ORIGINAL COMMUNICATION
Validation of a semi-quantitative adolescent food
frequency questionnaire applied at a public school
o Paulo, Brazil
in Sa
B Slater1*, ST Philippi1, RM Fisberg1 and MRDO Latorre2
1
Department of Nutrition, School of Public Health, University of Sao Paulo, Sao Paulo, Brazil; and 2Department of Epidemiology,
o Paulo, Sa
o Paulo, Brazil
School of Public Health, University of Sa
Objective: To develop a food frequency questionnaire for adolescents (AFFQ) and demonstrate its relative validity.
Design: The final version of the AFFQ was composed of 76 food items previously identified according to their contribution in
nutrients and overall importance within the eating habits of this population group. The validation study, which was undertaken
during a 6 month period (June to November 1999), was administered to a sample of 79 who answered at least three 24 h dietary
recalls (R24 h) applied at intervals of 45 days and one AFFQ at the end of the study. Applying the paired t-test and calculating
Pearson correlation coefficients on nutrient data, differences in the mean of nutrients were obtained. Correlation coefficients
between the mean energy-adjusted nutrients computed by the two methods were calculated, and correction was made for
within-person variability. Agreement was evaluated by distribution of the adolescents according to quartiles of consumption.
Locus: A public school within the metropolitan region of Sao Paulo city.
Results: A high variability in the dietary intake of adolescents was observed, with high rates of variability for cholesterol, retinal
and vitamin C. The Pearson correlation coefficients, after being adjusted and corrected for variability, ranged from 0.10 to 0.72
among females and from 0.16 and 0.91 among males. The mean correlation coefficient for the entire group was 0.52.
Conclusions: These results indicate that the AFFQ provides a potentially reliable scale for categorizing individuals by level of past
intake of most nutrients, excluding retinol and iron.
European Journal of Clinical Nutrition (2003) 57, 629 635. doi:10.1038=sj.ejcn.1601588
Keywords: nutritional assessment; food frequency questionnaire; adolescents
Introduction
In the last few years, evidence has emerged concerning the
possible relations between diets of children and adolescents
and chronic diseases in adult life (Ludwig et al, 2001).
Adolescent eating habits are highly influenced by family
patterns, habits and the peer group, and by the growing
o da Faculdade de
*Correspondence: B Slater, Departamento de Nutrica
Saude Publica=USP, Av. Dr Arnaldo, 715-CEP: 01246-904, Cerqueira Cesar,
o Paulo, SP, Brazil.
Sa
E-mail: betzy_slater@hotmail.com
Guarantor: B. Slater
Contributors: BS responsible for research design, conduction of the
study and for the writing the document. STP was the advisor of
the study. RMF collaborated in the interpretation and discussion
of the studys results. MRDOLT collaborated in the statistical
analyses and interpretation of the results.
Received 27 February 2002; revised 1 July 2002;
accepted 2 July 2002
630
second, because they are relatively inexpensive when selfadministered (Jime nez & Martn-Moreno, 1995; Ocke et al,
1997). The literature indicates that studies which utilize
FFQs in order to evaluate childrens and adolescents food
intake are scarce. There is, therefore, a need to develop new
instruments which can be easily administered to large
groups and which are simple so they can be self-administered
and which will be able to evaluate habitual diet among
adolescents.
The main objective of this study was to develop a food
frequency questionnaire for adolescents, AFFQ , and evaluate
the relative validity of the estimates of energy, macronutrients, retinal, cholesterol, dietary fiber, vitamin A, calcium
and iron consumption.
Methods
Development of the AFFQ structure
Data concerning food habits registered in a study on nutritional evaluation conducted by Nuzzo (1998) was utilized in
the elaboration of the AFFQ list of foods. The information
regarding diet in the former study corresponds to the application of 2 days of the dietary record method in 200 adolescents of both sexes. The foods were grouped together
according to their physical characteristics and nutritive
value in 140 food items. Based upon this grouping, the
foods that made a greater contribution, both energetically
and in terms of their nutrients, were identified (Block et al,
1985a,b).
Having defined a food list, the AFFQ was tested in a pilot
study, in which 76 food items were included. This questionnaire evaluated the quantity of foods and nutrients consumed during the preceding 6 month period. The frequency
of consumption was evaluated by means of simple questions
Figure 1
Validation study
Description of the population and of the design of the
validation study. The validation study was conducted
between June and December 1999, the period during
which data were collected. The population study consisted
of 106 adolescent volunteers of both sexes, aged 14 18 y and
11 months, registered in the first year of a junior high school
located in the western region of Sao Paulo city. The relative
validity of the study was evaluated by comparing the standard measure with the data collected by the AFFQ. During
this period the students completed three 24 h dietary recalls
administered on nonconsecutive days including one day
during a weekend, with 45 day intervals between them and
one AFFQ applied at the end of the period.
Data processing
Once the data was collected, the food intake registered by
the 24 h dietary recall and the AFFQ were converted into
energy and nutrients by the Virtual Nutri program (Philippi
et al, 1996).
631
Statistical analysis
The differences in mean consumption of energy and nutrients obtained by the AFFQ and by the 24 h dietary recalls
were analyzed using paired t-test. The correlation coefficient (r), which compared the energy and nutrient values
obtained by both methods, was calculated before and after
adjustment for total energy intake (Willett & Stampfer,
1986).
Within-person variance (ie day-to-day variation in diet)
estimated in the recalls could attenuate correlations between
the AFFQ and the 24 h recalls, due to the relatively low
number of replicates. In order to obtain a measure of validity,
the correlation coefficient was multiplied by factor
(1 (s2w=s2b)n)0.5) where n represents the number of replicate
measurements, s2w is the within-person variation and s2b is
the between person variation. The result thus obtained is the
de-attenuated correlation coefficient (Willet, 1998). To evaluate the agreement of classification according to the levels
of nutrient intake between the AFFQ and the 24 h-dietary
recall, we compared the quartile classification obtained by
Results
Within the study period, from June to November 1999, 79 of
the 106 students completed three 24 h dietary recalls and the
AFFQ. Among the students who participated in this study,
50.6% were female and 49.4% were males. The average age
was 15.8 y (s.d. 1.09).
No statistically significant differences were found between
the mean values estimated by the three 24 h recalls and the
AFFQ for energy intake, total fat, vitamin C and calcium. On
the other hand, the AFFQ overestimated the values of carbohydrates and fiber and underestimated those of proteins,
cholesterol, unsaturated fat and iron (Table 1). Upon
Table 1 Mean, standard deviation and median daily intake assessed by three 24 h recall and AFFQa among 79 adolescents
attending public high school in Sao Paulo, Brazil
24 h recall
Nutrients
Energy (kcal)
Female
Males
Total proten (g)
Female
Males
Total carbohydrates (g)
Female
Males
Total fat (g)
Female
Males
Polyunsaturated fat (g)
Female
Males
Dietary fiber (g)
Female
Males
Cholesterol (mg)
Female
Males
Retinal (mg ER)
Female
Males
Vitamin C (mg)
Female
Males
Calcium (mg)
Female
Males
Iron (mg)
Female
Males
AFFQ
Median
Mean
s.d.
Mean
s.d.
24 h recall
AFFQ
Pb
2004.87
1731.94
2284.79
79.52
66.12
93.27
242.18
210.89
274.29
78.63
67.87
89.65
31.63
26.08
37.33
10.56
7.87
13.31
239.70
204.60
275.70
745.64
608.28
886.52
69.83
64.28
75.52
584.88
481.30
691.12
11.73
9.09
14.45
728.76
572.58
766.87
34.83
28.10
35.80
90.23
78.22
90.73
39.02
30.60
43.53
20.79
16.41
23.21
6.32
4.82
6.50
153.50
138.75
160.02
831.98
830.23
813.39
86.59
81.43
91.58
384.17
301.13
429.83
5.62
4.18
5.62
2023.59
1763.71
2290.11
69.14
62.27
76.18
265.40
226.82
304.97
76.60
68.52
84.90
25.74
22.93
28.62
11.70
10.31
13.13
204.56
182.98
226.70
614.94
596.55
633.80
79.91
69.84
90.24
561.40
507.00
617.20
8.43
8.44
10.15
563.36
451.10
546.46
20.04
17.95
19.83
75.96
61.96
68.76
24.83
19.96
26.78
9.54
7.82
10.35
4.40
4.08
4.30
83.38
89.30
71.35
243.50
272.89
211.06
41.62
33.43
46.83
223.49
232.67
201.63
2.49
2.49
2.79
1927.28
1686.84
2125.93
77.29
62.48
86.61
222.78
206.77
253.05
74.72
62.91
82.25
28.49
23.92
36.21
9.95
7.11
12.26
222.74
182.11
249.78
612.66
547.53
695.74
55.33
44.23
57.69
577.55
421.44
622.18
11.04
8.84
13.45
1920.47
1681.79
2270.48
65.34
63.27
74.39
264.94
218.90
318.34
72.28
66.68
79.91
24.87
22.32
28.00
10.90
9.86
12.47
198.16
175.34
221.99
581.04
511.91
611.98
74.35
61.82
81.44
569.07
471.23
597.04
8.63
7.89
9.39
0.560
0.407
0.919
0.001
0.213
0.001
0.001
0.029
0.001
0.370
0.826
0.167
0.001
0.061
0.001
0.012
0.001
0.787
0.003
0.197
0.003
0.048
0.895
0.010
0.103
0.526
0.096
0.386
0.387
0.105
0.001
0.209
0.001
632
examining the 11 dietetic variables, it was found that only
two of them presented different mean intakes among the
female participants of this study, being the overestimation of
dietetic fiber in approximately 30% (P > 0.001) the most
relevant difference found. Among the male participants,
five nutrients presented different mean intakes, these were:
protein, carbohydrate, dietetic fiber, unsaturated fat, cholesterol and iron (P < 0.05).
The ratio of the components of within-person and
between-person variability which ranged from 1 to 2 for
energy, protein, carbohydrates, dietetic fiber and iron for
all individuals and which was greater than 2 for the rest of
the nutrients studied, being that patterns of within-person
variability were more pronounced among the adolescent
boys than among the adolescent girls. However, retinol was
an exception for the pattern of within-person variability
being, in this case, more pronounced among the girls. The
ratios for cholesterol and vitamin C among the adolescent
boys caught our attention, as they were extremely high
(Table 2).
Table 2 Variance components of the nutrient intakes estimated by three 24 h recalls among 79
adolescents attending public high school in Sao Paulo, Brazil
24 h dietary recalls
Nutrients
Energy (kcal)
Female
Males
Total protein (g)
Female
Males
Total carbohydrates (g)
Female
Males
Total fat (g)
Female
Males
Polyunsaturated fat (g)
Female
Males
Dietary fiber (g)
Female
Males
Cholesterol (mg)
Female
Males
Retinal (mg ER)
Female
Males
Vitamin C (mg)
Female
Males
Calcium (mg)
Female
Males
Iron (mg)
Female
Males
a
l Sw=Sb.
Variance ratios
1.4
1.3
2.9
1.7
2.1
3.8
1.2
1.4
1.7
2.5
2.5
4.1
2.5
2.9
3.6
1.6
2.6
3.4
5.9
7.6
10.4
4.7
26.4
2.5
8.2
5.1
14.35
2.7
3.2
3.8
1.4
2.4
3.6
Within-person variance
Between-person variance
311 761.95
189 326.37
437 336.91
772.93
533.38
1018.64
4383.26
3542.82
5245.24
1090.01
673.40
1517.30
310.55
200.58
423.33
24.74
16.91
32.77
20 186.06
17 056.30
23 396.08
571 837.34
664 563.15
476 333.94
6687.93
5554.74
7850.17
107 741.82
69 433.99
147 031.90
18.51
12.37
24.80
221 210.21
140 889.1
153 400.34
444.05
260.64
267.79
3790.12
2619.84
3040.60
426.41
267.95
383.97
122.78
69.91
117.36
15.33
6.52
9.60
3411.42
2232.16
225 013
121 374.90
25 135.21
88 109.09
816.73
1094.60
546.97
40 183.66
21 610.00
38 379.83
13.17
5.21
6.95
633
Table 3 Pearson correlation coefficients between daily intake of energy and nutrients assessed by three 24 h recalls and AFFQa among 79
adolescents attending public high school in Sao Paulo, Brazil
Correlation coefficients
Non-adjusted b r
Nutrients
De-attenuated c r
Energy-adjusted r
Total
Female
Male
Total
Female
Male
Total
Female
Male
Energy (kcal)
Total protein (g)
Total carbohydrates (g)
Total fat (g)
Polyunsaturated fat (g)
Dietary fiber (g)
Cholesterol (mg)
Retinol (mg ER)
Vitamin C (mg)
Calcium (mg)
Iron (mg)
0.87**
0.63**
0.77**
0.72**
0.57**
0.59**
0.44**
0.28*
0.42**
0.61**
0.46**
0.86**
0.47**
0.70**
0.68**
0.46**
0.48**
0.33*
0.18
0.27
0.60**
0.30
0.82**
0.66**
0.71**
0.69**
0.60**
0.56**
0.44**
0.39*
0.56**
0.50**
0.43**
0.31**
0.58**
0.40**
0.35**
0.56**
0.26
0.06
0.47**
0.51**
0.17
0.17
0.12
70.07
0.26
0.48**
0.20
0.01
0.37
0.50**
0.10
0.21
0.43**
0.33*
0.37*
0.62**
0.30
0.18
0.56**
0.46**
0.11
0.38
0.68
0.54
0.48
0.69
0.52
0.10
0.91
0.70
0.22
0.22
0.15
70.10
0.36
0.66
0.37
0.03
0.61
0.72
0.13
0.32
0.53
0.51
0.55
0.91
0.64
0.24
0.69
0.16
Mean
0.57
0.48
0.58
0.37
0.36
0.52
0.34
0.51
0.23
Table 4 Cross-classification of nutrient distribution quartiles from 24 h recalls and AFFQa calculated from energy-adjusted nutrient intake (except
for total calories), Sao Paulo, Brazil
Lowest quartile 24 h recall
Overall
Lowest quartile
AFFQ (%)
Highest quartile
AFFQ (%)
Highest quartile
AFFQ (%)
Lowest quartile
AFFQ (%)
Exact
agreement (%)
Opposite
(%)
26
42
32
37
47
26
32
47
63
16
16
11
16
5
0
16
21
0
11
10
30
45
25
30
45
30
30
55
55
25
20
10
10
20
5
5
10
5
10
20
30
35
28
29
39
23
25
43
56
20
9
5
6
6
1
5
8
1
5
8
Mean
37
11
37
12
33
Nutrients
Discussion
The AFFQ was judiciously designed following the recommendations of Block et al, (1985a,b), Willett (1998), and
Nelson (1997) for determining a list of foods and portions
sizes according to gender and coherent with dietary patterns
and eating habits of the population group which was the
object of this study.
Comparing the AFFQ with the 24 h dietary recalls, similar
values for energy, carbohydrates, total fat and calcium intake
were observed, suggesting a high consistency in estimating
these nutrients. However there was a significant difference
European Journal of Clinical Nutrition
634
for the seven remaining nutrients (protein, polyunsaturated
fat, dietary fiber, cholesterol, retinal, vitamin C and iron).
The results of this validation study explain why there is
such considerable variability in the daily consumption calculated by means of the 24 h recalls. Our results are similar to
those of Beaton et al (1983), Sempos et al (1985) and Field et al
(1999), although only the latter was administered to children.
Patterns of within-person variability derive primarily from
individuals social behavior, although other factors are also
responsible. Harbottle and Duggan (1994), Tsubono et al
(1998) and Friis et al (1997) suggest that nutrients which
are part of the daily diet, that is, which are consumed on a
regular basis (eg macronutrient intake) have smaller variance
ratios (l) than those nutrients with a high within-person
variability, as in the of retinol, vitamin C and cholesterol in
the present study.
Considering the unadjusted correlation, higher values
(mean r 0.57) were observed compared with studies conducted by Rockett et al (1997), r 39, and Field et al (1999),
r 0.28.
The fact that the correlation diminished considerably
after adjustment for total energy intake led us to search for
possible answers. According to Willet (1998), this procedure
increases the correlation coefficient when variability of
nutrient consumption is related to energy intake, but
decreases when variability of the nutrient depends on systematic errors of overestimation and underestimation. In
studies focusing on adults by Willett et al (1985), Overvad
et al (1991) and Rimm et al (1992), this procedure confers
more relevant r values. In studies conducted in Greece, by
Gnardellis et al (1994) and in the USA by Munger et al (1992),
the two effects (increasing and decreasing the correlation)
happen simultaneously for the different nutrients analyzed.
On the other hand, in the study conducted by MartnMoreno et al (1993), in Spain, the effect is particularly
insignificant.
Although the 24 h recall is conceptually different from the
AFFQ , both methods have a common characteristic, namely,
they share some of the same sources of errors such as the sub
under- or overestimation of the quantities of foods consumed due to memory flaws. In this sense, artificially high
correlations, as is the case with energy (female, r 0.87;
male, r 0.82), may be explained.
Given a situation in which all elements of the diet are
informed proportionally, we can presume that energy adjustment may compensate for errors in general information. The
modest correction using the strategy of energy adjustment in
Flegal and Laarkins study, and the correlation coefficient
decrease in this study, led us to question the validity of this
statement. We may suppose that the subject in this study did
not report the nutrients in a similar way when answering
both instruments. Therefore, although energy adjustment
has made it possible to remove general and common differences between the methods, it did not permit differences
resulting from disproportional information to be removed
(Flegal & Laarkin, 1990; Flegal, 1999).
European Journal of Clinical Nutrition
References
Beaton GH, Milner J, McGuire V, Feather TE & Little JA (1983): Source
of variance in 24-hour dietary recall data: implications for
nutrition study design and interpretation. Carbohydrate sources,
vitamins, and minerals. Am. J. Clin. Nutr. 37, 986 995.
Munger RG, Folsom AR, Kushi LH, Kaye SA & Sellers TA (1992):
Dietary assessment of older Iowa women with a food frequency
questionnaire: nutrient intake, reproducibility and comparison
with 24-hour dietary recal intrerviews. Am. J. Epidemiol. 136,
192 200.
Nelson M (1997): The validation of dietary assessment. In: Design
concepts in nutrition epidemiology, 2nd edn. B Margetts & M Nelson,
pp 241 272. Oxford: Oxford University Press.
Nuzzo L (1998): Avaliacao do estado nutricional de adolescentes de uma
o particular de ensino. Sao Paulo: Tese de Mestrado, Faculinstituica
dade de Sau de Pu blica da USP.
Ocke MC, Bueno-de-Mesquita HB, Goddijn HE, Jansen A, Pols MA,
Staveren WAV & Kromhout D (1997): The Dutch EPIC food
frequency questionaire. I. Description of the questionnaire, and
relative validity and reproducibility for food groups. Int. J. Epidemiol. 26(Suppl. 1), 37S 48S.
Overvad K, Tjonneland A, Haraldsdottir J, Ewerte M & Jensen OM
(1991): Development of semiquantitave food frequency questionnaire to assess food, energy and nutrient intake in Denmark. Int.
J. Epidemiol. 20, 900 905.
Philippi ST, Szarfarc SC & Latterza AR (1996): Virtual Nutri (software),
Version 1.0 for Windows. Sao Paulo: Departamento de Nutric ao,
Faculdade de Sau de Pu blica, Universidade de Sao Paulo.
Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB &
Willett WC (1992): Reproducibility and validity of an expanded
self-administered semi-quantitative food frequency questionnaire
among male health professionals. Am. J. Epidemiol. 135,
1114 1126.
Rockett RH & Colditz GA (1997): Assessing diets of children and
adolescents. Am. J. Clin. Nutr. 65(Suppl), 1116S 1122S.
Rockett HRH, Breitenbach M, Frazier AL, Witschi J, Wolf AM, Field AE
et al (1997): Validation of a youth=adolescent food frequency
questionnaire. Prev. Med. 26, 808 816.
Sempos CT, Johnson NE, Smith EL & Gilligan C (1985): Effects of
intraindividual and interindividual variation in repeated dietary
records. Am. J. Epidemiol. 121, 120 130.
Thompson FE & Byers T (1994): Dietary assessment resource manual.
J. Nutr. 124(Suppl 11).
Tsubono Y, Fahey MT, Takahashi T, Iwase Y, Iitoi Y, Akabane M &
Tsugane S (1998): Interpopulation and intrapopulation variability
of nutrient intake in five regions of Japan. Eur. J. Clin. Nutr. 52,
176 179.
Willett WC (1998): Nutritional Epidemiology, 2nd edn. Oxford: Oxford
University Press.
Willett WC & Stampfer MJ (1986): Total energy intake: implications
for epidemiological analyses. Am. J. Epidemiol. 124, 17 20.
Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J et al
(1985): Reproducibility and validity of a semiquantitative food
frequency questionnaire. Am. J. Epidemiol. 122, 51 65.
635