Validity and reproducibility of a dietary history

method estimating the usual food intake

during one
Wija A van Staveren, MSc, Janna 0 de Boer, MSc, and Jan Burema, MSc

ABSTRACT The validity and reproducibility of a dietary history method with a time of
reference of one month was assessed with 44 young adults (aged 19-32 yr). The concurrent validity
of the method was assessed by means of the 24-h urine nitrogen excretion. The mean difference

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
between N-intake and N-excretion (24-h urine N-excretion plus 2 g for extra renal nitrogen losses)
was 0.0 g with 95%-confidence limits of ±1.1 g. These limits for the mean difference between
excretion and intake indicate a valid assessment of the protein intake of this group. The
reproducibility was evaluated in the same group thmugh a test-retest design. The intraclass
correlation coefficients were high over a weighted average of weekdays and for an average workday
with regard to the intakes of energy and selected nutrients. For the Saturday and Sunday intakes,
the intraclass correlation coefficients were lower for the energy intakes and most of the nutrients
(except alcohol), indicating a poorer reproducibility for weekend assessments. Am J C/in Nuir
1985;42:554-559.

KEY WORDS Dietary methods, validity, reproducibility, nitrogen excretion

Introduction chemical criteria. In previous reports, the

relative validity (3-7) as well as the concurrent
In the context of a longitudinal study on validity (8- 10) of the dietary history method
the etiology of obesity in young adults a has been evaluated. In a review on the validity
methodological study was carried out on the ofdietary assessment methods Block (1) con-
validity and reproducibility of a dietary his- cludes that it is difficult to gather from the
tory method. This method is one of the reports published, whether a dietary history
preferred methods available to estimate the is a valid method in the sense that it reflects
usual dietary intakes of individuals. The the nutrient intake accurately. In Block’s
quality of methods estimating food con- report no validation of nitrogen intake against
sumption can be expressed by their validity 24-h urine nitrogen excretion (10-12) is in-
and reproducibility. Validity tries to give an cluded. As this method is considered (13)
answer to the question whether the method one of the best biochemical criteria against
actually measures what the investigator in- which to validate a dietary intake method,
tends to measure. The central methodological we decided to validate the dietary history
problem of food consumption studies is that method on an aggregate level against this
there is no golden standard for directly as- biochemical marker.
sessing the validity (1, 2). This is especially a
problem for the dietary history method, since I From the Department of Human Nutrition, Agri-
this method estimates the habitual food intake cultural University, De Dreijen 12, 6703 BC Wageningen,
over a longer period of time up to one year. The Netherlands.
2 Address reprint requests to: Wija A van Staveren,
Therefore investigators usually determine ei-
Department of Human Nutrition, De Dreijen 12, 6703
ther the relative validity, ie comparison with BC Wageningen, The Netherlands.
the weighed record method, or the concurrent Received November 1, 1984.
validity, ie comparison with clinical or bio- Accepted for publication February 28, 1985.

554 The American Journal ofClinical Nutrition 42: SEPTEMBER 1985, pp 554-559. Printed in USA
C 1985 American Society for Ginical Nutrition
 QUALITY OF DIETARY HISTORY METHOD 555

Hankin et al (14) have recently reported as butter, contents ofcups, glasses, and dishes for porridge
or pudding; and sugar and milk in coffee and tea. Other
on the reproducibility of a dietary history
portions of food not directly available at home were
questionnaire. An assessment is called repro- estimated.
ducible if it gives the same results in the Food consumption was converted into nutrient intake
same situation. In Hankin’s study the dietary using a computer program based on a nutrient file
history covered a usual week of the subjects; compiled from the Dutch food composition table (19).
Twenty-four-hour urine was collected twice in 2#{189}-I
in order to test the reproducibility of the containers with 5 ml of a 10% thymol solution in
method, another usual week was selected isopropanol to prevent oxidation. For the whole group,
within three months after the first interview. collecting was evenly distributed over all days of the
They suggested that in studying the food week in order to obtain the mean urine nitrogen excretion
of an average weekday. The total nitrogen content of
consumption ofCaucasians for most nutrients
urine was determined by the Kjeldahl method. To
a longer period was required to estimate the express nitrogen excretion in terms of protein or nitrogen
usual intake. Our study concerns the concur- intake the formula of Isaksson (10) was applied:
rent validity and reproducibility of a dietary

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
Protein intake in 24 h
history method in which the usual food = N1 fl 24 h

consumption ofa specific month was recalled.

= j 24 h urine + 2 g.

Methods Two grams of nitrogen were added to the 24-h nitrogen

excretion in urine to compensate for losses via feces and
Selection of samples skin. As this amount depends on the quantity and
quality ofprotein consumed and on the nutritional status
Young adults aged 19-32 yr were contacted by tele-
and physical activity, it was tested first whether extra
phone in a town (more than 30,000 inhabitants) near renal nitrogen losses for young Dutch adults also were
Wageningen. The methodological character of this study
about 2 g. For this purpose data were available from a
does not require a representative sample. Based on data previous experimental study in this group. The protein
found in the literature, about 50 young adults would be content of their diet was in accordance with the average
sufficient to obtain an acceptable imprecision in the Dutch diet. Food consumption had been assessed by a
estimated mean difference between nitrogen excretion 4-day weighing method, and conversion into nitrogen
and nitrogen intake (10, 1 1), ie the half-width of the intake had been carried out by analyses of nitrogen in
95%-confidence interval of the mean difference is 10%
foods. Furthermore nitrogen excretion had been deter-
of the level of nitrogen excretion or less.
mined in 24-h urine as well as in feces (20). Table 1
shows that the amount of 2 g/day is applicable to Dutch
Collection of data
circumstances. Both the difference between daily nitrogen
A dietary history method based on Burke’s method intake and urine nitrogen excretion, and the sum of
(15) was carried out in March 1980 (first interview) and fecal and dermal nitrogen losses are 2 g.
in April 1980 (second interview). The participants were
visited at home by dietitians and postgraduate students Statistical methods
in human nutrition trained for this purpose. The training In order to assess the concurrent validity, the mean
included interview training, estimating portion sizes, and of the differences between individual pairs of measure-
coding. To assure comparability in the data, all inter- ments and 95% confidence limits were calculated. The
viewers coded several times the same interview. Each mean difference reflects the systematic departure between
subject was visited by the same interviewer for the test the estimation of nitrogen intake as assessed by the
interview as well as for the retest interview. In both dietary history and as assessed by the nitrogen excretion
interviews the participants were questioned about their in urine.
usual food consumption of the preceding month. The To evaluate the reproducibility, the standard deviation
questionnaire was open-ended and consisted of three of individual differences and the intraclass correlation
parts: one for the usual food consumption on a workday, coefficients between the first and the second estimation,
one for the usual consumption on a Saturday, and one as advocated by Lee (22), were calculated.
for the usual consumption on a Sunday. Workdays, This type of project is usually not referred to the
Saturdays, and Sundays were recorded separately because Ethical Committee. However, we have examined the
earlier studies have shown systematical weekend effects project and are of the opinion that there
are no risks or
on nutrient intake in comparable groups (16-18). From ethical problems associated with carrying it out. Thus
these data a weighted average ofweekdays was calculated the Ethical Committee approves the project.
according to the formula
(5 X workday + 1 X Saturday + 1 X Sunday) Results
7
Population
Portions of food frequently consumed were checked for
weight. A Soehnle 8600 balance was used by the dietitian Out of the 47 persons who agreed to
for weighing, eg slices of bread; spreads on bread such participate, 44 (22 men and 22 women) were
556 VAN SIAVEREN El AL

TABLE 1
Nitrogen intake and nitrogen excretion via urine, feces, and skin during 24 h in young adults (aged 18-28 yr).
Data from Van Raaij (20)
Mean SE Method N

S g

N-intake (N1) 13.9 0.5 4-Day weighed record Nitrogen 64

analyses in foods consumed
Urine N-excretion (NE) 1 1.9 0.4 Kjeldahl 64
N1 - NE 2.0 0.45 64
Fecal N-excretion 1.6 0. 1 Kjeldahl 68
Dermal N-excretion 0.4 0.01 Estimated as 0.005 g/kg body 69
weight (21)
Fecal and dermal N-excretion 2.0 0. 1 68

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
able to complete the whole task: two dietary 1 1 for men and women
. respectively. The
history interviews and collecting 24-h urine 95% confidence limits ofthis mean difference
twice. Three women completed the dietary are rather wide, especially for the men. In
history interviews, but were not able to collect fact, the number of subjects in the subcate-
the urine. gories men and women is too small to draw
firm conclusions.
Concurrent validity
Table 2 shows the results of a comparison Reproducibility
between the daily nitrogen intake for a Table 3 shows the results of both dietary
weighted average of weekdays as estimated history interviews over a weighted average of
by the first dietary history interview and as weekdays, a workday, a Saturday, and a
estimated from the 24-h urine nitrogen ex- Sunday. The assessed nutrient intake resulting
cretion. The mean difference was 0.0 g with from the second interview does not differ
95% confidence limits of ±1. 1 g. This result significantly from that of the first interview
seems to support the hypothesis that there is on an aggregate level. However, whereas va-
no difference between excretion and intake lidity concerns systematic deviations between
for the aggregated group values. On the other two methods, reproducibility concerns ran-
hand, it should be borne in mind that there dom fluctuations in repeated measurements
is considerable variation in individual differ- by the same method on various occasions.
ences between estimated intake and excretion. The intraclass correlation coefficient quanti-
The sex-specific mean difference between fies the extent of overall agreement on an
intake and excretion was 0.2 g and -0.2 g individual level between the two interviews.
with confidence limits of 1.6-2.0 and -1.5-
- The high correlation coefficients indicate

TABLE 2
Comparison of N-intake in g/day as determined by the dietary history method (N1) and as estimated from urine
N-excretion/24 h including 2 g for extra renal N losses (NE) in young adults

Total oup Men Women

(n - 44) (n - 22) (n = 22)

g g g

N-intake (N1) 13.3 (0.51) 14.6 (0.77) 1 1.9 (0.55)

N-intake (NE) estimated from:

urine N-excretion 1 1.3 (0.46) 12.4 (0.60) 10. 1 (0.63)
and extra renal losses 2.0 2.0 2.0

N1 - NE 0.0 (0.50) 0.2 (0.94) -0.2 (0.57)

Confidence limits N1 - NE -1.1 and 11 -1.6 and 2.0 -1.5 and 1.1
. Mean; standard error between parentheses.
 QUALITY OF DIETARY HISTORY METHOD 557

TABLE 3
Reproducibility of a dietary history interview on an aggregate level (mean) and on an individual level (SD of
difference) over a weighted average of weekdays, a workday, a Saturday, and a Sunday (n = 47)

Saturated Unoleic Dietary

Energy Protein Fat fat acid Carbohydrate fiber AlcohOl

kcal g g g g g g g
Weighted weekday
First interview (mean) 2352 82 101 45 12 256 24 12
Second interview (mean) 2327 79 103 46 14 248 23 12
SD of differences 434 14 26 9 7 46 6 8
Intraclass correlation 0.86 0.80 0.81 0.89 0.67 0.87 0.75 0.91
Workday
First interview (mean) 2321 84 101 45 12 255 26 8
Second interview (mean) 2247 78 100 46 13 243 24 8
SD of differences 442 16 27 8 8 51 7 8
Intraclass correlation 0.87 0.79 0.85 0.93 0.64 0.86 0.72 0.87

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
Saturday
First interview (mean) 2573 82 107 46 14 271 20 28
Second interview (mean) 2641 82 1 12 47 16 280 20 27
SD ofdifferences 1026 35 54 25 9 1 12 10 21
Intraclass correlation 0.54 0.70 0.52 0.46 0.64 0.47 0.50 0.85
Sunday
Firstinterview(mean) 2286 77 96 43 11 248 20 17
Second interview (mean) 2410 79 1 10 47 15 245 20 17
SDofdifferences 608 28 44 15 11 62 87 11
Intraclass correlation 0.68 0.57 0.53 0.62 0.41 0.76 0.59 0.90

agreement between the first and the second tein. It might be concluded that it would also
dietary history interview over a weighted be an unbiased estimation for energy, fat,
average weekday. The lowest correlation coef- and carbohydrate from these products. How-
ficient was found for linoleic acid (0.67) and ever, various foods do not contain any pro-
the highest for alcohol (0.91). High correlation tein, for example sugar, oils and butter, al-
coefficients were also found for energy and coholic drinks and soft drinks. This makes it
nutrient intakes during an average workday. unwarranted to extrapolate the concurrent
The sex-specific results of the intraclass cor- validity simply towards other nutrients. As
relation coefficient were very similar to those to energy, Isaksson (10) suggests that the
for men and women pooled. contribution from protein to the total energy
The intraclass correlation coefficients for may be a fixed proportion in groups of
Saturday and Sunday are much lower. Very subjects. In comparable groups of Dutch
low correlation coefficients were found for adults the percentage of energy from protein
the Saturday intakes of saturated fat (0.46) varies from 12-15 (17, 18, 23), which is
and total carbohydrate (0.47); the agreement indeed a small fluctuation.
between the first and the second interview In the literature the validity of the dietary
on Saturday as regards alcohol intake seems history method is primarily determined by
to be much better (0.85), but this is partly comparing this method with a seven-day
due to nonalcohol users. The correlation food record or a 24-h recall method. The
coefficients for the two interviews on Sunday results of these studies show that the dietary
are also low, especially as regards linoleic history method, in general, produces higher
acid (0.41). estimates than the two other methods (4-7).
Our study does not indicate an overestimation
Discussion ofthe protein intake. However, in comparing
The positive results found on the concur- our results with those of the other studies it
rent validity of the dietary history method should be realized that in the former the
correspond with the findings of Isaksson (10) estimated usual consumption refers to a fixed
and indicate that a valid assessment has been rather short period of time as opposed to a
made on the intake of foods containing pro- longer period (6 mo to 1 yr) in the latter.
558 VAN STAVEREN El AL

C.-
c
I#{149}L -o
n ‘ : Our study demonstrates a good reproduc-
0 . r- o r- n #{176} ibility of the dietary history method (Table
. § 0 0 0 : 3). This confirms earlier findings of studies
l (. evaluating the same method over longer pe-
riods of time (24-26).
r- 0 Table 4 compares the reproducibility in
U a’rr-.
E 0 the daily intake of fat, saturated fat, and
. ; animal protein by women in the present
et -

E i E study on the one hand and by Hankin et al

I ! ‘ i ‘. o (14) with a dietary history method questioning
li T© a usual week on the other hand. Hankin et
nE
0 et al have tested the reproducibility of a dietary
E , history questionnaire in a case control study
g.a’ 0
r-o
o#{243}
r r .
a
#{149}E on breast cancer with Caucasian and Japanese

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
.- t . , r’ - - a women in Hawaii. Their findings seem to
a , c
R +1+1+1
,r n indicate that the dietary history method is
v
c #{176}#{176}. . , not reproducible for the Caucasian controls.
v, 2
a a a The higher correlation coefficients in our
.0 > _ - a study reflect a greater consistency in the
: 0 t #{149}
0
-‘
r’”1

‘
-
‘
-
+1
8’...
0
o
intakes measured
In comparing
over 1 mo than over 1 wk.
our results with those of
t E . c a . Hankin’s study, it should be realized that, in
. , ., addition to differences in the time of refer-
a U a a a ence, the time lag (1 mo vs 3 mo) and the
. a n . . a technique of measuring the amounts of foods
2 i#{149}8 - a
0.0 e’- a>n consumed on differed.
. . § 0 ci ci ‘. et
u; subjects 43 items Hankin
or food questioned
groups, the
which
. ‘ 2 . I contributed about 85% of the total fat and
. - E animal protein intakes. Also, she estimated
(o t) I I I aa 8 E a the amounts of foods consumed by means
E I8’ C I ‘ ; of pictures of standard-size servings.
‘- . 0 In summary the results of our study mdi-
I
. ;
. I od
.
a I
cate that
reference
a dietary history
of one month
with a time of
may give a valid
ta to a
‘, fi , a estimation of the nitrogen and thus protein
, E N intake, and reproducible estimations for in-
t , 0 r- ‘ ; . takes of energy and macronutrients in adult
.otI, (N
. ri #{149} +1 +1 .,.‘ a a Caucasians.
#{149}-.a “ne ‘.cn
0 a
: ‘
U
#{149} ! ?
a
E l The authors
for their invaluable
thank those who
cooperation.
took
They
part in the study
also thank Mrs
U C... JM van Duijvendijk and the dietitians who conducted
ca
, ‘O - #{176}#{176}
I1 2 the interviews and Mrs GJC van Oosten-van der Goes
a
, et C,) r - . 2 and Mr BA Scholte who assisted in the computer analyses

E ‘ +1 +1 +1 #{149} of the data.

cJa
- . Eo-.o
#{149}g
,,1:r%iod
ori”s aa 0

t, E References
e‘a - .-

1. Block 0. A review ofvalidations ofdietary assessment

,-
o
( : a E “ g methods. Am J Epidemiol 1982;l 15:492-505.
a 2 ‘#{149} 2. Marr JW. Individual dietary surveys: purposes and
>. >. 0 methods. World Rev Nutr Diet 197 l;l3:105-64.
I ‘ . .. , 3. Huenemann RL, Turner D. Methods of dietary
mE
< 0 0 a #{149} . investigation. J Am Diet Assoc 1942;18:562-8.
f-, U n ‘ ci < - 4. Young CM, Chalmers FW, Church HN, et al. A
 QUALITY OF DIETARY HISTORY METHOD 559

comparison of dietary study methods I. Dietary 16. Beaton GH, Milner i, Corey P, et al. Sources of
history vs seven day record. J Am Diet Assoc variance in 24-hour dietary recall data: implications
l952;28: 124-8. for nutrition design and interpretation. Am J Gin
5. Den Hartog C, van Schaik IFSM, Dalderup LM, et Nutr 1979;32:2546-59.
al. The diet of volunteers participating in a longterm 17. Baecke JAH, Staveren WA van, Burema J. Food
epidemiological field survey on coronary heart disease consumption, habitual physical activity and body
at Zutphen, The Netherlands. Voeding 1965;26: fatness in young Dutch adults. Am J Gin Nutr
184-208. l983;37:278-86.
6. Jam M, Howe GR, Johnson KC, et al. Evaluation 18. Van Staveren WA, Hautvast JGAJ, Katan MB, et
of a diet history questionnaire for epidemiological al. Dietary fiber consumption in an adult Dutch
studies. Am J Epidemiol l980;l 11:212-21. population. A methodological study using a seven-
7. Van Leeuwen FE, de Vet HCW, Hayes RB, et al. day record. J Am Diet Assoc l982;80:324-30.
An assessment ofthe relative validity of retrospective 19. Hautvast JGAJ. Ontwikkeling van een systeem om
interviewing for measuring dietary intake. Am J gegevens van voedingsenqu#{234}tes met behulp van de
Epidemiol l983;l 18:752-8. computer te verwerken. Voeding l975;36:356-6l.
8. Burke BS, Stuart HC. A method of diet analysis. J 20. Van Raaij JMA. Influence ofhuman diets containing
Pediatr 1938;l2:493-503.

Downloaded from www.ajcn.org at Univ of Otago (AJCN) Sci Lib on March 9, 2008
casein, soy protein isolate, and soy protein concen-
9. Reed RB, Burke BS. Collection and analysis of trate on serum cholesterol and lipoproteins in hu-
dietary intake data. Am J Public Health 1954;44: mans, rabbits and rats. Thesis Wagcningen, l9g2.
1015-26. 21. Energy and protein requirements. Report of a joint
10. Isaksson B. Urinary nitrogen output as a validity test FAO/WHO ad hoc expert committee. World Health
in dietary surveys. Am J Clin Nutr l980;33:4-5. Organization. Technical report series, no 522, Ge-
1 1. Huse DM, Nelson PA, Briones ER, et al. Urinary neva, 1973.
nitrogen excretion as objective measure of dietary 22. Lee J. Alternate approaches for quantifying aggregate
intake. Am J Gin Nutr 1974;27:77 1.
and individual agreements between two methods for
12. Warnold IG, Carlgen 0, Krotkiewski M. Energy
assessing dietary intakes. Am J Gin Nutr 1980;33:
expenditure and body composition during weight 956-8.
reduction in hyperplastic obese women. Am J Gin
23. Niessen Ch, Brussaard JH, Katan MB. Serumlipiden
Nutr 1978;3 1:750-63.
en samenstelling van de voeding van 371 Wageningse
13. Bingham S. Recent developments in dietary meth-
studenten. Voeding 1983;44:350-5.
odology. In: Hautvast JGAJ, Kiaver W, eds. The
diet factor in epidemiological research. Euro Nut 24. Trulson MF, McCann MB. Comparison of dietary
Report 1. Wageningen, 1982:106-22. survey methods. i Am Diet Assoc 1959;35:672-6.
14. Hankin JH, Nomura AMJ, Lee J, et al. Reproduc- 25. Dawber IR, Pearson G, Anderson A, et al. Dietary
ibility of a diet history questionnaire in a case- assessment in the epidemiologic study of coronary
control study of breast cancer. Am J Gin Nutr heart disease: The Framingham Study. II. Reliability
1983;37:98 1-5. of measurement. Am J Gin Nutr l962;l 1:226-34.
15. Burke BS. The dietary history as a tool in research. 26. Reshef A, Epstein LM. Reliability of a dietary
J Am Diet Assoc 1947;23:1041-6. questionnaire. Am J Gin Nutr 1972;25:9l-5.