this task (Du­

The number of 24-h recalls required to esti­ four et al., 1999).

mate the usual nutrient intake of individuals Food portion size can be estimated by the

depends on the day-to-day variation in food respondent in a variety of ways. Standard

intake within one individual (i.e., within-sub­ household measuring cups and spoons should

ject variation) (Basiotis et al . , 2002). In tum, be used if possible, supplemented by meas­

this variation is affected by the nutrient un­ urements with a ruler (for meat and cake)

der study, the study population, and seasonal and counts (for eggs and bread slices). Un­

variations in intake. Nonconsecutive days fortunately, errors may arise because the re­

should be selected, when possible (Beaton spondent may fail to quantify portion sizes

etal., 1979). correctly. Additional errors may also arise

Repeated 24-h recalls were recommended during the conversion of volumes to weights

as part of a system for measuring food con­ (Section 5 . 2 . 5 ) . The latter step is usually

sumption patterns in the United States (NRC, completed by the investigator. Details on how

1981). For the CSFII 1994-1996 (USDA, to convert portion sizes to weight equivalents

1998), for example, food intakes on two non­ are given in Gibson and Ferguson (1999).

consecutive days were recorded through an The number of days included in an esti­

in-person 24-h recall. mated record varies, depending on the ob­

If it is not feasible to carry out repeated jective of the study. When the objective

observations on all respondents, the recalls is to obtain an average intake for a group,

should be repeated on a subsample of the then only one day per person is required,

population. In the 1988-1994 NHANES III provided all days of the week are equally

(NCHS, 1994) and the New Zealand Na­ represented in the final sample. However,

tional Nutrition Survey (MOH, 1997), 24-h when estimates of usual intakes of each per­

recalls were repeated on 5% and 15% of son are required, then the number, selection,

the population, respectively, across the entire and spacing of the days required per person

age range. Section 6 . 1 . 1 provides a detailed depends on the factors described for the re­

discussion of the reproducibility of the 24-h peated 24-h recall (Section 3 . 1 . 2 ) . Weekend

recall method. days should always be proportionately in­

cluded in the dietary survey period for each

person, to account for potential day-of-the­

3.1.3 Estimated food records
week effects on food and nutrient intakes.

For the estimated food record, the respondent This problem is discussed in more detail in

is asked to record, at the time of consumption, Sections 6 . 1 . 2 and 6.2.

all foods and beverages (including snacks) The European Prospective Investigation of

eaten in household measures, for a speci­ Cancer (EPIC) study in Norfolk, U.K. col­

fied time period. Detailed descriptions of all lected food and nutrient intakes from 2 1 1 7
Measuring food consumption of individuals 45

men and women. EPIC is a large multicen­ of the mixed dish. The method of record­

ter prospective study aimed at investigating ing is similar to that shown for a 24-h recall

the relationship between nutrition and vari­ (Table3.l), with the weight of the food items

ous life-style factors and the etiology of can­ being recorded under "Amount."

cer and other chronic diseases. The study If occasional meals are eaten away from

involved 23 regional centers located in ten home, respondents are generally requested to

countries, and involved a total cohort of about record descriptions of the amounts of food

480,000 subjects. The respondents of the eaten. The nutritionist can then buy and

EPIC study in Norfolk were provided with weigh a duplicate portion of each recorded

a "diet diary" - a 45-page colored booklet food item, where possible, to assess the prob­

in which they were asked to record the de­ able weight consumed. Alternatively, if ap­

scription, preparation, and amounts of foods propriate, the nutritionist can telephone a

eaten over seven consecutive days. Food por­ restaurant to obtain details of the portion sizes

tion sizes were estimated by the respondents consumed.

in terms of household measures, with the help As with the estimated record, the number,

of 17 sets of color photographs of small, spacing, and selection of days necessary to

medium, and large portions of the different characterize the usual nutrient intakes of an

foods. A semiquantitative food frequency individual using the weighed record depend

questionnaire (Section 3 . 1 . 6 ) was also used. on the within-subject variation in food intake,

Details are given in Bingham et al. (2001 ). which, in turn, depends on the nutrient of in­

terest, the study population, and any seasonal

variation of intake. Again, weekend days

3.1.4 Weighed food records
should be proportionately included to account

Weighed food records are more frequently for any weekend effect on the nutrient intake.

used in the United Kingdom and Europe be­ If a weighed food record method is to be

cause weighing scales are often used for food used, respondents must be motivated, numer­

preparation in these areas. In the British Na­ ate, and literate. However, respondents may

tional Diet and Nutrition Surveys of adults change their usual eating pattern to simplify

(Gregory et al., 1990; Finch et al., 199 8; Hen­ the measuring or weighing process or, alter­

derson et al., 2002, 2003a, 2003b) and chil­ natively, to impress the investigator (Cameron

dren (Gregory e t al . , 1995, 2000), 7-d weigh­ and van Staveren, 1 9 8 8 ) (Section 7 . 1 . 3 ) . Re­

ed food records were used. spondent burden for food records is higher

A weighed food record is the most pre­ than for the 24-h recall, so individuals may

cise method available for estimating usual be less willing to cooperate. Reproducibility

food and nutrient intakes of individuals. It is greater in the weighed record than in the

is the preferred method when diet counsel­ estimated record method because the portion

ing or correlation of intakes with biological sizes are weighed. Significant underreporting

parameters are involved. (Section 5.2.2) may, however, still occur.

In a weighed record, the subject, parent, or

caretaker is instructed to weigh all foods and

3.1.5 Dietary history
beverages consumed by the subject during

a specified time period. Details of meth­ The dietary history method (Burke,1947) at­

ods of food preparation, description of foods, tempts to estimate the usual food intake and

and brand names (if known) should also be meal pattern of individuals over a relatively

recorded. For mixed dishes such as spag­ long period of time - often a month. This

hetti bolognese, the weight of the portion interview method was originally designed

consumed should be recorded, along with the to be carried out by a nutritionist trained

weights and description of all the raw ingre­ in interviewing techniques. More recently,

dients, including flavors and spices used in computerized versions have been developed

the recipe, as well as the final total weight (Kohlmeier et al., 1997). Such versions pro-
46 Principles of Nutritional Assessment

vide standardized methods for data collection quently consumed in this study were weighed

and probing, and minimize potential inter­ by a dietitian in the home. A weighted daily

viewer bias in responses. average intake was then calculated from the

Initially, the dietary history had three com­ data, using the following formula:

ponents. The first component was an inter­

((5 x weekday)+ Saturday + Sunday) 1 7
view about the usual overall eating pattern of

the subject, both at mealtimes and between A modified version of this dietary history

meals. Such information included detailed method was adopted in the Survey in Eu­

descriptions of foods, their frequency of con­ rope on Nutrition in the Elderly: A Concerted

sumption, and usual portion sizes in common Action (SENECA). This multicenter survey

household measures. "What do you usually was designed to examine cross-cultural vari­

eat for breakfast?" is a typical question that ations in the nutrition, lifestyle, health, and

might have been included in the interview. performance of elderly Europeans (Euronut­

The second component served as a cross­ SENECA, 1 9 9 1 ) . The method involved the

check and consisted of a questionnaire on completion of a 3-d estimated record, fol­

the frequency of consumption of specific food lowed by an interview during which respon­

items. This part was used to verify and clar­ dents were questioned about their usual diet­

ify the information on the kinds and amounts ary intake over the past month. Portions of the

of foods given as the usual intake in the first most commonly eaten foods were weighed by

component. Questions asked related to spe­ the interviewer (van Staveren et al., 1996).

cific foods, such as: "Do you like or dislike The recording of a dietary history can be

milk." A 24-h recall of actual intake may also very labor intensive, with interviews taking

have been included at this stage. up to 2 h per subject (Slattery et al., 2000).

In the third component, subjects recorded Several investigators have reported that the

their food intake at home for three days. Por­ dietary history tends to overestimate nutrient

tion sizes at this stage were estimated using intakes, when compared with results from

a variety of techniques, including standard weighed records. Nes et al. ( 1 9 9 1 ) , for ex­

measuring cups and spoons, common uten­ ample, used the dietary history developed

sils, commercial plastic food models, photo­ for the SENECA study and showed that the

graphs, or real foods. Today, the original history generated consistently higher intakes

dietary history method is seldom used in this of energy and nutrients than the 3-d weighed

three-part format, the third component being records. Livingstone and Robson (2000) re­

commonly omitted. ported similar findings in a study of children

The time periods covered by the dietary and adolescents, but claimed that the results

history method vary. The maximum time obtained from the dietary history were more

period that can be used has not been defi­ representative of habitual intake than those

nitely established. When shorter time frames obtained from 7-d weighed records. In gen­

(i.e., � 1 mo) are used, reproducibility and eral, because dietary histories, unlike food

validity are apparently higher than for longer frequency questionnaires, do not limit the

periods (see Section 7 . 1 . 2 ) . Measurements variability in the responses, they overcome

of food intake over 1-y periods are probably many of the limitations of a food frequency

unrealistic unless seasonal variations in food questionnaire.

intakes are taken into account.

Dutch investigators used a three-part diet­

3.1.6 Food frequency questionnaire
ary history method covering 1 mo to record

usual food consumption on weekdays, Satur­ The food frequency questionnaire aims to as­

days, and Sundays separately (van Staveren et sess the frequency with which food items or

al., 1 9 8 5 ) . This approach takes into account food groups are consumed during a speci­

the potential effect of weekends on nutrient fied time period. It was originally designed

intake. The portion size of foods most fre- to provide descriptive qualitative informa-
 Measuring food consumption of individuals 47

For each food item, indicate with a checkmark the category that best

describes the frequency with which you usually eat that particular food item.

More than Once 3-6 Once or Twice per

Food item once per per times twice month

day day per week per week or less Never

Beef, hamburger
D D D D D D
Pork, ham
D D D D D D
Liver
D D D D D D
Poultry
D D D D D D
Eggs
D D D D D D
Dried peas/beans
D D D D D D
. . .

D D D D D D

Enter other foods not listed that are eaten regularly:

1 . . . . . . . . . . . . . . .
D D D
2 . . . . . . . . . . . . . . .
D D D
3 . . . . . . . . . . . . . . .
D D D

Table 3.2: Abbreviated food frequency questionnaire. A few foods and food categories are shown as

examples. A complete questionnaire might contain more than 100 items.

tion about usual food-consumption patterns. of fresh fruits and fruit juices as predictors

With the addition of portion-size estimates of vitamin C intake (Tsugane et al., 1998);

and the introduction of improved comput­ green leafy vegetables and carrots as pre­

erized self-administered questionnaires, the dictors of carotenoid intakes (O'Neill et al.,

method has become semi-quantitative, al­ 2001); whole grain cereals, legumes, nuts,

lowing the derivation of energy and selected fruits, and vegetables as predictors of diet­

nutrient intakes (Willet et al., 1 9 8 5 ; Block et ary fiber intakes (Merchant et al., 2003); and

al., 1986). dairy products as predictors of calcium in­

In its simplest form, the questionnaire con­ takes (Barr et al., 2001). The method can

sists of a list of foods and an associated also be used to assess the intake of fats and

set of frequency-of-use response categories cholesterol (Feunekes et al., 1993), artificial

(Table 3.2). The list of foods may focus on sweeteners, certain contaminants present in

specific groups of foods, particular foods, specific foods (Macintosh et al., 1997), al­

or foods consumed periodically in association cohol (Kesse et al., 2001), and condiments

with special events or seasons. Alternatively, (Maskarinec et al., 2000).

the food list may be extensive to enable es­ The food frequency questionnaires should

timates of total food intake and dietary di­ feature simple, well-defined foods and food

versity to be made. The frequency-of-use categories, and open-ended questions should

response categories may be daily, weekly, be avoided as preformatted lists of food cate­

monthly, or yearly, depending on the study gories act as a memory prompt. The method

objective. may use a standardized interview, a self­

Specific combinations of foods can be used administered machine-readable printed ques­

as predictors for intakes of certain nutrients or tionnaire, or a computer-administered ques­

non-nutrients, provided that the dietary com­ tionnaire. Most questionnaires take from 15-

ponents are concentrated in a relatively small 30 min to complete (see abbreviated example

number of foods or specific food groups. Ex­ given in Table 3.2). Hence, the food fre­

amples include the frequency of consumption quency questionnaire imposes less burden on
48 Principles of Nutritional Assessment

respondents than most of the other dietary

assessment methods. The results are easy to • Children 2-3

• Children 4-6
collect and process and are generally taken %
80 D Children 7-9
to represent usual intakes over an extended

period of time. Nevertheless, the validity

and feasibility of the food frequency ques­

60
tionnaire for estimating food intakes in the

remote past has not been clearly established

(van Staveren et al., 1986; Dwyer and Cole­

40
man, 1997; Ambrosini et al., 2003).

Food scores can be calculated from quali­

tative food frequency data and the frequency

20
of consumption of certain food groups. The

U.S. Food Guide Pyramid (USDA, 1992), or

an equivalent standard, listing the optimum

0
number of servings of the major food groups
Good diet Needs improvement Poor diet

per person per day, can serve as a basis for

the scores. The scores can then be exam­ Figure 3 . 1 : Healthy eating index rating for children

aged from two to nine, 1998. From Bowman et al.

ined in relation to psychosocial influences
(1998).
(e.g., level of education, income), as well

as vital statistics, season, geographic distri­

bution, and so on. preferably based on age- and sex-specific por­

Figure 3.1 illustrates the use of a food tion size data generated from country-specific

scoring system- termed the Healthy Eat­ national nutrition surveys (Willett et al., 1 9 8 5 ;

ing Index (HEI) - developed by the USDA Block et al., 1986). Note that inclusion of

for evaluating the quality of diets of U.S. information on portion sizes produces semi­

children, using data from USDA's Contin­ quantitative food frequency data. This can

uing Survey of Food Intakes by Individuals be converted to data on energy and nutrient

(Basiotis et al., 2002). The HEI has ten com­ intakes by multiplying the fractional portion

ponents in total, five of which are based on size of each food consumed per day by its

the suggested number of servings of each of energy and nutrient content, obtained from

the five food groups recommended in the U . S . appropriate food composition data. The re­

Food Guide Pyramid. The application of the sults are then summed to obtain an estimate

HEI shown in Figure 3 . 1 demonstrates that, of an individual's total daily intake.

of these subjects, it is the youngest children Block et al. (1986) derived a food list with

who consume the best diets. A more detailed portion sizes from the NHANES II results.

description of the HEI and other food scoring Food items selected contributed significantly

systems designed to evaluate the overall qual­ to the total population intake of energy and

ity of the diet is given in Section 8 . 4 . 1 . each of 17 nutrients. Serving sizes were

Kant et al. (2000) used food frequency in­ estimated from observed portion size distri­

formation to calculate a Recommended Food butions in the NHANES II data. Medium

Score (RPS) in a prospective study of diet serving sizes for each food were specified in

quality and mortality in women from the the food frequency questionnaire, and the re­

United States. They showed that the RPS was spondent indicated whether his or her usual

inversely associated with all-cause mortality. serving size was small, medium, or large, as

Many recent users of food frequency ques­ shown in Table 3 . 3 .

tionnaires have quantified portion sizes of A specialized food composition database

food items of interest, often using photo­ was developed for use with the food fre­

graphs (Nelson et al., 1994). Portion sizes quency questionnaire developed by Block

can be ranked as small, medium, and large, et al. (1986), based on the frequency of con-
 Measuring food consumption of individuals 49

Medium Serving How Often?

Food serving s M L D w M y N

Apples, apple sauce, pears ( 1 ) or 1/2 cup

Bananas 1 medium

Peaches, apricots (canned) ( 1 ) or 1/2 cup

Peaches, apricots (fresh) 1 medium

Cantaloupe 1/4 medium

Watermelon 1 slice

Strawberries 1/2 cup

Oranges 1 medium

Orange juice 6-oz glass

Grapefruit or grapefruit juice 1/2 or 6-oz glass

. . .

Table 3 . 3 : An example of part of the self-administered semiquantitative food frequency question­

naire. Abbreviations: S M L = small, medium, and large relative to the medium serving; D W M Y

N = daily, weekly, monthly, yearly, and never. From Block et al., American Journal of Epidemiology

124: 453-469, 1986, with permission of the Society for Epidemiologic Research.

sumption of certain specific food items ob­ of interviewer time (Block et al., 1990). In

served during NHANES II. A very similar some countries, a semiquantitative food fre­

approach has been used to design semiquan­ quency questionnaire has been used in na­

titative multi-ethnic food frequency question­ tional nutrition surveys (e.g., 1995 Australia

naires (Dreon et al., 1 993 ; Deurenberg-Yap National Dietary Survey) (McLennan and

et al., 2000). Podger, 1998).

The semiquantitative food frequency ques­ Food frequency questionnaires are often

tionnaire has become a widely used tool in used by epidemiologists studying associa­

dietary assessment. Country-specific semi­ tions between dietary habits and disease (Wil­

quantitative food frequency questionnaires lett, 1994; Levi et al., 2000; Kesse et al.,

containing between 130 and 300 food items 2001). In such studies, the food frequency

were used in the EPIC study to estimate questionnaires must be semiquantitative, with

individual usual food intakes (Margetts et al., the ability to rank subjects on the basis of

1997). In the EPIC study in Norfolk, U.K. their intakes, so that subjects with low in­

(Bingham et al., 2001), for example, respon­ takes can be separated from those with high

dents (n = 23,003) estimated how frequently intakes. This permits the calculation of the

foods were eaten over the past year, from odds ratio or relative risk of disease in rela­

. nine possible frequency-of-use response cat­ tion to intake of certain foods, food groups, or

egories from a list of 1 3 0 foods. Reduced nutrients (Masson et al., 2003). This approach

versions containing only 60 food items that was followed by Holick et al. (2002) in their

require only 17 min to administer by an in­ study of the relationship between dietary car­

terviewer are available; even the full 98-item otenoids and the risk of lung cancer; results

Block questionnaire requires only 30-35 min are shown in Table 3.4.
50 Principles of Nutritional Assessment

much greater than those for mail follow-up.

Quintile of nutrient
As a result, both NHANES III (1988-1994)
or intake per day (n) Median RR 95%CI
(NCHS, 1994), and the 1994-1996 USDA

Carotenoids (µg) Continuing Survey of Food Intake by Indi­

I ( < 2770) (397) 2170 1.00 Reference viduals (CSFII/DHKS, 1994-1996; USDA,
2 (2770-3786) (364) 3281 0.94 0.82, 1.08 1998) included telephone follow-up surveys
3 (3787-4988) (320) 4344 0.80 0.69, 0.93
rather than mail follow-up.
4 (4989-6792) (276) 5777 0.70 0.60, 0.81
Casey et al. (1999) carried out a valida­
5 (> 6792) (287) 8577 0.72 0.62, 0.84
tion study in which results of 24-h recalls
p for trend< 0.0001
conducted over the telephone were compared
Fruits + vegetables (g) with in-person recalls collected in the 1994-
1 ( < 1 1 6 ) (407) 80 1.00 Reference
1996 USDA CSFII; good correlations were
2 ( 1 1 6- 1 7 6 ) (362) 147 0.88 0.76, 1.02
reported. Based on these results, the USDA
3 (177-241) (326) 207 0.79 0.68, 0.91
nationwide food survey conducted in 2002
4 (242-332) (293) 280 0.71 0 . 6 1 , 0.83
also used telephone follow-up surveys.
5 (> 332) (256) 415 0.64 0.55, 0.75
Several other large-scale telephone diet­
p for trend< 0.0001
ary surveys have been used in the United

States, some of which have used a food fre­

Table 3.4: Relative risk of lung cancer according to

categories of baseline carotenoid and fruit + vegetable quency instrument rather than a 24-h recall.

intake in a cancer-prevention prospective study 1985- Lyu et al. (1998) successfully showed that
1998. RR, relative risk; CI, confidence interval. Data
agreement between telephone and face-to­
from Holick et al., American Journal of Epidemiology
face interviews of a semiquantitative food
1 5 6 : 536-547, 2002, with permission of the Society for
frequency questionnaire made up of 1 1 5 food
Epidemiologic Research.

items was good and unaffected by age, gen­

der, ethnicity, or education of the Hawaii

3.2 Technical improvements in food respondents. These investigators did recom­

consumption measurements mend mailing photographs of foods in three

portion sizes in advance, to help respondents

The increasing evidence of the relationship estimate amounts eaten more accurately.

between diet and chronic disease has led to Many of the smaller telephone dietary stud­

a number of technical advances in measure­ ies have been conducted on adult women

ments of food consumption for individuals. (Galasso e t al . , 1994; Casey etal., 1999; Tran

These aim to improve the speed and accu­ et al., 2000; Yanek et al., 2000); very few have

racy and reduce the cost of collecting and been carried out on adolescents and adult men

analyzing dietary intake data during large­ (Bogle et al., 2001). In college students, food

scale epidemiological studies and nutrition intakes by telephoned recalls have been com­

surveys. pared against actual intakes determined sur­

reptitiously in the college cafeteria (Krantzler

et al., 1982). More studies are needed, how­

3.2.1 Telephone
ever, among certain life-stage groups, to es­

A telephone survey that is well-designed and tablish the validity of 24-h recalls or food fre­

carefully-administered appears to be a prom­ quency questionnaires administered over the

ising method of obtaining dietary informa­ telephone. Underreporting of self-reported

tion. In particular, telephoned 24-h recalls food intakes may still occur in telephone

are being increasingly used. The USDA recalls, as they do with in-person recalls;

has conducted several studies to examine this occurred, for example, when total energy

the feasibility of using telephone follow-up intakes derived from telephone-administered

surveys instead of mail follow-up for 24-h 24-h recalls were compared with total en­

recalls. Results have been promising: re­ ergy expenditure measured by doubly labeled

sponse rates for telephone follow-up were water (Tran et al., 2000) (Section 7 . 2 . 1 ) .
 Measuring food consumption of individuals 51

No. of Portions Order of Series Table of

Country Reference series per series presentation per page Instructions contents

France Hercberg et al. ( 1994) 245 3 Increasing size 3 Yes Yes

Portugal Marques et al. (1996) 110 3 Varies 2 Yes No

Portugal Rombo et al. ( 1996) 58 4 Increasing size 2 No No

Portugal Galeazzi et al. (1996) 71 3 Decreasing size 2 Yes Yes

Poland Szczyglowa et al. ( 1 9 9 1 ) 135 3 Increasing size 3 No No

Finland Haapa et al. ( 1985) 126 3 Increasing size 3 Yes No

Russia Martintchik et al. ( 1995) 63 3 Increasing or varies 1 No No

U.K. Nelson et al. (1997) 98 8 Increasing size 1 Yes Yes

EPIC van Kappel etal. (1994) 140 4-6 Increasing size 1 Yes Yes

Table 3 . 5 : Some photographic atlases of food portion sizes. From Nelson and Haraldsdottir, Public

Health Nutrition 1 : 231-237, 1998.

Telephone surveys do have several advan­ able, the elderly who may have hearing dif­

tages. These include their ability to reach a ficulties, and persons with less education.

large number of persons at perhaps less than In the future, telephone dietary surveys

half the cost of face-to-face surveys. As may become a practical, economical, and

well, with the advent of computer-assisted valid alternative to the conventional face-to­

telephone interviewing, the interviews can be face methods for large-scale epidemiological

readily standardized, queries can be clarified, studies and nutrition surveys and for devel­

and responses can be coded immediately, dur­ oping and evaluating community-based nutri­

ing an interviewing time that is much shorter tion interventions in industrialized countries.

than a face-to-face interview. As a result, the

response rate is enhanced.

3.2.2 Photographs
Potential interviewer biases can be elim­

inated by using computer-assisted telephone Photographs can be used as memory aids or to

interviewing. However, other sources of bias quantify portion sizes (Section 5.2.6). Either

may occur with telephone surveys; these may photographs depicting a portion (i.e., amount

arise from noncoverage and nonresponse. In consumed on any one occasion) or a serving

the United States, although over 87% of the size (amount served in one helping) can be

population owns a telephone, subgroups such used. To quantify the portion sizes, a se­

as the poor, certain minorities, and the el­ ries of graduated portion-size photographs

derly still have fewer telephones than the for each food item is used, often bound to­

general population. Such differential cover­ gether in a photographic atlas. Table 3.5 lists

age can introduce bias in national surveys some of the photographic atlases currently

unless alternative compensating strategies are available. Existing atlases should be used,

employed. Strategies may include using sup­ whenever possible, provided they have been

plementary face-to-face or mail interviews validated by using respondents with char­

for persons without telephones; statistical ad­ acteristics similar to those who will partic­

justments employing weighting for sex, age, ipate in the planned study. For assessing

race, and income; selective over-sampling of intakes of individuals, a range of portion­

the non-telephone users; and random-digit di­ size photographs for each food is needed, but

aling to contact those with unlisted numbers. at the group level, single average portion size

Nonresponse is also a source of poten­ photographs may suffice (Robson and Living­

tial bias that influences telephone surveys. stone, 2000).

This is especially a problem among certain Practical guidelines on how to develop a

subgroups such as the ethnic minorities for photographic atlas are given in Nelson and

whom language barriers may be consider- Haraldsdottir ( 1 9 9 8 ) . Factors that must be

52 Principles of Nutritional Assessment

considered in relation to the format of the

photographs include size of the image, num­

ber and range of portion sizes depicted, and

the interval between portion sizes. In the

EPIC study, for example, there was a 25%

difference between the portion sizes to al­

90 sq. cm 30 sq. cm
low a real visual perception of differences Diam. = 10. 7 cm Diam. = 6 . 1 8 cm 120 sq. cm 30 sq. cm

10.95 x 10.95 cm 5.48 x 5·48 cm

in size (van Kappel et al., 1994). Nelson

et al. ( 1994) used portion weights from the

British Adult Dietary Survey ranging from

the 5th to the 95th percentile (Gregory et

Thickness indicator.
al., 1990), for a series of eight photographs
Each 'leaf' is 1 mm thick.

for each food. Other important factors that

should be standardized include the order of

Figure 3 . 2 : Models for use in the estimation of portion
presentation of the photographs, labels used,
size developed by Health and Welfare Canada (1973).
angle at which photographs are taken, back­

ground and use of reference objects for scale,

color versus black and white, and use of one cookies. Use of thickness indicators is crit­

versus several foods on a plate. ical for assessing portion sizes of intact cuts

Photography has also been used to reduce of meat, especially when irregular in shape.

the respondent burden imposed by complet­ Note that the use of a range of graduated food

ing food records. Elwood and Bird ( 1 9 8 3 ) models, like photographs depicting a range of

instructed subjects to photograph, at a spec­ portion sizes, prevents the tendency to gen­

ified distance and angle, all food items and erate a "direct" response. This phenomenon

leftovers and to record descriptions of each is observed when plastic food models repre­

foodstuff, including the method of prepara­ senting only "average" portion sizes are used

tion. Estimates of the weights of the food (Samuelson, 1970).

items consumed were obtained by viewing Since 1973, graduated food models have

the photographs alongside previously pre­ been used in many national food consump­

pared standard photographs of food portions tion surveys, including those in the United

with known weights. Such an approach ap­ States (USDA, 1 9 9 8 ; NCHS, 1994) and New

pears less demanding for the subject than Zealand (MOH, 1997; Parnell etal., 2003).

the conventional weighed record, and is rela­ The United States has developed a new tool

tively easy and acceptable. Nevertheless, this to measure portion sizes in the 24-h recalls

approach has not been widely adopted. conducted in their nationwide food surveys.

The tool consists of a food model booklet

of 32 life-size two-dimensional drawings of

3.2.3 Graduated food models
household vessels (glasses, mugs, bowls), ab­

Canada was one of the first countries to use stract shapes (mounds and spreads), and geo­

a collection of three-dimensional graduated metrical models (circles, a grid, wedges, and

food models in its National Nutrition Sur­ thickness bars), together with a set of meas­

vey (Health and Welfare Canada, 1973). The uring cups, spoons, and rulers. This tool

models consisted of papier-mache, wooden, is now used in the dietary component of

or hardboard shapes of various volumes and NHANES (Cleveland and Ingwersen, 2001);

surface areas. The surface-area models were more details are given in Section 5 . 2 . 5 .

accompanied by standard thickness indica­

tors made of hardboard squares (Figure 3.2).

3.2.4 Tape recorders
These were used for the 24-h recalls to assist

in assessing the overall size and thickness of A portable electronic set of tape-recording

foods such as cheese, cold meats, cakes and scales (PETRA) has been developed in the
Measuring food consumption of individuals 53

United Kingdom (Cherlyn Electronics, Cam­ 55 color-coded food record keys. The key­

bridge, England). In this system, the respon­ board was fitted with a removable transparent

dent places an empty plate onto the PETRA keyboard overlay, to assist in the correct iden­

digital recording scale, presses a switch in tification of the food keys. The respondent

front of the machine, and describes the plate. burden was reduced when using this food­

The scale simultaneously records the spoken recording device because the subject did not

words and the weight in digitally coded form. have to read the balance or keep a written di­

The respondent then adds each food item sep­ ary (Stockley et al., 1986b). The device also

arately onto the plate and, at the same time, eliminated the process of coding the food

dictates a description of the food into the mi­ records, a task considered to be the most

crophone. At the end of the study, tapes are time-consuming part of a quantitative diet­

retrieved and read by the investigator using ary study (Black, 1982). Details of some of

the PETRA Master Console. The latter plays the problems with coding foods are given in

back the description of the food and displays Section 5 . 2 . 8 .

the decoded weight information. In 1990, the USDA Western Human Nu­

The PETRA food scales are simple to op­ trition Research Center developed the Nutri­

erate, and the system makes it difficult for the tion Evaluation Scale System (NESSy) (Fong

subject to modify the digitally coded food and Kretsch, 1990) to replace the manually

record. Consequently, the habitual food in­ weighed food record. NESSy is a comput­

take is more likely to be truthfully recorded erized method that uses interactive software

than in the conventional weighed food record to prompt and guide participants through the

(Bingham, 1987; Bingham et al., 1994). As recording of food weights and descriptions.

a result, the PETRA food scales were used in It is designed to speed up the recording of

the validation of the dietary assessment meth­ a weighed food record and can be used by

ods used in the U.K. EPIC study (Bingham participants in their own homes. Use of

etal.,2001). NESSy to record food intake saves about

Tape recorders have been perceived to be 80% in time and labor, and it yields ac­

especially useful for populations facing mem­ curate data at both the group and individual

ory or visual limitations such as the elderly levels (Kretsch and Fong, 1990). The USDA

and children. However, when Lindquist et al. Agricultural Research Service is evaluating

(2000) compared energy intakes of children, the use of NESSy in association with internet

based on both tape-recorded food records data transfer, allowing expansion of the capa­

and recall interviews, with energy expendi­ bilities of the system for use by the lay person

ture measured by the doubly labeled water (Consumer NESSy) and for the dietetic pro­

technique, the use of a tape recorder did fessional (ProNeSSy).

not result in more accurate assessments of Several large-scale national nutrition sur­

energy intake than the recall method, espe­ veys, including the EPIC study, the 1988-

cially among the older children ( 1 2 - 1 5 y). 1994 NHANES ill, the USDA nationwide

household food consumption survey, and the

New Zealand National Nutrition Survey and

3.2.5 Microcomputers
Children's Nutrition Survey (Parnell et al.,

Stockley et al. ( 1986a, 1986b) were among 2003), have adopted the use of microcom­

the first investigators to develop a computer­ puter-based automated dietary interviews to

ized system for recording food intake. The standardize 24-h recall procedures and to

device consisted of a digital balance with a automate data entry. In general, these sys­

capacity of 1 kg interfaced to a specially de­ tems allow all the recall data to be entered

veloped microcomputer with a keyboard. The directly into the computer at the time of the

latter had an upper bank of color-coded con­ interviews.

trol keys which registered "start," "waste," In the EPIC study, standardization of the

"mixed waste," "no waste," and "done," and 24-h recall interviews in the 23 European
54 Principles of Nutritional Assessment

centers ( 1 0 countries in total) was achieved ethnicity, and education. The program auto­

with the aid of a software program (EPIC­ mates data entry and ensures that responses

SOFf) (Slimani et al., 1999). EPIC-SOFf are complete by encouraging subjects to re­

was designed specifically to prevent memory view and correct inconsistent data.

deficiency, to standardize the identification Computerized systems have also been de­

and description of foods, for quantification veloped especially for use by persons with

of portion sizes, and for handling of recipes. low literacy (Ammerman et al. 1994 ). These

The specific approaches used are described in interactive multimedia-based dietary assess­

Slimani et al. (2000). ment tools will have an increasing role in

Laptop computers were used in the CS­ improving the validity of the dietary data col­

FII for the first time in 1 9 8 7 - 1 9 8 8 , to assist lected and in advancing our understanding of

with interviewing (Section 2 . 3 ) . The com­ relationships between diet and disease.

puters were programmed to handle a food

list of nearly 3000 items and default codes

for foods that subjects have difficulty remem­ 3.3 Selecting an appropriate method

bering. The default codes represent the most

commonly used items in a specific food cat­ The method of choice for assessing food or

egory. For example meat that is not further nutrient intakes depends primarily on the ob­

specified may default to regular ground, pan­ jectives of the study. No method is devoid

cooked hamburger (USDA, 1998). of random or systematic errors (Chapter 5),

In 2002, a new, automated, multiple-pass or prevents alterations in the food habits of

24-h recall interview procedure was intro­ the subjects. Table 3.6 summarizes the most

duced by USDA in their nationwide food appropriate methods for assessing food or

survey. In this revised method, there are nutrient intakes in relation to four possible

a number of built-in cues to help jog the levels of objectives.

respondent's memory. In addition, all ques­ Note that the number and selection of repli­

tions, prompts, and details about the food and cate 24-h recalls, or weighed or estimated 1-d

how it was prepared are computerized. food records required to obtain level two,

In the New Zealand 1997 National Nutri­ level three, or level four data, depends on

tion Survey (MOH, 1997), and the Children's the day-to-day variation within one individual

Survey (Parnell et al., 2003), a multiple-pass (i.e., within-subject variation) of the nutri­

24-h recall direct data-capture program was ent of interest (Section 6.2). This variation

developed to reflect the unique needs of the depends on the nutrient, the study popula­

New Zealand surveys. The program was tion, the dietary survey method used, and the

based on the direct data-capture program de­ seasonal variations of intake. Nonconsecu­

veloped by the University of Minnesota (Mc­ tive days should be selected when possible,

Dowell e t al . , 1990). Bar code scanners were to enhance the statistical power of the in­

also used in these New Zealand surveys to formation: day-to-day correlations between

improve the accuracy of the product name intakes often occur when food intake data

information for branded items. are collected over consecutive days. The

A different approach has been developed length of time needed between the observa­

by Kohlmeier et al. (1997). The system is tion days also depends on the nutrient (IOM,

known as computer-assisted self-interviewing 2000). Generally, for nutrients found in high

(CASI), and it permits the use of in-depth concentrations in only a few foods, such as

questionnaires using a microcomputer but vitamins A and D and cholesterol, the number

without interviewers. The system also pro­ of replicates needed is greater than for those

vides respondents with an evaluation of their found in a wide range of foods (e.g., protein).

reported nutrient intakes. A prototype CASI Additional factors that should be consid­

diet history program has been tested in sev­ ered when choosing a method for assess­

eral U.S. focus groups of mixed age, sex, ing the food consumption of individuals are
 Measuring food consumption of individuals 55

Level Desired information Preferred approach

One Mean nutrient intake of a group A single 24-h recall, or single weighed or estimated food

record, with large number of subjects and adequate repre-

sentation of all days of the week

Two Proportion of population "at risk" Replicate observations on each individual or a subsample

using 24-h recalls or weighed or estimated l-d food records

Three Usual intakes of nutrients in individ- Multiple replicates of 24-h recalls or food records or a semi-

uals for ranking within a group quantitative food frequency questionnaire

Four Usual intakes of foods or nutrients Even larger number of recalls or records for each indiv-

in individuals for counseling or for idual. Alternatively, a semiquantitative food frequency

correlation or regression analysis questionnaire or a dietary history can be used.

Table 3.6: Selection of methodology to measure nutrient intakes to meet four possible levels of

objectives. Modified from Gibson (2002).

the characteristics of the subjects within the required and the day-to-day vanation bet­

study population, the respondent burden of ween subjects in the nutrient intakes (NRC,

the method, and the available resources. For 1 9 8 6 ; Sempos et al., 1 9 9 1 ; IOM, 2000). The

instance, certain methods are unsuitable for appropriate formula is

elderly subjects with poor memories, for busy 2

n = s�/e
mothers with young children, or for illiter­

ate subjects. Other methods require highly where st is the between-subject variance of

trained personnel and specialized laboratory the nutrient of interest, and e is the desired

and computing facilities, which may not be standard error- a measure of precision re­

available. Generally, the more accurate meth­ quired for the estimate of the mean intake of

ods are associated with higher costs, greater the nutrient of interest.

respondent burden, and lower response rates. Obviously, to use this formula, an estimate

Unfortunately, compromises often have to be of between-subject variation for the nutri­

made between the collection of precise data ent of interest is required. This is usually

on usual nutrient intakes of individuals and a obtained from the literature but may be deter­

high response rate. mined during a pilot study. This calculation

should be repeated for each of the nutrients

of interest, and the largest n (i.e., the worst

3.3.1 Determining the mean nutrient
case) should be used if possible.
intake of a group: level one
As an example, assume the expected mean

Level one is the easiest objective to achieve iron intake obtained from the literature is

and can be met by measuring the food intake 10 mg/d, with an anticipated standard devi­

of each subject in the group using a single ation (s) of 3 mg/d. Also assume that the

24-h recall or a 1-d food record, provided desired margin of error in the expected mean

the subjects are representative of the study is from 9.2 to 10.8 mg/d, or 1 . 6 mg/d. If

population and all the days of the week are a precision of 95% is required, the margin

equally represented in the final sample (Cole of error used is ± twice the standard error
and Black, 1984). Data on average usual and e = 1.6/4 (i.e., e = 0.4). If a precision of

nutrient intakes of a group can be used for 99% is required, the margin of error used is

international comparisons across countries of ± three times the standard error, and then

the relationship of nutrient intakes to health e = 1.6/6 (i.e., e = 0 . 2 7 . Using a precision of

and disease. 95%, then

The size of the group (n) necessary to 2 2

n = 3 / 0.4 = 56 subjects
characterize the group mean usual nutrient

intake depends on the degree of precision Alternatively, if a precision of 99% is used

56 Principles of Nutritional Assessment

then more subjects must be studied: using the method outlined by the National

2 2
Research Council (NRC, 1986), or a more­
n = 3 / 0.27 = 123 subjects
refined NRC approach developed by Nusser

If the study objective is to demonstrate a sig­ et al. ( 1996) using SAS and PROC IML, or

nificant difference in the mean intakes of two using the program PC-SIDE. The latter pro­

groups, or a significant change in the mean in­ gram can be downloaded from

takes, based on unpaired or paired data, then

http://www.iastate.edu/
alternative formulae must be applied; details

are given in Gibson and Ferguson (1999). The adjustment process provides estimates of

the usual nutrient intakes for each specified

age- and gender-specific subgroup. An ex­

3.3.2 Calculating the population
ample comparing the adjusted distributions
percentage "at risk": level two
of usual zinc intakes (using the refined NRC

To determine the percentage of the popula­ approach of Nusser et al., 1996) with the ob­

tion "at risk" of inadequate nutrient intakes, served zinc intakes for New Zealand adult

estimates of the usual intakes of the subjects females aged 19-50y is shown in Figure3.3.

are required. This, in turn, requires that The adjustment process used yields a distri­

the food consumption of subjects be meas­ bution with reduced variability that preserves

ured over more than 1 day. Hence, repeated the shape of the original observed distribution

24-h recalls, or replicate weighed or esti­ (Gibson et al., 2003).

mated 1-d food records are the methods of The adjusted distribution of "usual" nutri­

choice. Often, it is not feasible to carry out ent intakes can then be used to predict the

repeated observations on all the subjects, as proportion of the population at risk of nutri­

in the case of a national dietary survey, and ent inadequacy using either the full probabil­

the recalls or records are repeated on a sub­ ity approach, or the Estimated Average Re­

sample of the subjects only. quirement (EAR) cutpoint method; details are

To achieve a level two objective, at least given in Sections 8 . 3 . 1 and 8 . 3 . 2 . Figure 3 . 3

two independent measurements of food in­ shows that, in this particular case, adjusting

take should be obtained on at least a represen­ the distribution significantly reduces the pro-

tative subsample of individuals in the survey.

The U.S. Food and Nutrition Board (IOM,

2000) recommends that the replicate meas­ I '\ - 1 - d intake

0.12 I I
urements should be independent and made I I
- - Usual intake

I I
on nonconsecutive days. However, if the I I
Q)
I

data can be collected only on consecutive ""

ra I

-� 0.08 I

days, then three daily measurements should I

be used. The subsample should consist of 30 1,- I

c I
Q)
I
to 40 individuals who represent the age range
5- 0.04 \

of the sample. Note that it is more important e

u,

to have a minimum number of replicate ob­

servations in the subsample than a minimum 0.00 J'._.L_-1, __;_::�:::::===�-.J

proportion of replicate observations. 0 10 20 30 40

Zinc (mg/d)
Once a series of replicate observations on

at least 30 individuals have been obtained,

Figure 3 . 3 : Estimates of usual intake distribution for
an adjustment can be made to the observed zinc for New Zealand adults obtained from 24-h recall

distribution of intakes to remove the vari­ data and adjusted with replicate intake data using the

refined NRC method. The y-axis (frequency of intake)

ability introduced by day-to-day variation in
shows the likelihood of each level of intake in the
nutrient intakes within an individual, i.e., to
population. EAR, Estimated Average Requirement.
remove the within-subject variation (Section From Gibson et al., Nutrition Today 38: 63-70, 2003

6.2.3). Such an adjustment can be performed © Lippincott Williams & Wilkins.

 Measuring food consumption of individuals 57

portion of subjects considered to have intakes days required to achieve the level three ob­

below the EAR. jective can be calculated from the ratio of

Within-subject variation can also have a the within- to the between-subject variation

significant effect on estimates of the preva­ in nutrient intakes (i.e., the variance ratio).

lence of abnormally high nutrient intakes. Sometimes the latter can be obtained from

Table 3 . 7 shows data from NHANES II. The the literature, again preferably from an earlier

large differences between the observed preva­ study on a comparable group. Alternatively, a

lence and the calculated "true" prevalence pilot study may be necessary to obtain this in­

represent the effect of removing the within­ formation.

subject variation by calculation. In this case, Several authors have developed equations

very large numbers of repeated measurements for calculating the number of replicate days

on each subject are required to reduce the ob­ required to meet level three objectives (Liu

served prevalence of abnormally high intakes et al., 1978; Black et al., 1983; Marr and

of cholesterol and calcium to within 5% of Heady, 1986; Basiotis et al., 1 9 8 7 ; Nelson

the true prevalence (Sempos etal., 1 9 9 1 ) . et al., 1989). Black et al. ( 1 9 8 3 ) suggest using

Data on the distribution of usual nutrient the following formula for the number of days

intakes of a population are essential for na­ (n) of diet records needed:

tional food policy development and food for­ 2 2))

n = (r /(1 - r x (s�/s�)
tification planning. Food patterns associated

with inadequate nutrient intakes can also be In this equation, r is the unobservable cor­

identified using this approach, enabling food­ relation between the observed and true mean

assistance programs to be designed and im­ intakes of individuals over the period of ob­

provements in nutrition education made. servation, and s! and sl are the observed

within- and between-subject variances, re­

spectively. This equation should be used

3.3.3 Ranking individuals by food or
in association with Table 3 . 8 which shows
nutrient intake: level three
the proportion of subjects correctly and in­

When the study objective is at level three and correctly classified in the extreme fractions

involves ranking individuals within a group, for different values of the correlation coeffi­

often for the purpose of linking dietary in­ cient between the observed and true intakes

takes with risk of chronic disease, the pre­ (r). The value of r chosen will depend on the

ferred approach is to obtain multiple obser­ degree of misclassification that the investiga­

vations on each individual. The number of tor is prepared to accept.

As an example, assume that the investiga­

tor requires that when the subjects are divided

Prevalence (%) No. of repeated
into terciles, fewer than 5% ( < 0.05) of the

Variable "True" Observed measurements subjects are grossly misclassified into the op­

posite tercile. This will require an r value of

Cholesterol
0.75 (Table 3 . 8 ) . Assumings!/ sl = 1 . 7 , then
>300mg 15 37 39
2 2))
number of days= (r /(1 - r x 1.7
Calcium
2 2)
> 800mg 12 21 9 = 0.75 / ( 1 - 0.75 x 1.7

= 3days

Table 3 . 7 : "True" and observed prevalence estimates

The number of days needed to generate a
and the number of repeated measurements needed to

reduce the observed prevalence to within 5% of the given r increases as the chosen r increases.

"true" prevalence. Data for women aged 45-54 y If the size of the within-subject variation (s!)
from NHANES II (1976-1980). From: Sempos CT,
in nutrient intake is small compared with the
Looker AC, Johnson CL, Woteki CE. (1991). The
size of the between-subject (sl) variation,
importance of within-person variability in estimating

prevalence. In: Macdonald I (ed.), Monitoring Diet­ then fewer replicate days are needed to meet

ary Intakes, pp. 99-109. © Springer-Verlag, Berlin. level three objectives.

58 Principles of Nutritional Assessment

the most difficult to obtain. Large num­

Correctly and incorrectly

classified into extreme fraction

bers of measurement days for each individual

are required using 24-h recalls or estimated or

r Thirds Fourths Fifths
weighed food records.

An estimate of the within-subject variation

0.75 a 0.69 0.63 0.59

b 0.049 0.013 0.004 for each nutrient of interest should be ob­

tained from the literature, preferably from an

0.80 a 0.72 0.68 0.65

b 0.033 0.006 0.002 earlier study on a comparable group or a pilot

0.85 a 0.76 0.72 0.69

study, as noted earlier. This estimate may be

b 0.018 0.002 <0.001 expressed as the variance, s!; standard de­

0.90 a 0.80 0.77 0.75 viation, Sw; or as the coefficient of variation

b 0.006 < 0.001 <0.001 (CV w) expressed as a percentage:

0.95 a 0.86 0.84 0.83

CV w = Sw /mean level of intake x 100%
b <0.001 <0.001 <0.001

This estimate can be used in the following

Table 3.8: Proportion of subjects correctly and incor­
equation to determine the number of days re­
rectly classified in the extreme fractions for different
quired per subject to estimate an individual's.
values of the correlation coefficient (r). a, correctly
nutrient intake to within 20% of their true
classified in the extreme thirds, fourths, or fifths of

the distribution of intakes; b, misclassified into the mean 95% of the time (Beaton et al., 1979):
opposite extreme fraction. From Nelson et al., Amer­

ican Journal of Clinical Nutrition 50: 155-167, 1989 n = (Z 0 CVw/Do)2

© Am J Clin Nutr. American Society for Clinical Nu­

where n = the number of days needed per
trition.
subject, Za = the normal deviate for the per­

The U . S . subcommittee on criteria for diet­ centage of times the measured value should

ary evaluation (NRC, 1986) recommended be within a specified limit (i.e., 1 . 9 6 in the ex­

using independent days for replicating the ample below), CV w = the within-subject co­

measurements of l-d nutrient intakes to re­ efficient of variation (as a percentage), and

duce any effect of autocorrelation between Do = the specified limit (as a percentage of

intakes on adjacent days. long-term true intake) (i.e., 20% in the exam­

An alternative approach to achieving level ple given below).

three objectives is to use a semiquantitative The following example illustrates how to

food frequency questionnaire This approach calculate the number of days required to

is often used in epidemiological investiga­ estimate a Malawian woman's zinc intake

tions to study associations between intakes using 24-h recalls to within 20% of the true

and risk of disease and does not require mean, 95% of the time. In this example, the

a measurement of absolute nutrient intakes. CVw (i.e., 34%) for zinc intakes on Malaw­

Although this approach is much simpler, in­ ian women via 24-h recalls is taken from

volving only a single interview with each the literature (Hotz and Brown, 2004). Thus

subject, it is difficult to quantify the errors Z0 = 1 . 9 6 and CV w = 34%. Then:

involved and to separate the effects of within­ 2

n = (1.96 x 34%/20%) = 1 1 days
and between-subject variance.

If a pilot study is undertaken in which repli­

cate 24-h recalls are conducted, then the

3.3.4 Determining usual intakes of nutri-
actual CV w for each nutrient of interest can
ents of individuals: level four
be calculated. In this way, the estimate

Reliable estimates of usual food or nutrient of the number of days required to measure

intakes of individuals that can be used with the usual intake of each of the nutrients of

confidence to meet a level four objective, interest in an individual, with a required de­

involving correlation or regression analysis gree of precision, can be defined. In general,

with individual biochemical measures, are considerably more days are required to obtain
 Measuring food consumption of individuals 59

Method and Procedures Uses and Limitations

24-h recall. Subject or caretaker recalls food Useful for assessing average usual intakes of a large population, provided

intake of previous 24-h in an interview. Quan- that the sample is truly representative and that the days of the week are

tities estimated in household measures using adequately represented. Used for international comparisons of relation-

food models as memory aids or to assist ship of nutrient intakes to health and susceptibility to chronic disease.

in quantifying portion sizes. Nutrient intakes Inexpensive, easy, quick, with low respondent burden so that compliance

calculated using food composition data. is high. Large coverage possible; can be used with illiterate individuals.

Element of surprise so less likely to modify eating pattern. Single 24-h

recalls likely to omit foods consumed infrequently. Relies on memory

and hence unsatisfactory for the elderly and young children. Multiple

replicate 24-h recalls used to estimate usual intakes of individuals.

Estimated food record. Record of all food Used to assess actual or usual intakes of individuals, depending on

and beverages "as eaten" (including snacks), number of measurement days. Data on usual intakes used for diet

over periods from one to seven days. Quanti- counseling and statistical analysis involving correlation and regression.

ties estimated in household measures. Nutrient Accuracy depends on the conscientiousness of subject and ability to

intakes calculated using food composition data . . estimate quantities. Longer time frames result in a higher respondent

burden and lower cooperation. Subjects must be literate.

Weighed food record. All food consumed over Used to assess actual or usual intakes of individuals, depending on the

a defined period is weighed by the subject, care- number of measurement days. Accurate but time consuming. Setting

taker, or assistant. Food samples may be saved must permit weighing. Subjects may change their usual eating pattern to

individually, or as a composite, for nutrient simplify weighing or to impress investigator. Requires literate, motivated,

analysis. Alternatively, nutrient intakes calcu- and willing participants. Expensive.

lated using food composition data.

Dietary history. Interview method consisting Used to describe usual food or nutrient intakes over a relatively long

of a 24-h recall of actual intake, plus informa- time period, which can be used to estimate prevalence of inadequate

tion on overall usual eating pattern, followed intakes. Such information is used for national food policy development,

by a food frequency questionnaire to verify and for food fortification planning, and to identify food patterns associated

clarify initial data. Usual portion sizes recorded with inadequate intakes. Labor-intensive, time-consuming, and results

in household measures. Nutrient intakes calcu- depend on skill of interviewer.

lated using food composition data.

Foo d frequency qu estionnaire. Uses com- Designed to obtain q ualitative, descriptive data on usual intakes of foods

prehensive or specific food item list to record or classes of foods over a long time period. Useful in epidemiological

intakes over a given period (day, week, month, studies for ranking subjects into broad categories of low, medium,

year). R ecord is obtained by interview or and high intakes of specific foods, food components, or nutrients, for

self-administered questionnaire. Questionnaire comparison with the prevalence or mortality statistics of a specific dis-

can be semiquantitative when subjects asked to ease. C an also identify food patterns associated with inadequate intakes

quantify usual portion sizes of food items, with of specific nutrients. Method is rapid, with low respondent burden and

or without the use of food models. high response rate, but accuracy is lower th an for other methods.

Table 3.9: Uses and limitations of methods used to assess the food consumption of individuals.

reliable estimates of intakes of individuals to finite time period are required. For such data,

meet the level four objective, compared with weighed food records .(Section 3 . 1 . 4 ) , com­

level three (i.e., relative ranking of subjects pleted for the duration of the study period, are

into groups) (Palaniappan et al., 2003). the recommended method. Nutrient intakes

Sometimes, dietary histories or semiquant­ can then be calculated using food composi­

itative food frequency questionnaires are used tion data. Alternatively, duplicate meals can

to obtain this level four data on usual nutri­ be collected throughout the period for later

ent intakes for correlation with biomarkers analysis; details are given in Section4.7.

(Jacques et al., 1993). Some investigators

emphasize, however, that the accuracy of a

semi-quantitative food frequency question­ 3.4 Summary

naire is only equivalent to two to three repeat

24-h recalls (Sempos et al., 1999). Details of the available methods for assess­

In some experimentally controlled studies ing the food consumption of individuals, and

such as balance studies, information on the their uses and limitations, are summarized in

actual nutrient intakes of an individual over a Table3.9.

60 Principles of Nutritional Assessment

Technical improvements for measuring the tion on patterns of food use over a relatively

food consumption of individuals include tele­ long time period. With certain modifications,

phone-assisted approaches, even in national they can also provide data on usual nutrient

surveys; use of photographs as memory aids intakes for level three, and sometimes four

and to quantify portion sizes consumed; elec­ objectives.

tronic devices for recording food intakes di­

rectly (e.g., tape recorders); and use of mi­

crocomputers, which can be used to automate References

both dietary interviews and data entry, in an

effort to standardize interviews, as well as the Acheson KJ, Campbell IT, Edholm OG, Miller DS,

Stock MJ. (1980). The measurement of food and

identification and description of foods, espe­
energy intake in man - an evaluation of some tech­
cially in multicenter epidemiological studies.
niques. American Journal of Clinical Nutrition 3 3 :
Such developments aim to reduce respon­ 1147-1154.

dent and interviewer burden and hence in­ Ambrosini GL, van Roosbroeck SAR, Mackerras D,

Fritschi L, de Klerk NH, Musk AW. (2003). The

crease compliance; reduce errors resulting
reliability of ten-year dietary recall: implications for
from memory lapses and incorrect estimation
cancer research. Journal of Nutrition 133: 2663-
of portion sizes; and, in the case of electronic 2668.

devices, reduce interviewer biases and elim­ Ammermann AS, Kirkley BG, Dennis B., Hohenstein

inate the tedious process of coding the food C, Allison A, Strecher VJ, Bulger D. (1994). A

dietary assessment for individuals with low literacy

records.
skills using interactive touch-screen computer tech­
To characterize the average usual intake of
nology. American Journal of Clinical Nutrition 59:
a large group (i.e., level one), a 24-h recall 289 (abstr.).

or record over a l-d period is the method Barr SI, Petit MA, Vigna YM, Prior JC. (2001). Eating

attitudes and habitual calcium intake in peripubertal

of choice, provided all days of the week are
girls are associated with initial bone mineral content
proportionately represented and the sample is
and its change over 2 years. Journal of Bone and
representative of the population under study. Mineral Research 16: 940-947.

To determine the proportion of the popula­ Basiotis PP, Welsh SO, Cronin FJ, Kelsay JL, Mertz W.

tion "at risk" (i.e., level two) of inadequate (1987). Number of days of food intake records

required to estimate individual and group nutrient

nutrient intakes, replicate observations on
intakes with defined confidence. Journal of Nutri­
each individual or a subsample of subjects
tion 1 1 7 : 1 6 3 8 - 1 64 1 .
are required, whereas for assessing usual Basiotis PP, Carlson A, Gerrior SA, Juan WY, Lino M.

nutrient intakes in individuals for ranking (2002). The Healthy Eating Index: 1999-2000, U.S.

Department of Agriculture, Center for Nutrition

within a group (i.e., level three), multiple
Policy and Promotion., Washington DC. CNPP-12.
replicates of 24-h recalls or weighed or es­
www.cnpp.usda.gov
timated l-d food records are recommended.
Beaton GH, Milner J, Corey P, McGuire V, Cousins M,
Alternatively, a semiquantitative food fre­ Stewart E, de Ramos M, Hewitt D, Grambsch PV,

quency questionnaire can be used. An even Kassim N, Little JA. (1979). Sources of variance

in 24-hour dietary recall data: implications for nu­

larger number of multiple replicates of 24-h
trition study, design and interpretation. American
recalls or food records for each individual
Journal of Clinical Nutrition 32: 2546-2559.
are needed for individual diet counseling or Bingham SA. (1987). The dietary assessment of in­

correlation and regression analysis (i.e., level dividuals; methods, accuracy, new techniques and

recommendations. Nutrition Abstracts and Reviews

four). The number, spacing, and days se­
(Series A) 57: 705-742.
lected for all these measurements depend
Bingham SA, Gill C, Welch A, Day K, Cassidy A,
on the day-to-day variation (within-subject
Khaw KT, Sneyd MJ, Key TJ, Roe L, Day NE.

variation) of the nutrient of interest, which, (1994). Comparison of dietary assessment meth­

in turn, is affected by the study population, ods in nutritional epidemiology: weighed records

v. 24 h recalls, food frequency questionnaires and

dietary survey method used, and seasonal
estimated-diet record. British Journal of Nutrition
variations in intake.
72: 619-643.
The food frequency and dietary history Bingham SA, Welch AA, McTaggart A, Mulligan AA,

questionnaires yield retrospective informa- Runswick SA, Luben R, Oakes S, Khaw KT, Ware-
 Measuring food consumption of individuals 61

ham N, Day NE. (2001). Nutritional methods in intakes of energy, fats and cholesterol among Singa­

the European Prospective Investigation of Cancer in poreans. Asia Pacific Journal of Clinical Nutrition

Norfolk. Public Health Nutrition 4: 847-858. 9: 282-288.

Black AE. (1982). The logistics of dietary surveys. Dreon DM, John EM, DiCiccio Y, Whittemore AS.

Human Nutrition: Applied Nutrition 36: 85-94. (1993). Use of NHANES data to assign nutrient

Black AE, Cole TJ, Wiles SJ, White F. ( 1 9 8 3 ) . Daily densities to food groups in a multi-ethnic diet history

variation in food intake of infants from 2 to 18 questionnaire. Nutrition and Cancer 20: 223-230.

months. Human Nutrition: Applied Nutrition 37: Dufour D, Staten LK, Waslien CI, Reina JC, Spurr GB.

448-458. (1999). Estimating energy intake of urban women in

Block G, Hartman AM, Dresser CM, Carroll MD, Gan­ Colombia: comparison of diet records and recalls.

non J, Gardner L. (1986). A data-based approach American Journal of Physical Anthropology 108:

to diet questionnaire design and testing. American 53-63.

Journal of Epidemiology 124: 453-469. Dwyer JT, Coleman KA. (1997). Insights into dietary

Block G, Hartman AM, Naughton D. (1990). A re­ recall from a longitudinal study: accuracy over four

duced dietary questionnaire: development and val­ decades. American Journal of Clinical Nutrition 65:

idation. Epidemiology 1 : 58-64. 1153S-1158S.

Bogle M, Stuff J, Davis L, Forrester I, Strickland E, Eck LH, Klesges RC, Hanson CL. (1989). Recall of a

Casey PH, Ryan D, Champagne C, McGee B, Mel­ child's intake from one meal: are parents accurate?

lad K, Neal E, Zaghloul S, Yadrick K, Horton J. Journal of the American Dietetic Association 89:

(2001). Validity of a telephone-administered 784-789.

24-hour dietary recall in telephone and non-tele­ Elwood PC, Bird G. (1983). A photographic method

phone households in the rural Lower Mississippi of diet evaluation. Human Nutrition: Applied Nutri­

Delta region. Journal of the American Dietetic As­ tion 37: 474-477.

sociation 1 0 1 : 216-222. Euronut-SENECA. (1991). Nutrition and the elderly

Bowman SA, Lino M, Gerrior SA, Basiotis PP. (1998). in Europe. European Journal of Clinical Nutrition

The Healthy Eating Index: 1994-96. U.S. Depart­ 45(suppl 3): 1-196.

ment of Agriculture, Center for Nutrition Policy and Ferguson EL, Gadowsky SL, Huddle J-M, Cullinan TR,

Promotion, Washington, DC. CNPP-5. Gibson RS. (1995). An interactive 24-h recall tech­

http://www.cnpp.usda.gov/ nique for assessing the adequacy of trace mineral

Burke BS. (1947). The dietary history as a tool in intakes of rural Malawian women: its advantages

research. Journal of the American Dietetic Associa­ and limitations. European Journal of Clinical Nu­

tion 23: 1041-1046. trition 49: 565-578.

Buzzard IM, Sievert YA. (1994). Research priori­ Feunekes GU, van Staveren WA, De Vries JHM, Bu­

ties and recommendations for dietary assessment rema J, Hautvast JGAJ. (1993). Relative and bio­

methodology: First International Conference on marker based validity of a food-frequency ques­

Dietary Assessment Methods. American Journal of tionnaire estimating intake of fats and cholesterol.

Clinical Nutrition 59: 275S-280S. American Journal of Clinical Nutrition 58: 489-496.

Cameron ME, van Staveren WA. (eds.) (1988). Man­ Finch S, Doyle W, Lowe C, Bates CJ, Prentice A,

ual on Methodology for Food Consumption Studies. Smithers G, Clarke PC. (1998). National Diet and

Oxford University Press, Oxford. Nutrition Survey: People Aged 65 Years and Over.

Casey PH, Goolsby SL, Lensing SY, Perloff BP, Bogle Vol. 1 : Report of the Diet and Nutrition Survey. The

ML. (1999). The use of telephone interview method­ Stationery Office. London.

ology to obtain 24-hour dietary recalls. Journal of Fong AK, Kretsch MJ. (1990). Nutrition Evaluation

the American Dietetic Association 99: 1 4 0 6- 1 4 1 1 . Scale System reduces time and labor in recording

Cleveland LE, Ingwersen LA. (2001). Was it a slab, a quantitative dietary intake. Journal of the American

slice, or a sliver? High-tech innovations take food Dietetic Association 90: 664-670.

survey to new levels. Agricultural Research, March Galasso R, Panico S, Celentano E, Del Pezzo M.

2001: 4-7. www.ars.usda.gov/is/AR/archive/ (1994). Relative validity of multiple telephone ver­

Cole TJ, Black AE. (1984). Statistical aspects in the sus face-to-face 24-hour dietary recalls. Annals of

design of dietary surveys. In: The Dietary Assess­ Epidemiology 4: 332-336.

ment of Populations. Medical Research Council En­ Galeazzi MAM, de Meireles AJ, de Toledo Viana RP,

vironmental Epidemiology Unit. Scientific Report Zabotto CB, Domene SAM. (1966). Registro fo­

No. 4. Medical Research Council, Southampton, tografico inqueitos dieteticos. Ministry of Health,

U.K. National Institute of Food and Nutrition, Goiania,

CSFII/DHKS. (1994-96). The Continuing Survey of Portugal.

Food Intakes by Individuals (CSFII) and the Diet Gibson RS. (2002). Dietary assessment. In: Mann J,

and Health Knowledge Survey (DHKS), 1994-96. Truswell AS. (eds.) Essentials of Human Nutrition

www.barc.usda.gov/bhnrc/foodsurvey/ (2nd ed.). Oxford University Press, Oxford, pp.

Deurenberg-Yap M, Li T, Tan WL, van Staveren WA, 449-466.

Deurenberg P. (2000). Validation of a semiquantita­ Gibson RS, Ferguson EL. (1999). An Interactive

tive food frequency questionnaire for estimation of 24-Hour Recall for Assessing the Adequacy of Iron
62 Principles of Nutritional Assessment

and Zinc Intakes in Developing Countries. Internat­ ence Intakes: Applications in Dietary Assessment.

ional Life Sciences Institute Press, Washington, DC. National Academy Press, Washington, DC.

Gibson RS, McKenzie JE, Ferguson EL, Parnell WR, Jacques PF, Sulsky SI, Sadowski JA, Phillips JCC,

Wilson NC, Russell DG. (2003). The risk of inade­ Rush D, Willett WC. (1993). Comparison of micro­

quate zinc intake in United States and New Zealand nutrient intake measured by a dietary questionnaire

adults. Nutrition Today 38: 63-70. and biochemical indicators of micronutrient sta­

Gregory J, Foster K, Tyler H, Wiseman M. (1990). The tus. American Journal of Clinical Nutrition 57:

Dietary and Nutritional Survey of British Adults. 182-189.

The Stationery Office, London. Kant AK, Schatzkin A, Graubard B, Schairer C. (2000).

Gregory J, Collins DL, Davies PSW, Hughes JM, A prospective study of diet quality and mortality in

Clarke PC. (1995). National Diet and Nutrition women. Journal of the American Medical Associa­

Survey: Children Aged 1.5-4.5 Years. Volume 1: tion 283: 2109-2115.

Report of the Diet and Nutrition Survey. The Sta­ Kesse E, Clavel-Chapelon F, Slimani N, van Liere M,

tionery Office, London. E3N Group. (2001). Do eating habits differ accord­

Gregory J, Lowe S, Bates CJ, Prentice A, Jackson LV, ing to alcohol consumption? Results of a study of

Smithers G, Wenlock R, Farron M. (2000). National the French cohort of the European Prospective In­

Diet and Nutrition Survey: Young People Aged 4 vestigation into Cancer and Nutrition (E3N-EPIC).

to 18 Years. Volume 1: Report of the Diet and American Journal of Clinical Nutrition 74: 322-327.

Nutrition Survey. The Stationery Office, London. Kohlmeier L, Mendez M, McDuffie J, Miller M. ( 1997).

Haapa E, Toponen T, Pietinen P, Rasanen L. (1985). Computer-assisted self-interviewing: a multimedia

Annoskuvakirja (Portion picture booklet). National approach to dietary assessment. American Journal

Public Health Institute and the Department of Nu­ of Clinical Nutrition 65: 1275S-1281S.

trition, University of Helsinki. (Available from Krantzler NJ, Mullen BJ, Schutz HG, Grivetti LE,

authors.) Holden CA, Meiselman HL. (1982). Validity of

Health and Welfare Canada. (1973). Nutrition Canada telephoned diet recalls and records for assessment

National Survey. Health and Welfare, Ottawa. of individual food intake. American Journal of Clin­

Henderson L, Gregory J, Swan G. (2002). National ical Nutrition 36: 1234-1242.

Diet and Nutrition Survey: Adults Aged 19 to 64. Kretsch MJ, Fong AKH. (1990). Validation of a new

Volume 1 : Types and Quantities of Foods Con­ computerized technique for quantitating individual

sumed. The Stationery Office, London. dietary intake: the Nutrition Evaluation Scale Sys­

Henderson L, Gregory J, Irving K, Swan G. (2003a). tem (NESSy) vs the weighed food record. American

Diet and Nutrition Survey: Adults Aged 19 to 64. Journal of Clinical Nutrition 5 1 : 477-484.

Volume 2: Energy, Protein, Carbohydrate, Fat and Levi F, Pasche C, Lucchini F, La Vecchia C. (2000).

Alcohol Intake. The Stationery Office, London. Selected micronutrients and colorectal cancer, a case

Henderson L, Irving K, Gregory J, Bates CJ, Prentice A, control study from the canton in Vaud, Switzerland.

Perks J, Swan G, Farron M. (2003b). National European Journal of Cancer 36: 2 1 1 5 - 2 1 1 9 .

Diet and Nutrition Survey: Adults Aged 19 to 64. Lindquist CH, Cummings T, Goran MI. (2000). Use

Volume 3: Vitamin and Mineral Intake and Urinary of tape-recorded food records in assessing children's

Analytes. The Stationery Office, London. dietary intake. Obesity Research 8: 2 - 1 1 .

Hercberg S, Deheeger M, Preziosi P. (1994). Portions Liu K, Starnler J, Dyer A, McKeever J, McKeever P.

alimentaires: manuel photos pour !'estimation des (1978). Statistical methods to assess and mini­

quantites. Institut Scientifique et Technique de la mize the role of intra-individual variability in ob­

Nutrition et de I' Alimentation, Conservatoire Na­ scuring relationships between dietary lipids and

tional des Arts et Meetiers, (CNAM) Paris Candia serum cholesterol. Journal of Chronic Diseases 3 1 :

Polytechnica, France. 399-418.

Holick CN, Michaud DS, Stolzenberg-Solomon R, Livingstone MB, Robson PJ. (2000). Measurement

Maynes ST, Pietinen P, Taylor PR, Virtamo J, Al­ of dietary intake in children. Proceedings of the

banes D. (2002). Dietary carotenoids, serum beta­ Nutrition Society 59: 279-293.

carotene, and retinol and risk of lung cancer in the Lyu LC, Hankin JH, Liu LQ, Wilkens LR, Lee JH,

alpha-tocopherol, beta-carotene cohort study. Amer­ Goodman MT, Kolonel LN. (1998). Telephone

ican Journal of Epidemiology 156: 536-547. vs face-to-face interviews for quantitative food fre­

Hotz C, Brown KH (eds.). (2004). Supplement 2: quency assessment. Journal of the American Di­

International Zinc Nutrition Consultative Group etetic Association 98: 44-48.

(IZiNCG) Technical Document No. 1 . Assessment Macintosh DL, Williams PL, Hunter DJ, Sampson LA,

of the risk of zinc deficiency in populations and op­ Morris SC, Willet WC, Rimm EB. (1997). Evalua­

tions for its control. Food and Nutrition Bulletin 25: tion of a food frequency questionnaire- food com­

S96-S203. position approach for estimating dietary intake of

Hughes BA. (1986). Nutrition interviewing and coun­ inorganic arsenic and methylmercury. Cancer Epi­

selling in public health: the North Carolina experi­ demiology, Biomarkers & Prevention 6: 1043-1050.

ence. Topics in Clinical Nutrition 1 : 43-50. Margetts BM, Pietinen P, Ribolo E. (1997). EPIC

IOM (Institute of Medicine). (2000). Dietary Refer- European prospective investigation into cancer and
 Measuring food consumption of individuals 63

nutrition. Validation studies on dietary assessment from photographs. British Journal of Nutrition 72:

methods. International Journal of Epidemiology 26 649-663.

(Suppl. I): 1-189. Nelson M, Atkinson M, Meyer J. (1997). A Photo­

Marques M, Pinho 0, de Almeida MDV. (1996). Man­ graphic Atlas of Food Portion Sizes. Pub. No.

ual de quantificao de alimentos. Curso da Ciencias PB3006. Ministry of Agriculture, Fisheries and

de Nutricao de Universidade do Porto, Porto, Portu­ Food, London.

gal. Nes M, van Staveren WA, Zajkas G, Inelmen EM,

Marr JW, Heady JA. (1986). Within- and between-per­ Moreiras-Varela 0. (1991). Validity of the dietary

son variability in dietary surveys: number of days history method in elderly subjects. European Jour­

needed to classify individuals. Human Nutrition: nal of Clinical Nutrition 45: 97-104.

Applied Nutrition 40A: 347-364. NRC (National Research Council). (1981). The pro­

Martintchik AN, Baturin AK, Baeva VS, Peskova EV, posed system. In: Committee on Food Consumption

Larina TI, Zaburkhna TG. (1995). Alborn portisy Patterns, Food and Nutrition Board, National Re­

produktov i bljud [Album of portions of food and search Council (eds.) Assessing Changing Food

dishes]. Institute of Nutrition, Moscow. (Available Consumption Patterns. National Academy Press,

from the authors) Washington, DC., pp. 1 3 - 1 8 .

NRC (National Research Council). (1986). Nutrient

Maskarinec G, Novotny R, Tasaki K. (2000). Diet­
Adequacy: Assessment using Food Consumption
ary patterns are associated with body mass index
Surveys. National Academy Press, Washington,
in multi-ethnic women. Journal of Nutrition 130:
DC.
3068-3072.
Nusser SM, Carriquiry AL, Dodd KW, Fuller WA.
Masson LF, McNeill G, Tomany JO, Simpson JA,
(1996). A semi-parametric transformation approach
Peace HS, Wei L, Grubb DA, Bolton-Smith C.
to estimating usual daily intake distributions. Jour­
(2003). Statistical approaches for assessing the
nal of American Statistical Association 9 1 : 1440-
relative validity of a food-frequency questionnaire:
1449.
use of correlation coefficents and the kappa statistic.
O'Neill ME, Carroll Y, Corridan B, Olmedilla B, Gran­
Public Health Nutrition 6: 3 1 3 - 3 2 1 .
ado F, Blanco I, Van den Berg H, Rininger I, Rous­
McDowell M, Briefel RR, Warren RA Buzzard M,
sell AM, Chopra M, Southon S, Thurnham DI.
Seskanich D, Gardner S. (1990). The dietary data
(2001). A European carotenoid database to as­
collection system: an automated interview and cod­
sess carotenoid intakes and its use in a five-country
ing system for NHANES III. In: Proceedings of
comparative study. British Journal of Nutrition 85:
the 14th National Nutrient Databank Conference,
499-507.
CBORD Group Inc, Ithaca, New York, pp. 1 2 5 - 1 3 1 .
Palaniappan U, Cue RI, Payette H, Gray-Donald K.
McLennan WM, Podger A. (1998). National Nutri­
(2003). Implications of day-to-day variability on
tion Survey Nutrient Intakes and Physical Meas­
measurements of usual food and nutrient intake.
urements, Australia 1995. Australian Bureau of
Journal of Nutrition 1 3 3 : 232-235.
Statistics, Canberra.
Parnell W, Scragg R, Wilson N, Schaaf D, Fitzgerald
Merchant AT, Hu FB, Spiegelman D, Willett WC,
EL. (2003). NZ Food NZ Children: Key results
Rimm EB, Ascherio A. (2003). Dietary fiber periph­
of the 2002 National Children's Nutrition Survey.
eral arterial disease risk in men. Journal of Nutrition
Ministry of Health, Wellington, New Zealand.
1 3 3 : 3658-3663.
Piwoz EG, Creed de Kanashiro H, Lopez de Romana
MOH (Ministry of Health). (1997). Food Comes First:
G, Black RE, Brown KH. (1995). Potential for
Methodologies for the National Nutrition Survey of
misclassification of infants' usual feeding practices
New Zealand. Ministry of Health, Wellington, New
using 24-hour dietary assessment methods. Journal
Zealand. http://www.moh.govt.nz/
of Nutrition 125: 57-65.
NCHS (National Center for Health Statistics). (1994).
Robson PJ, Livingtone MB. (2000). An evaluation
Plan and operation of the Third National Health and
of food photographs as a tool for quantifying food
Nutrition Examination Survey, 1988-94. Vital and
and nutrient intakes. Public Health Nutrition 3:
Health Statistics 1 (32).
183-192.
Nelson M, Haraldsdottir J. (1998). Food photographs: Rambo M, Silveira D, Martins I, Cruz A. (1996). Mod­
practical guidelines II: Development and use of pho­ elos fotograficos para inqueritos alimentares. Centro
tographic atlases for assessing food portion size. de Estudos de Nutricao do Instituto Nacional de
Public Health Nutrition 1 : 231-237. Satide, Lisbon, Portugal.

Nelson M, Black AE, Morris JA, Cole TJ. (1989). Samuelson G. (1970). An epidemiological study of

Between- and within-subject variation in nutrient in­ child health and nutrition in a northern Swedish

take from infancy to old age: estimating the number county. 2: Methodological study of the recall tech­

of days required to rank dietary intakes with desired nique. Nutrition and Metabolism 12: 321-340.

precision. American Journal of Clinical Nutrition Sanjur D. (1982). Food consumption survey: issues
50: 1 5 5 -:- 1 6 7 . concerning the process of data collection. In: Social

Nelson M, Atkinson M, Darbyshire S. (1994). Food and Cultural Perspectives in Nutrition. Prentice­

photography I: The perception of food portion size Hall, Englewood Cliffs, NJ, pp. 169-194.
64 Principles of Nutritional Assessment

Sempos CT, Looker AC, Johnson CL, Woteki CE. men. Annals of Epidemiology 8: 378-383

(1991). The importance of within-person variabil­ USDA (U.S. Department of Agriculture). (1992). The

ity in estimating prevalence. In: Macdonald I Food Guide Pyramid. Home and Garden Bulletin

(ed.), Monitoring Dietary Intakes. Springer-Verlag, No. 252. Washington, DC.

Berlin, pp. 99-109. USDA (U.S. Department of Agriculture, Agricultural

Sempos CT, Liu K, Ernst ND. (1999). Food and Research Service). (1998). Food and Nutrient In­

nutrient exposures: what to consider when evalua­ takes by Individuals in the United States by Sex

ting epidemiologic evidence. American Journal of and Age, 1994-96. USDA Nationwide Food Sur­

Clinical Nutrition 69: 1330S-1338S. veys Report No. 96-2. Washington, DC.

Slattery ML, Benson J, Curtin K, Ma KN, Schaeffer D, van Kappel A, Amoye!J, Slimani N, Vozar B, Riboli E.

Potter JD. (2000). Carotenoids and colon can­ (1994). EPIC-SOFT Picture Book for Estimation

cer. American Journal of Clinical Nutrition 71: of Food Portion Sizes. International Agency for

575-582. Research on Cancer, Lyons, France. (Available from

Slimani N, Deharveng G, Charrondier RU, van Klap­ the authors.)

pel AL, Ocke MC, Welch A, Lagiou A, van Liere van Staveren WA, de Boer JO, Burema J. (1985). Va­

M, Agudo A, Pala V, Brandstetter B, Andrea C, lidity and reproducibility of dietary history method

Stripp C, van Staveren WA, Riboli E. (1999). Struc­ estimating the usual food intake during one month.

ture of the standardized computerized 24-h diet re­ American Journal of Clinical Nutrition 42: 554-559.

call interview used as reference method in the 22 van Staveren WA, West CE, Hoffmans MDAF, Bos

centers participating in the EPIC project. Com­ P, Kardinaal AFM, van Poppel GAFC, Schipper

puter Methods and Programs in Biomedicine 58: HJ-A, Hautvast JGAJ, Hayes RB. (1986). Compari­

251-266. son of contemporaneous and retrospective estimates

Slimani N, Ferrari P, Ocke M, Welch A, Boeing H, van of food consumption made by a dietary history

Liere M, Pala V, Amiano P, Lagiou A, Mattisson I, method. American Journal of Epidemiology 123:

Stripp C, Engeset D, Charrondiere R, Buzzard M, 884-893.

van Staveren WA, Riboli E. (2000). Standardization van Staveren WA, Burema J, Livingstone MBE, van

of the 24-hour recall calibration method used in the den Broek T, Kaaks R. (1996). Evaluation of the

European Prospective Investigation into Cancer and dietary history method used in the SENECA study.

Nutrition (EPIC): general concepts and preliminary European Journal of Clinical Nutrition 50(Suppl. 2):

results. European Journal of Clinical Nutrition 54: S47-S55.

900-917. Weiner JS, Lourie JA. (1969). Human Biology: A

Soho EJ, Rock CL, Neuhauser ML, Maciel TL, Neu­ Guide to Field Methods. International Biological

mark-Sztainer D. (2000). Caretaker-child interac­ Programme. Handbook No. 9. Blackwell Scientific

tion during children's 24-hour dietary recalls: who Publications, Oxford.

contributes what to the recall record? Journal of the WHO (World Health Organization). (1998). Comple­

American Dietetic Association 100: 428-433. mentary Feeding of Young Children in Developing

Stockley L, Chapman RI, Holley ML, Jones FA, Pres­ Countries: A Review of Current Scientific Knowl­

cott EHA, Broadhurst AJ. (1986a). Description of edge. World Health Organization, Geneva.

a food recording electronic device for use in dietary Willet WC. (1990). Nutritional Epidemiology. Oxford

surveys. Human Nutrition: Applied Nutrition 40: University Press, Oxford.

13-18. Willett WC. (1994). Future directions in the develop­

Stockley L, Hurren CA, Chapman RI, Broadhurst AJ, ment of food frequency questionnaires. American

Jones FA. (1986b). Energy, protein, and fat intake Journal of Clinical Nutrition 59: 171S-174S.

estimated using a food recording electronic device Willett WC, Sampson L, Stampfer MJ, Rosner B,

compared with a weighed diary. Human Nutrition: Bain C, Witschi J, Hennekens CH, Speizer FE.

Applied Nutrition 40: 19-23. ( 1 9 8 5 ) . Reproducibility and validity of a semiquan­

Szczyglowa H, Szczepanska A, Ners A, Nowicka L. titative food frequency questionnaire. American

(1991). Album porcji produktow i potraw. Instytut Journal of Epidemiology 122: 5 1 --6 5 .

Zywnosci i Zywienia, Warsaw, Poland. (Available Yanek LR, Moy TF, Raqueno N, Becker DM. (2000).

from the authors) Comparison of the effectiveness of a telephone 24-h

Tran KM, Johnson RK, Soultanakis RP, Matthews DE. dietary recall method vs an in-person method among

(2000). In-person vs telephone administered mult­ urban African-American women. Journal of the

iple-pass 24-hour recalls in women: validation with American Dietetic Association 100: 1172-1177.

doubly labeled water. Journal of the American Di­ Young CM. (1981). Dietary methodology. In: Com­

etetics Association 100: 777-783. mittee on Food Consumption Patterns, Food and

Tsugane S, Fahey MT, Kobayashi M, Sasaki S, Tsub­ Nutrition Board (ed.). Assessing Changing Food

ono Y, Akabane M, Gey F. (1998). Four food-fre­ Consumption Patterns. National Research Coun­

quency categories of fruit intake as a predictor of cil. National Academy Press, Washington, DC.,

plasma ascorbic acid level in middle-aged Japanese pp. 8 9 - 1 1 8 .