Canada has the world's highest burden of inflammatory bowel disease (IBD) both in children and in adults. Findings from twin studies and familial aggregation studies indicate that >50% of the etiology of IBD may be related to nongenetic exposures such as diet. However, evidence on diet has been equivocal, perhaps owing to the use of retrospective designs and nonvalidated food frequency questionnaires (FFQ) in dietary assessment. Studies using validated FFQ close to the diagnosis have suggested that specific dietary patterns may increase risk of IBD in children (1), with Western diets (including high quantities of refined sugar and total fat) conferring risk and prudent diets (comprising high quantities of fruits and vegetables) conferring protection. Other studies focusing on individual nutrients or dietary components have shown that diets with higher ratios of linoleic acid (n-6 polyunsaturated fatty acids [PUFA]):docosahexaenoic acid (n-3 PUFA) and low intakes of dietary fiber, fruits, and vegetables (1–4) could increase risk of IBD. A recent prospective study (5) based on a European cohort has reported that higher consumption of n-6 PUFA and a lower consumption of n-3 PUFA was significantly associated with adult-onset ulcerative colitis, a form of IBD, lending support to the findings of earlier studies.
In IBD risk factor studies, errors in diet recall could result if data are collected close to disease onset or postdiagnosis, because patients could have already changed their diet to manage digestive symptoms or may have started a therapeutic diet on the advice of their physician or dietitian. To avoid such recall bias, a planned study will recruit a cohort of healthy siblings of individuals affected with IBD. Given that usual diet, rather than diet consumed at a specific time period, may be related to the risk of developing IBD (2), an FFQ was sought for assessment of diet as an etiological factor in this planned IBD study. However, no validated Canadian FFQ was available for assessing usual diet among children. The objective of this research was to assess the accuracy of a Canadian adult FFQ in children and adolescents ages 7 to 18 years.
PATIENTS AND METHODS
Population and Sampling
The FFQ validation study targeted healthy children and adolescents ages 7 to 18 years, consistent with the ages of the intended IBD study population. Participants were recruited from youth attending the outpatient acute trauma center of the orthopedic clinic at Hôpital Ste-Justine in Montréal, Canada, where a computerized list is maintained of children scheduled to visit the clinic for follow-up of their injuries. These lists were screened to identify children who had experienced minor injuries or fractures, and an information letter was sent to potential recruits before they came into the clinic. On the day of the visit, the children (and/or parents) were approached to seek their participation in the study. Eligible children had to be free of any current disease, including digestive disease; they should not be taking medication that could modify their appetite or recent diet (ascertained using an interviewer-administered supplementary questionnaire), should speak and read either French or English, and should have a parent willing to be involved in reporting dietary data. Participants were recruited by age group (7–9 years, 10–12 years, 13–15 years, 16–18 years) with the goal of including approximately 25 children in each age group, and assembling a sample including both French-speaking and English-speaking children/youths to reflect the linguistic composition of the population. The age groups broadly reflect developmental stages in children's abilities to reflect on their food intakes and develop the cognitive capacities necessary for completing a dietary questionnaire. Recruitment and data collection for the FFQ validation study took place from February to December 2008. The study was approved by the institutional review board of the study center, and informed written consent was acquired from the participants.
FFQ Development and Validation
Development and validation of the semiquantitative FFQ have been presented elsewhere (6). It was designed to collect information on usual consumption frequency and quantity of 78 foods and beverages during the previous 12 months, as well as the types of fats and oils used in cooking and on foods. Background consumption frequency and portion size information and nutrient values were initially derived from population-based food consumption data extracted from a nutrition survey of adults in Québec (7). Although we met our objective of developing a relatively brief FFQ food list, foods retained contributed to at least 80% of nutrient intakes among consumers. To aid in accurate estimation, the respondent was directed to food-specific photographs of sample portion sizes on the facing pages. The FFQ contains detailed instructions for completion and takes around 30 minutes on average to complete when self-administered. Estimates of absolute nutrient values and food group servings are output, and the tool permits ranking of individual intakes.
The instrument was designed to be self- or interviewer-administered. It was developed, pretested, and validated in both French and English in several adult populations against 4-day nonconsecutive food records (FR). Results of previous validation studies among nonpregnant women and men ages 18 to 82 years (6), in older adults ages 70 years and older (8–10), and among pregnant women ages 18 to 47 years (11) indicate that the FFQ is a valid instrument for determining usual diet in the adult Canadian population. Version 3.15 of the FFQ was used in the present FFQ validation study.
FFQ Validation Procedure
The study coordinator was trained in the administration and use of the FFQ. The FFQ was self-administered by the youth or child/parent in the clinic setting, with the study coordinator present to answer questions and provide assistance. It was completed by the child's parent for children ages 7 to 9 years. The study protocol stipulated that the parent should be present to assist children/youths ages 10 to 12 years and 13 to 15 years, but that subjects ages 16 to 18 years should independently complete the FFQ, in the absence of the parents, to enhance the accuracy of the reporting.
Following our standard practice, the FFQ was validated using a 3-day non consecutive estimated food record (3D-FR). A data collection package was distributed to willing recruits. The package included the information letter on the validation study and an FR booklet (6) with a structured form for recording details such as the meal or snack, how to provide exhaustive information on each food and beverage taken (eg, detailed description, brand name, label information), food preparation methods, quantities consumed, and home recipes used, where relevant. Two-dimensional measuring aides were provided in the booklet along with a completed, sample FR. The respondent was provided with dates for completion of the 3D-FR (2 weekdays and 1 weekend day). A stamped, addressed envelope was included in the package for returning the completed 3D-FR to the dietary data handling center (Centre de Recherche, Institut Universitaire de Gériatrie de Montréal).
Data Entry and Nutrient Analysis
The FFQs were data entered using Microsoft Access software (Microsoft, Redmond, WA) for customized data entry. Analysis was based on algorithms developed for computing energy and nutrient values from the instrument food list, frequency options, and portion sizes (6). Nutrient values were calculated from the food and nutrient database developed for the FFQ, using the CNF2007b (12) incorporated into the FFQ data entry utility. Data entry included systematic double entry to verify accuracy. FFQ data were exported to Microsoft Excel and then transferred to SPSS (version 16, SPSS Inc, Chicago, IL) for preliminary analysis of questionnaire plausibility. The preliminary analysis of the FFQ was conducted by the nutrition team to detect outliers and assess the plausibility of the FFQ data based on a set of established criteria. This process excludes FFQs having full blank page(s), ≥10% of foods with missing answers for both frequency and portion size, or energy intakes ≤800 kcal or ≥4000 kcal (13). None of the 65 FFQ was excluded. The FFQ food and nutrient files were output in Microsoft Excel format along with reports on the plausibility of the FFQs to aid in interpretation and best use of the FFQ dietary and nutritional analyses.
The FRs were analyzed using the CANDAT dietary analysis system (version 6.0, Godin London, London, Ontario), based on the CNF2007b (12); results were exported to Microsoft Excel and then transferred to SPSS for further analyses. The 3 days of FR data were aggregated in SPSS to provide mean daily values for energy and nutrients.
Data are reported only for food intakes (no supplements). Because data were skewed, as is common with dietary and nutrient data, nonparametric statistics were used (14). Analyses were done on the whole sample and then stratified by sex and by age group. Initially, age was stratified by the 4 recruitment age groups; because of small numbers in these 4 groups they were then grouped into 2 categories: 7 to 12 years and 13 to 18 years. Unadjusted nutrient intakes (energy and 24 nutrients) estimated from the FFQ were compared with the mean of the 3D-FR using the Wilcoxon signed-rank test. The percentage of over- or underestimation of nutrients from the FFQ in reference to the mean of the 3D-FRs was determined ([(FFQ − 3DFR)/3DFR] × 100). Spearman rank correlation analyses were conducted to test associations between unadjusted FFQ energy and nutrient estimates and those obtained from the reference method (3D-FR). To determine the extent of misclassification of nutrient estimates from the FFQ, joint classification of respondents into quartiles of the distribution was compared for nutrients from the FFQ and the mean of the 3D-FR after the method of Willett et al (15). Proportions of respondents classified into the same quartile, the same and contiguous quartile, or frankly misclassified (lowest quartile in 1 method classified as highest quartile in the other method) (15) were calculated for each nutrient of interest. Finally, Bland-Altman analysis (16) was carried out to assess the extent of agreement between absolute estimates from the FFQ and 3D-FR. The differences between FFQ and 3D-FR were plotted against the mean of the 2 dietary measures, and the 95% limits of agreement (LOA) were calculated on pairs of measurements (LOA ± 2SD). SPSS version 16 was used in the analyses. Differences were considered significant at P < 0.05.
In all, 131 participants were recruited (80% Francophone). Of these, 82 respondents (63%) returned their FFQ to the study coordinator. A maximum of 5 reminders were given to the nonrespondents. From these 82 respondents, 70 (85%) also provided a complete 3D-FR.
Of the 70 participants who provided a complete set of FFQ and 3D-FR, the FFQs from 5 participants were subsequently found to have been poorly completed and were eliminated from analyses. There were no differences between participants and nonparticipants on the selection criteria of age, sex, and language (data not presented). The final sample retained for FFQ validation analyses thus consisted of 65 children/adolescents with acceptable FFQ and 3D-FR (54% girls). Average age was 11.7 ± 2.6 years with most (90.8%) in the 3 younger age groups (7–9 years, 10–12 years, and 13–15 years). The majority (86%) were Francophone. Almost two-thirds of FFQ and a similar proportion of FR were completed by the child's/youth's mother, with an additional 9% to10% completed by the father/other adult. Only 26% of all FFQ and 21% of 3D-FR were completed by the child/youth (Table 1).
Absolute intake estimates are shown in Table 2. Overall, the FFQ overestimated intakes from the 3D-FR. The average median difference between the 2 instruments was approximately 15% for energy and macronutrients, with greater discrepancies for certain nutrients, notably vitamins A and C. The differences in intakes were lower for those in the 7 to 9 years age group (data not shown).
Table 3 provides results of Spearman rank correlation analyses between nutrients from the FFQ and 3D-FR for the entire sample, and by the 2 broader age categories (7–12 years, 13–18 years) and by sex. Although there was some variability depending on the nutrient, the strongest associations overall between nutrients estimated by FFQ and 3D-FR were found in adolescents ages 13 to 18 years, whereas stronger associations were found for most nutrients among girls compared with boys. Results for the sample as a whole indicated an acceptable degree of relative validity between the test (FFQ) and reference (3D-FR) instruments. Spearman correlation coefficients were positive and most were statistically significant (0.05 < P < 0.0001). With the exception of low correlations between FFQ and FR for dietary fiber, folate, and PUFA, the remainder ranged from 0.22 for vitamin C to 0.57 for saturated fat, with a mean Spearman correlation for energy and the 24 nutrients of 0.38. The average increased to 0.41 when dietary fiber, folate, and PUFA were excluded (data not shown).
More than three-fourths (77%) of participants were cross-classified into the same half of the distribution for both instruments, with 39% exact agreement. Only 6% overall were frankly misclassified, with vitamins E, B6, and folate showing the highest extent of misclassification (Table 4).
Results of the Bland-Altman analyses for the sample as a whole are shown for selected nutrients of frequent interest among children and youth for energy (Fig. 1), for protein (Fig. 2), and for calcium (Fig. 3). The difference in intake between the FFQ and 3D-FR is plotted on the y-axis and the mean intake from the 2 tools is on the x-axis. Plots produced for the other nutrients, and for stratified analyses by the 2 broad age groups and by sex resembled those shown in Figures 1 to 3 (data not shown).
The data were further examined to establish whether the FFQ and 3D-FR had been completed by the same person as instructed, determine which FFQ were of doubtful plausibility, and ascertain whether 1 or more of the FR had been reported to be nonrepresentative of usual dietary intakes. We found that contrary to instructions, the instruments had been completed by different respondents (child/youth vs parent) in 6 instances. In addition, 15% of the FFQs retained in analyses initially raised some questions as to their plausibility, and 34% of respondents reported that 1 or more of their FRs was not representative of their usual eating habits. Boys were more likely than girls to have FFQs of questionable plausibility and to have reported that their 3 FRs were not representative of their regular diet. The distributions by age suggested that the “young teens” may have had more erratic eating behavior as evidenced by their reporting that the FRs were unrepresentative. However, because of the limited numbers in each subgroup, no statistical tests were performed to assess these differences (data not shown).
Due to their ease of administration and relatively low cost, FFQs are the method of choice in large population studies that seek to understand relations between usual diet and development of diseases, particularly those with a long latency period, such as IBD. In preparation for a Canada-wide study examining the contribution of diet to the etiology of IBD, we examined the accuracy of our previously developed adult FFQ for assessing diet in unaffected children. Although the FFQ is frequently criticized for its conceptual limitations resulting in subject misclassification due to respondent difficulties in estimating frequency and portion sizes (17), our results showed that the adult FFQ performed reasonably well overall in this sample of children and adolescents, despite its tendency to overestimate intakes relative to the 3D-FR.
It is often observed that the FFQ tends to overestimate intakes in adult validation studies (18), but data among children are limited. An Italian study was conducted in a small sample of children and adolescents (n = 37) to assess macronutrients and calcium using an FFQ validated by a 7-day weighed FR (19). Significant differences were reported between estimates from the FFQ and the FR; overestimates ranging from 11% to 55% were reported, depending on nutrient and age group. The degree of overestimation in our study (on average, 15%) was, however, considerably lower.
In 1 of the few North American FFQ validation studies conducted among youngsters, Rockett et al (20) reported on the validation of the YAQ (Youth/Adolescent Food Frequency Questionnaire) adapted from the Willett FFQ (21), in a sample of 261 participants ages 7 to 18 years. Although it was acknowledged that FRs are considered to be the criterion standard, 24-hour diet recalls were used as the reference method in consideration of issues of literacy and accessibility in their target group. The mean Spearman correlation for energy and 24 nutrients in the present FFQ validation sample was r 0.38, which compares favorably with the average Pearson r 0.39 reported by Rockett et al (20), and with the Spearman correlation coefficients reported by Marshall et al (22) in 2 groups of American children in the validation of the Iowa Target Nutrient Questionnaire and the Block Kids' Food Questionnaire, relative to 3-day food diaries. In the latter study, unadjusted correlations for nutrients ranged from 0.462 to 0.528 for the Target Nutrient Questionnaire and from 0.203 to 0.515 for the Block Kids' Food Questionnaire versus the 3-day diaries. Our results also resembled those from another American group in a study among 248 children in grades 6 through 8 to assess the performance of a brief instrument targeting calcium intakes in relation to three 24-hour diet recalls. Intraclass correlations between estimates from these instruments ranged from 0.33 to 0.59 across the demographic groups and were lower among boys compared with girls, and younger compared with older adolescents (23).
Stratified analyses on our data suggested that accuracy of the FFQ could vary by age group and sex, with some inconsistencies across nutrients among girls. In addition, initial stratified analyses on the 4 recruitment age groups produced greater inconsistencies in results. For example, although there was a strong agreement between FFQ and 3D-FRs in the youngest age group (7–9 years) this could presumably be explained by the fact that parents—usually the mother—uniformly completed both instruments for their children as instructed. The oldest group (16–18 years) included only 6 participants (data not shown). Consequently, there was little stability in the data, which precludes our ability to further clarify relations between the test and reference instruments for older adolescents. Although this can be partly explained by the relatively small sample size, and even lower numbers in the stratified groups, such inconsistencies are difficult to interpret. When regrouped into broader age categories, however, more similarities were observed in the strength of association between nutrients from the test and reference methods among the children and adolescents.
Others have also noted that FFQ reporting accuracy differs by age group (19,21,23). Bertoli et al (19) ascribed this difference to differing levels of age-related stability in food intakes, reporting that the younger children in their study had a more stable diet than adolescents. We also observed more accurate reporting in the youngest age group (7–9 years), but suggest that this was most likely because the parents were the respondents on behalf of young children. Indeed, agreement between the test and reference instruments was considerably better in this group than among adolescents, particularly those in the 10 to 15 years age range (data not shown).
While carrying out FFQ data entry, we noted that a number of participants had routinely selected large portion sizes, leading us to initially suspect the plausibility of their responses. However, further examination of the FFQ of participants who reported eating large portions showed that most were boys ages 10 to 15 years (data not presented). Their possible overestimation of intakes in the FFQ may have attenuated the correlations between the FFQ and 3D-FRs. In addition, differential reporting of large portion sizes by young adolescent boys could be responsible for the discrepancy in the extent of association between instruments observed in the different age groups. As suggested by Marshall et al (22) this may reflect conceptual difficulties experienced by young adolescents in estimating frequencies and portion sizes, rather than the instrument's capacity to assess intakes. Clearly, this signals the need to provide precise FFQ completion instructions and to be particularly attentive to the quality of FFQs completed by certain age-sex groups such as young adolescent boys.
In the present study, the children and adolescents in the older age groups (13 years and older) were instructed to complete the instruments on their own, with parents asked to aid the younger adolescents when necessary. Because some parents who completed the instruments for their child or adolescent may not be fully aware of what their children eat (22), it is possible that completion of the FFQ by the youth and the 3D-FR by the parent (or vice versa) may have increased discrepancies between the 2 instruments. In addition, some 15% of the FFQ retained in analyses were initially considered to be of doubtful plausibility and more than one-third (34%) of participants had reported that 1 or more of their 3 FR were not representative of a usual day. Their reasons—they had eaten in a restaurant, gone visiting, or had been on a vacation trip, among others—are consistent with true intraindividual variability and should not contribute to measurement error because they would be picked up as less frequently consumed items on the FFQ. Because the FFQ queries “usual diet,” comparing the FFQ with potentially nonrepresentative FRs could attenuate the strength of association between methods. However, careful reexamination of the foods reported on “nonrepresentative” FR did not reveal great diversity in consumption on the “nonrepresentative” days compared with the other FR days, and most of the days reported as nonrepresentative were weekend days. Indeed, study respondents are asked to provide information for 1 weekend day because it is well known that weekend food consumption typically differs from the weekday diet. Interestingly, the latter 2 issues were more prevalent among boys (70% and 59%, respectively) than girls (30% and 41%, respectively). Participants in the 10 to 12 years age group had higher proportions with doubtful FFQ plausibility, and those ages 13 to 15 years had higher proportions reporting that 1 or more of their FR were nonrepresentative of the usual diet (data not shown). Supplementary analyses excluded subjects with “doubtful” FFQs, with no impact on the results (data not shown), thereby providing support for retaining them in the analysis.
We found that overall, 77% of FFQ validation study participants were cross-classified into the same half of the distribution by both methods (identical and contiguous quartiles), and only 6% were frankly misclassified, that is, classified into the opposite quartile (lowest quartile in 1 method but into the highest quartile in the other) (15). This is consistent with findings from previous validation studies on this FFQ (6,8–11). Percent exact agreement in the present study was 39%, similar to the results reported by Marshall et al (22). Cross-classification of estimates for calcium from our FFQ and 3D-FR ranked 42% into the same quartile and 80% into the identical and contiguous quartile, which also closely resembles results reported by Jensen et al (24).
Finally, the Bland-Altman plots produced for energy (Fig. 1), protein (Fig. 2), and calcium (Fig. 3) show that results from the test and reference methods were within the LOA and were similar to those reported by Bertoli et al (19). Although Bland-Altman analysis is not typically reported in dietary assessment method validation studies, it provides an easily interpreted measure of agreement that complements correlation coefficients (25).
The present study has limitations, the chief one being the small sample size, which made it difficult to conduct meaningful stratified analyses. We had aimed to recruit 100 subjects, an adequate sample size for a validation study (26); however, we were unable to achieve that target. Unlike others, we did not carry out energy adjustment to improve correlations between the FFQ and 3D-FRs, or use the measurement error model (26) to deattenuate the correlations between the FFQ and FR. First, we, like others, have found that energy adjustment does not always improve associations between test and reference methods (11,27,28). Second, there is evidence that the measurement error model cannot always be successfully applied (29,30). In addition, study participants were recruited as volunteers from orthopedic clinics in a pediatric hospital, and had no vested interest in participating in the present study. Thus, their level of motivation may have had consequences for participation rates and attention paid to completion of study instruments. However, although the participation rate in the present study was modest, we could detect no selection bias because participants did not differ from nonparticipants. Finally, the FFQ was initially developed for adults. Still, assessment of mean portion sizes among different age groups at the time of FFQ development showed little variation across age groups within the adult population, and provision of a smaller portion size on the FFQ allows children who eat smaller serving sizes than adults to select the small portion size option.
In summary, analyses show that the FFQ performed reasonably well in this sample of children and adolescents, suggesting that the tool can be confidently used to rank subjects on a range of nutrient intakes with the potential to provide useful information on dietary risk factors in the etiology of IBD. However, it is recommended that the FFQ be completed by parents of younger children with input from other caregivers (eg, teachers, lunchroom monitors) where appropriate. In addition, if feasible, interviewer-based administration of the FFQ should be considered. When self-completion is the preferred mode, this should be carried out in a research setting, particularly for older children and adolescents, where a resource person could provide assistance when needed and on-the-spot verification and clarification of questionable responses can be carried out before the respondent leaves the clinic or research center. By paying particular attention to the quality of the completed FFQs, especially among young adolescent boys, the accuracy of the estimates could be enhanced.
1. D'Souza S, Levy E, Mack D, et al
. Dietary patterns and risk for Crohn's disease in children
. Inflamm Bowel Dis 2008; 14:367–373.
2. Bernstein CN, Shanahan F. Disorders of a modern lifestyle: reconciling the epidemiology of inflammatory bowel diseases. Gut 2008; 57:1185–1191.
3. Russel MG, Engels LG, Muris JW, et al
. Modern life' in the epidemiology of inflammatory bowel disease
: a case-control study with special emphasis on nutritional factors. Eur J Gastroenterol Hepatol 1998; 10:243–249.
4. Persson PG, Ahlbom A, Hellers G. Diet
and inflammatory bowel disease
: a case-control study. Epidemiology 1992; 3:47–52.
5. The IBD in EPIC Study Investigators. Linoleic acid, a dietary n-6 polyunsaturated fatty acid, and the aetiology of ulcerative colitis: a nested case-control study within a European prospective cohort study. Gut
6. Shatenstein B, Nadon S, Godin C, et al
. Development and relative validity of a food frequency questionnaire
in Montreal. Can J Diet
Prac Res 2005; 66:67–75.
7. Santé Québec, BL. Les Québecoises et les Québécois Mangent-ils Mieux? Rapport de l'Enquête Québécoise sur la Nutrition, 1990. Montréal: Ministère de la Santé et des Services Sociaux, Gouvernement du Québec;1995.
8. Shatenstein B, Payette H, Nadon S, et al. Validation
of a food frequency questionnaire
for home-dwelling elderly people in Québec: a pilot study. In: Proceedings from the 5th International Conference on Dietary Assessment Methods
. Chiang-Rai, Thailand; January 26–29, 2003.
9. Shatenstein B, Payette H, Nadon S, et al
. An approach for evaluating lifelong functional foods intakes in elderly people. J Nutr 2003; 133:2384–2391.
10. Shatenstein B, Payette H, Nadon S, et al
. Dietary intakes of home-dwelling elderly Quebecers obtained from a food frequency questionnaire
: a pilot study. J Nutr Health Aging 2003; 7:234–235.
11. Shatenstein B, Xu H, Luo ZC, et al. Relative validity of a food frequency questionnaire
for Canadian pregnant women. Can J Diet Pract Res
. In press.
12. Health Canada. Canadian Nutrient File
. Ottawa: Bureau of Nutritional Sciences, Health & Welfare Canada; 1982.
13. Shatenstein B, Huet C, Jabbour M. Plausibility assessment and quality assurance of food frequency questionnaires completed in studies of diet
and health. Paper presented at International Conference on Diet
and Activity Methods (ICDAM). Washington, DC; June 5–7, 2009. Conference abstracts PS4-6:281–282.
14. Millns H, Woodward M, Bolton-Smith C. Is it necessary to transform nutrient variables prior to statistical analyses? Am J Epidemiol
15. Willett WC, Sampson L, Stampfer MJ, et al
. Reproducibility and validity of a semiquantitative food frequency questionnaire
. Am J Epidemiol 1985; 122:51–65.
16. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8:135–160.
17. Thompson FE, Subar AF. Dietary assessment methodology. In: Coulston AM, Boushey C, editors. Nutrition in the Prevention and Treatment of Disease. 2nd ed. San Diego, CA: Academic Press; 2008. pp. 3–39.
18. Marks GC, Hughes MC, van der Pols JC. Relative validity of food intake estimates using a food frequency questionnaire
is associated with sex, age, and other personal characteristics. J Nutr 2006; 136:459–465.
19. Bertoli S, Petroni ML, Pagliato E, et al
of food frequency questionnaire
for assessing dietary macronutrients and calcium intake in Italian children
. J Pediatr Gastroenterol Nutr 2005; 40:555–560.
20. Rockett HRH, Breitenbach M, Frazier AL, et al
of a youth/adolescent food frequency questionnaire
. Prev Med 1997; 26:808–816.
21. Rockett HR, Wolf AM, Colditz GA. Development and reproducibility of a food frequency questionnaire
to assess diets of older children
. J Am Diet
Assoc 1995; 95:336–340.
22. Marshall TA, Eichenberger Gilmore JM, Broffitt B, et al
. Relative validity of the Iowa fluoride study Targeted Nutrient Semi-Quantitative Questionnaire and the Block Kids' Food Questionnaire for estimating beverage, calcium, and vitamin D intakes by children
. J Am Diet
Assoc 2008; 108:465–472.
23. Harnack LJ, Lytle LA, Story M, et al
. Reliability and validity of a brief questionnaire to assess calcium intake of middle-school–aged children
. J Am Diet
Assoc 2006; 106:1790–1795.
24. 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 2004; 104:762–769.
25. Flood VM, Smith WT, Webb KL, et al
. Issues in assessing the validity of nutrient data obtained from a food-frequency questionnaire: folate and vitamin B12
examples. Public Health Nutr 2004; 7:751–756.
26. Willett W. Nutritional Epidemiology. 2nd ed New York: Oxford University Press; 1998.
27. Lee KY, Uchida K, Shirota T, et al
. Validity of a self-administered questionnaire against 7-day dietary records in four seasons. J Nutr Sci Vitaminol (Tokyo) 2002; 48:467–476.
28. Deschamps V, de Lauzon-Guillain B, Lafay L, et al
. Reproducibility and relative validity of a food-frequency questionnaire among French adults and adolescents
. Eur J Clin Nutr 2009; 63:282–291.
29. Segovia-Siapco G, Singh P, Haddad E, et al
. Relative validity of a food frequency questionnaire
used to assess food intake during a dietary intervention study. Nutr Cancer 2008; 60:603–611.
30. Neuhouser ML, Tinker L, Shaw PA, et al
. Use of recovery biomarkers to calibrate nutrient consumption self-reports in the women's health initiative. Am J Epidemiol 2008; 167:1247–1259.