There is growing interest in the role of peak bone mass in children and adolescents as determinant of future bone health (1). Attention has focused on nutrient intake in the youth, especially on calcium (2,3) and vitamin D intake. Consequently, there is a need for a method of estimating dietary calcium intake that is practical for large-scale surveys and epidemiologic studies in the younger population. Moreover, as dietary sources of calcium are represented by foods over a wide range of energy density (7) and because excess energy intake (and subsequent risk of childhood obesity) represents a major concern in developing and implementing dietary guidelines in this age group, any evaluation of calcium intake must take total energy intake into account.
Accurate assessment of the nutrient intake of free-living persons remains a difficult and labor-intensive process. No single method of assessment of an individual's usual intake is optimal under all conditions. Available methods range from quantitative approaches involving measurement of food portions to qualitative methods (4). The choice of method depends on the aim of the study, the accuracy of the dietary data required and the funds and personnel available (5). Probably the most accurate method to calculate dietary intake is the weighted food record (4). However, this method is time consuming and generally suitable only for individuals or small groups of cooperative volunteers. Routine assessment of diet in a large numbers of individuals from a range of socioeconomic backgrounds requires a quicker and simpler method of estimating the intake of specific nutrients.
Food frequency methods are a practical and efficient approach (8), but these methods for assessing calcium intake in children and young people have not been widely validated.
The objective of this study was to compare the dietary calcium intakes assessed by a semiquantitative food frequency questionnaire designed for children (aged 6-10 years) and adolescents (aged 16-20 years) and the 7-day food record methods.
We have adapted a food frequency questionnaire (FFQ) (6) validated in adults to assess nutrient intakes. We added some food items, which represent favorite food choices in age groups of our interest, and in particular those foods accounting for a significant proportion of calcium intake in the young population.
SUBJECTS AND METHODS
Thirty-seven healthy subjects, were recruited among students of primary and secondary schools nearby Milan. They were either prepubertal children (n = 18; 9 boys and 9 girls), aged from 6 to 10 years or adolescent boys and girls (n = 19; 9 boys and 10 girls) aged from 16 to 20 years. Their height and weight measurements were between third and 97th percentiles for age. They were not following any special dietary regimen or taking any dietary supplement. All participants were asked to answer to the questions of a FFQ by a trained dietician. Subsequently, they were trained to keep a 7-day weighed record.
The study was performed according to the recommendation of the Declaration of Helsinki, and informed consent was obtained from each participant and his or her legal guardian.
Food Frequency Questionnaire
The food frequency questionnaire consisted of 16 main food groups (e.g., milk and dairy products, pasta and rice), non-alcoholic drinks, cereals and oven products (e.g., bread and biscuits). Grouping was based on food composition in nutrients and customary use in the pediatric Italian population. Inside each group there were different items chosen to better define the intake of specific nutrients. For example, we considered milk and dairy foods separately, according to the level of calcium content (e.g., cheese with less or more than 600 mg of calcium per 100 g of product). Moreover, the calcium content of water (tap water and bottled mineral water) was considered: the calcium content of tap water was provided by the local water authorities of each district and the calcium content of bottled mineral water must be declared on the label; therefore, this information was readily available. The calcium content of water ranged between 4.25 mg/L to 455.5 mg/L.
Food frequency consumption of each item was evaluated using three categories: daily, weekly and monthly, each one expressed as 1 to 6 times by the interviewed subject. Portions were classified in three sizes (A, B and C), displayed as a photograph for each item, corresponding to small, medium and large portions adjustable for 20% more or less than the chosen portion. The interview took about 30 minutes per person; all interviews were performed by the same dietician. Children who were not able to correctly answer to the questionnaire were helped by their parents.
Seven-day Weighed Food Record
The day after the food frequency interview, the subjects or, in the younger children group, their parents, started the 7-day weighed record diary. An electronic scale (Soehnle, Murrhardt, Germany) was given to each subject to quantify the food portions consumed and subjects and parents were taught how to use household measures. Subjects were also asked to provide detailed description of each food, including method of preparation and recipes, whenever possible. Participants were also instructed to follow their usual diet.
Foods and nutrient intakes obtained from FFQ and FWR were computed by the same dietician who performed the interview. The analyses have been done using a software program containing the food composition data of the Italian National Institute of Nutrition (Winfood 1.0b version, Medimatica, Martinsicuro, Italy).
The food frequency questionnaire and the 7-day data were analyzed comparing the intake of total energy (Kcal/day), protein (grams), fat (grams) and carbohydrate (grams) and calcium (milligrams/day). The mean intake data obtained from the two methods were compared using Student's t-test for paired samples because the variables were normally distributed (Shapiro-Wilk test). Analyses were conducted at the α = 0.05 level.
Correlations between the two methods were assessed by Pearson's correlations coefficients. The Bland and Altman statistical method was used to evaluate the agreement between the two dietary intake assessment methods (9).
The agreement of the two methods in classifying a subject according to the intake of energy, macronutrients and micronutrients was examined by determining the frequency of similar classification into terciles or gross misclassification into opposite extreme terciles and by calculating the kappa coefficient.
The nutrient intakes determined by the FFQ and the FWR in young adult and children expressed as mean ± standard deviation are reported in Table 1. The significance of the differences of dietary intake assessed by Student's t-test is reported in the same table.
Significant differences were found for all the nutrients (Table 1) in both groups. We also evaluated the percentage of FFQ values compared with FWR, expressed as the %FFQ/FWR ratio. The differences between the calcium intake obtained by the different methods in each subject were well distributed around the mean of the difference (on average 50% of values above and 50% below the line representing the mean of the differences), and they were, on average, within 2 standard deviations (Fig. 1) in all subjects. Figure 2 shows the relationship between FFQ and FWR values about calcium estimation. The line of identity is represented to show the gap between FFQ and FWR values.
Table 2 shows Pearson's correlation coefficients between nutrient intake estimated by the FFQ and FWR methods. The highest values of correlation obtained are for energy and macronutrients. The correlation (r) for calcium intake was high (r = 0.71).
The evaluation by the Bland and Altman method showed that there was a good agreement between the FFQ and the FWR for each nutrient estimate (data not shown).
The classification of nutrients into terciles was used to evaluate the agreement between classes of subjects for FFQ and FWR (Table 4). The two methods classified approximately two thirds of the subjects into the same tercile for most nutrients. Gross misclassification (from one extreme tercile to the other) occurred for 2% (range, 0% to 5%) of the entire sample. Energy intake had no grossly misclassification; carbohydrate intake had the highest grossly misclassification rate. Kappa coefficients for energy, fat, carbohydrate and calcium intake were 0.33, 0.32, 0.36 and 0.34, respectively. Lower kappa coefficient was found for protein intake (0.16).
The aim of this study was to evaluate the validity of a food frequency questionnaire designed to provide rapid information about calcium intake in pediatric and young adult populations. To validate the questionnaire, we used the food weighted record as a reference method because of its well known advantages: it does not depend on recall and it allows direct measurement of food quantities. The 7-day food weight record is an acceptable compromise between the usual dietary intake and a satisfactory compliance. Indeed, a prolonged record period does not necessarily improve the accuracy of the method (10). Moreover, no significant differences were found in dietary habits of the Italian population between weighed record periods of different days or between seasons (11).
No reproducibility study (that requires repeating several months apart) (12) was appropriate in these age groups because of the rapidly changing dietary habits and dietary intakes of growing children and adolescents. According to other authors, the food frequency questionnaire seems to overestimate intakes for calcium and the other studied nutrients (13,14). In our study, calcium intake by FFQ was about 48% higher than that obtained by FWR. An explanation for such a discrepancy is that the subjects included foods high in calcium when responding to the FFQ but failed to eat those same foods during the 7-day recording period. The long list of food items and food groups is also likely to cause overestimation of nutrient intakes.
Small differences between the two methods were observed for carbohydrate intakes. This could depend on the relative importance of carbohydrates in the Italian diet.
Overall, nutrient intake estimation by FFQ is more precise in children than in adolescents. This could be attributable to a more stable diet of prepubertal children than of 16- to 20-year- old boys and girls. The only exception is fat, the nutrient that could better represent dietary habits of the family itself. Children had a very different use of snacks, fried foods, oil, butter or margarine, cakes, biscuits and pastries. The adolescent group, instead, had a common way of eating snacks and foods containing fat.
On average the diet of our children and older boys and girls, according to the Italian recommended nutrient (LARN) (15) and American recommended daily allowance, was within the normal range for total energy, proteins, fat and carbohydrates and calcium intake. Total energy includes energy derived from alcohol; approximately 20% of older boys and girls had frequent social drinking.
There was a good correlation between weighed intake and questionnaire-estimated intake, indicating that the questionnaire is useful in ranking a pediatric population. Correlation coefficients found in literature between nutrient intakes recorded from a food frequency questionnaire and weighed records vary considerably (10,16,17). It is difficult to compare different results, as each questionnaire has its own characteristics on the way food items, portions and relative frequency are built up.
The higher associations were always found in young adult aged from 16 to 20 years in respect to children. The lowest correlation was found in children for fat (r = 0.7); this could reflect excessive parental attention in completing the 7-day weighed record regarding what their children usually eat as snacks and everything else containing fat.
We found the same correlation for calcium in young adults and children (r = 0.7), probably because they were, on average, well informed about the risks of a diet poor in calcium. The highest (r = 0.9) association was found for protein and fat in young males aged 16-20 years: this result could suggest that our adolescent population had a more stable diet, at least for the analyzed nutrients. This result is in accordance to the well-known change of dietary habits of growing children.
The correlation between the two methods is good for calcium intake; linear regression is almost parallel to the identity line, the mean of the differences of values being 348 mg/day, but the calcium estimate by FFQ could be 144 mg/day below or 840 mg/day above the FWR.
The classification into terciles, showing the similarity of the two methods in classifying relative intakes, was satisfactory (Table 4), as the result for total energy and all nutrients on average is 65% of similar tercile classification, with a grossly misclassification range between 2% and 5% for macronutrients. The best classification (67% on average) compared to all nutrients regards protein. Calcium terciles classification is satisfactory: on average 62% of the sample classified itself in the same tercile; the best classification (75%) is for the first tercile, in accordance with the previous result (Table 4). The grossly misclassification for calcium is quite low (2%).
Although the present questionnaire was developed and validated in an Italian pediatric population, we can presume that this could as well be used in its present form in populations with similar food habits (e.g., southern Europe) or could be adapted for use in other Western populations with minor changes.
In epidemiologic and clinical settings, the usefulness of the FFQ would be in its ability to classify individuals into categories of nutrient intake (18). A high degree of cross-classification agreement was found between the two methods in this study. An important purpose of the FFQ would be to correctly identify children and adolescents with low intakes to target intervention strategies.
There are few validation studies of food frequency questionnaire among for the estimation of macronutrient and calcium intakes the age groups we considered. Jensen et al. recently developed a food frequency questionnaire to estimate calcium intake of Asian, Hispanic and white youths (n = 161) aged between 10 to 18 years (19). The mean correlation coefficient between calcium intake estimates with FFQ versus two 24-hour dietary recalls was 0.54 for the total sample and 0.73 in the group aged 14 to 18 years, higher than our correlation. However, the age range of the participants was smaller and the methods used for validation were different from those in our study. Rockett et al. developed and validated a FFQ in 261 youths (aged 9 to 18 years) comparing the average of the three 24-hour dietary recalls (20). The mean correlation for the energy-adjusted nutrients in older subjects (14 to 18 years) was 0.49, similar to that found among our adolescents group. Our results in younger children compare well with results obtained by Taylor et al. (21), who validated a short food frequency questionnaire to assess calcium intake in children aged 3 to 6 years. They showed that FFQ overestimated actual calcium intake in young children with a reasonable ability to correctly classify the majority of subjects into quartiles of intakes.
We conclude that while the FFQ developed could not assess calcium intake accurately, the application of this validated FFQ for ranking children and adolescents in low, medium or high calcium intake class is extremely useful for epidemiologic studies evaluating the relation between nutrient intake and diseases.
The authors thank Prof. F. Fidanza, M. G. Gentile and M. Porrini for allowing the adaptation of their original food frequency questionnaire.
1. Mora S, Gilsanz V. Establishment of peak bone mass. Endocrinol Metab Clin North Am
2. Ruiz JC, Mandel C, Garabedian M. Influence of spontaneous calcium
intake and physical exercise on the vertebral and femoral bone mineral density of children
. J Bone Miner Res
3. Martin AD, Bailey DA, McKay H, Whiting S. Bone mineral and calcium
accretion during puberty. Am J Clin Nutr
4. Marr JW. Individual dietary surveys: Purposes and methods. World Rev Nutr Diet
5. Pekkarinen M. Methodology in the collection of food consumption data. World Rev Nutr Diet
6. Fidanza F, Gentile MG, Porrini M. A self-administered semiquantitative food-frequency questionnaire with optical reading and its concurrent validation. Eur J Epidemiol
7. Block G, Norris JC, Mandel RM, Di Sogra C. Sources of energy and six nutrients in diets of low-income Hispanic-American women and their children
: quantitative data from HHANES, 1982-1984. J Am Diet Assoc
8. Treiber FA, Leonard SB, Frank G, et al. Dietary assessment instruments for preschool children
: reliability of parental responses to the 24-hour recall and a food frequency questionnaire
. J Am Diet Assoc
9. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
10. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire
. Am J Epidemiol
11. Alberti Fidanza A, Alunni Paolacci, Chiuchiù MP, et al. Dietary studies on two rural Italian population groups of the Seven Countries Study. 1. Food and nutrient intake at the thirty-first year follow-up in 1991. Eur J Clin Nutr
12. Willet W. Nutritional Epidemiology
. 2nd ed. Oxford: Oxford University Press, 1998.
13. Pao EM. Obtaining data on food intake. In: Livingstone GE, ed, Nutritional Status Assessment of the Individual
. New York: Food & Nutrition Press, 1987.
14. Stuff JE, Garza C, Smith EO, Nichols BL, Montandon CM. A comparison of dietary methods in nutritional studies. Am J Clin Nutr
15. Livelli di Assunzione Raccomandata di Energia e nutrienti per la Popolazione italiana LARN. Rome, Italy: Società Italiana di Nutrizione Umana 1996.
16. Tjonneland A, Overvad K, Haraldsdottir J, et al. Validation of a semiquantitative food frequency questionnaire
in Denmark. Int J Epidemiol
17. Lindroos AK, Lissner L, Sjostrom L. Validity and reproducibility of a self-administered dietary questionnaire in obese and non-obese subjects. Eur J Clin Nutr
18. Block G. A review of validations of dietary assessment methods. Am J Epidemiol
19. Jensen JK, Gustafson D, Boushey CJ, et al. Development of a food frequency questionnaire
to estimate calcium
intake of Asian, Hispanic, and white youth. J Am Diet Assoc
20. Rockett HR, Breitenbach M, Frazier AL, et al. Validation of a youth/adolescent food frequency questionnaire
. Prev Med
21. Taylor RW, Goulding A. Validation of a short food frequency questionnaire
to assess calcium
intake in children
aged 3 to 6 years. Eur J Clin Nutr