The European Journal of Public Health Advance Access published online on June 4, 2008
The European Journal of Public Health, doi:10.1093/eurpub/ckn017
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Variation in energy and nutrient intakes among pre-school children: implications for study design
Inge Huybrechts1, Dirk De Bacquer1, Bianca Cox2, Elisabeth HM Temme2, Herman Van Oyen2, Guy De Backer1 and Stefaan De Henauw1,3
1 Department of Public Health, Ghent University, Ghent, Belgium
2 Unit of Epidemiology, Scientific Institute of Public Health, Brussels, Belgium
3 Department of Health Sciences, Vesalius, Hogeschool Gent, Ghent, Belgium
Correspondence: Inge Huybrechts, International Agency for Research on Cancer (IARC)—World Health Organization, Nutrition and Hormones Group, 150, cours Albert Thomas, 69372 Lyon Cedex 08, France, tel: +33 (0) 472 73 83 49, fax: +33 (0) 472 73 83 61, e-mail: inge.huybrechts{at}ugent.be
Received August 29, 2007, accepted February 21, 2008
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Background: Within- and between-person variation in nutrient intakes has been characterized in adult populations, but little is known about variation in the diet of pre-school-aged children. The aim of this study was to describe dietary variations in Flemish pre-schoolers and to estimate the number of record days required for studying diet–disease associations among pre-school-aged children. Methods: Data from 3-day estimated diet records, collected in 2002–03, were used from 661 pre-school children (2.5–6.5 years) in Flanders, using parents/caregivers as a proxy. Age categories for studying differences in dietary variations between age groups were based on the age groups of the Belgian dietary recommendations (2.5–3 years and 4–6.5 years). Results: Overall, micronutrient intakes had smaller variance ratios than macronutrients. The largest variance ratios were found for cholesterol followed by fat, fatty acids and sodium intakes and would result in attenuated linear regression estimates of diet–disease associations in children. Within/between variance ratios were
1 for most nutrients in the oldest group (4–6.5 years) of pre-school children, while <1 for most micronutrients in the youngest age group (2.5–3 years), resulting in fewer days required for this youngest age group in comparison with the oldest group. No consistent differences in variance components were found between genders. Overall, 7-day dietary records were sufficient for accurately estimating 15 of the 23 nutrients in both age groups. Conclusion: The number of record days required for reliably classifying pre-school children raises with increasing age category (from 2.5–3 years to 4–6.5 years) for most nutrients and varies from 3 or 4 days for some nutrients like carbohydrates to 2 or 3 weeks for others like dietary cholesterol or monounsaturated fatty acids.
Keywords: Belgium, between-person variation, children, diet records, nutrient intake, within-person variation
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Accurate and precise assessment of habitual dietary intake remains a major challenge in studying diet–disease relationships.1 Day-to-day variability in individual energy and nutrient intakes adversely affects the statistical precision and accuracy of intake assessments. Therefore, day-to-day variability must be taken into consideration in the design of studies aiming to assess individual usual nutrient intakes and in the interpretation of these results.2–4
The estimation of within- and between-person variation is further useful in determining the optimal sample size and the number of replicate measurements required per individual, to study diet–disease relationships and usual intake distributions. Ideally, observations of diet over many days, weeks or months are preferred, but due to cost and respondent burden, less ideal, shorter term methods are often employed to estimate usual diet intakes.5 When using these short-term methods, day-to-day fluctuations around the usual intake can be partially removed during analysis at the group level, when the contribution of within-person (day-to-day) variation is known. This adjustment is crucial when only a single 24 h recall or one diet record is collected for estimating usual intake distributions.6 In this latter case, within- and between-person variances can be calculated from multiple days of intake collected from at least a subsample of individuals, or by borrowing variance estimates from another study population.7 However, for the assessment of relationships between nutrient intake and health status in individuals, long-term data on intake are always required. These long-term data (habitual diet intakes) on the individual level can only be calculated by using a sufficient number of records/recalls, or by using another or additional method for long-term intake assessments (e.g. Food Frequency Questionnaire).
Although different studies characterized within- and between-person variation of nutrient intakes in adults,8–19 the information available among children is very limited.4,10,20,21 Since dietary intakes and variance estimates differ between adult populations (depending on gender, age, culture, etc.), we might expect that the variance estimates among children are also gender, age and culture specific. Nelson et al.10 already revealed important differences for variances in nutrients between different age groups (from infancy to old age) and genders, whereas Jahns and colleagues4 showed important variance differences between Russian and American children.
While studies have already been published concerning variance components of nutrient intakes in young children (<2 years old)10,20,22 and older children (>5 years old)4,10,21,23, only limited information is available for children between 2 and 5 years old.10,24 Furthermore, in most of these studies that have already been published, the data used were collected in the eighties. Since dietary habits changed during the last decades, the variance components of nutrient intakes might be different in current dietary habits.10 Therefore, it is important to study variance components of nutrient intakes in the present generation.
In addition, evidence is showing important difficulties in the assessment of dietary intake among this pre-school-aged population, caused by some typical characteristics in this age group (e.g. short memory, limited knowledge of foods, attending pre-school classes, etc.).25 Hence, it is important to investigate the feasibility of using dietary records in this specific age category by estimating the number of days required for accurately estimating nutrient intakes.
Therefore, this study has two main objectives: (i) to describe and discuss dietary variations in Flemish pre-schoolers and (ii) to estimate the number of replicates needed for future studies in pre-school children, requiring a high degree of precision and accuracy on the individual level (e.g. studies investigating diet–disease relationships among children).
Since dietary recommendations and habits change at the age of 4 years, the authors studied differences in variances between children aged 2.5–3 years and 4–6.5 years.26–28
|
Subjects and methods |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Study population and sampling design
Data used for these analyses derived from a cross-sectional study in pre-school children in Flanders, Belgium, using a multistage cluster sampling design, with schools as primary sampling units and classes as secondary sampling units. The study design and the ‘usual’ nutrient intakes, calculated with the Nusser-method (using C-side29), have been described in more detail previously by Huybrechts and De Henauw.30 In brief, 3-day estimated dietary records (EDRs) of the children were completed by an adult proxy who spent most of the time with the child outside the school (generally the mother). Within the instructions for completing the diet records, the following approach was recommended to the parents for recording portion size information (depending on the information available):
- Whenever the exact weight or volume was available in grams or millilitre, the portion sizes had to be recorded in these standard metrics.
- However, for some foods (like fruit) standard units/portions could be reported (e.g. one medium-sized apple).
- For items or recipes that were difficult to quantify or to describe as standard units/portions (e.g. for spaghetti bolognaise and French fries), parents could use household measures like ‘half a plate’. The dieticians who encoded the diaries used standard serving sizes to translate these portion sizes into grams, using the manual ‘Maten en gewichten (Measures and weights)’ for standardized quantification of food items.31
The days of the week that needed to be registered were allocated randomly so that both week and weekend days were included.
Schoolteachers of classes included in the study, received clear instructions for the recording of foods and drinks consumed during schooldays (e.g. snacking and lunches). Teachers were asked to pass this information on to the parents so that they could include the foods and drinks consumed at school in the diary of their child. Only good-quality food diaries, containing sufficiently detailed descriptions of the food products and portion sizes consumed, were included in the analysis. The quality control procedure has been described in more detail previously.30 After the quality checks, the remaining diaries were coded and entered in a ‘Diet Entry & Storage’ programme (Becel 32) that automatically calculated the energy and nutrient intakes.
The fieldwork of this study was carried out from October 2002 until February 2003. The Ethical Committee of the Ghent University Hospital granted ethical approval for the study.
Statistical methods
Data management was done using SPSS, Version 14.0 and data analysis was done using SAS Version 9.1. Estimates of the number of record days required for accurately estimating usual nutrient intakes rely on distributions of intake not departing substantially from normal. Although Kolmogorov's D-statistic for normality is not technically appropriate for repeated measures (when data are not independent),33 it was calculated and used as a guide to evaluate skewness from histograms and normal probability plots. Loge-transformed data are reported for nutrients with a D-statistic >0.08, while untransformed data are reported for all nutrients. However, untransformed and loge-transformed data were both calculated for all nutrients in order to evaluate the effect of loge transformations.
Dependent variables were intakes of energy and 22 nutrients, derived from food and drinks intakes. Nutrient intakes derived from dietary supplement intakes have not been taken into account in these analyses. A mixed-effects regression model with a restricted maximum likelihood estimator was used to estimate mean intake and SEM, and within- and between-person variation, Sw (square root of estimated within-person variance), and Sb (square root of estimated between-person variance), stratified by sex or age group. In our models, we controlled for the random effects of the cluster design (school and class) and for the fixed effects of parental education level, day of the week and sequence of the interview. Furthermore, we controlled for gender when stratifying by age and for age when stratifying by gender. The effect of including the covariates was to reduce the observed size of the between-person variation effect.
Variance ratios of within- and between-person variance were expressed as 
. Coefficients of variation (CVs) were calculated as: CVw = [Sw/mean intake (nutrient)] x 100; CVb = [Sb/mean intake (nutrient)] x 100. Calculation of the number of days needed to correctly classify individuals is based on a hypothetical correlation coefficient (r) between observed and actual intakes. For 90% confidence in which 80% of individuals are classified correctly into thirds of a distribution and <1% grossly misclassified, r must be at least 0.9.10,22 The number of days (D) of diet records needed to obtain the required r of 0.9 between observed- and true-nutrient intakes was calculated using the following formula from Black et al.:22
|
|
|
Results |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
In total, 1052 of the 2095 invited pre-schoolers returned a diary (Participation rate = 50%), of whom 1026 remained included after the quality check. From these 1026 diaries, 696 contained three good-quality days, 208 two days and 122 only one day. For the analyses reported in this article, only the 696 3-day EDRs have been used, since a minimum of three consecutive days is required for calculating within-subject variation.34 Since age and/or gender information was missing for 35 children, only 661 children (338 boys; 323 girls) were included in analysis among age groups.
Characteristics of study population
Some characteristics of the study population are given in table 1. Boys and girls were almost equally represented. In total, 30 and 70% of the children were, respectively 2.5–3 years and 4–6.5 years old. Less than 3% of the children were on a medical diet and <2% followed a special eating pattern (e.g. vegetarian). More characteristics have been described elsewhere.35
|
Intake distributions
Although boys and girls differed in magnitude of within- and between-person variation and variance ratios for the different nutrients, the directions of these gender differences were not consistent. When comparing the youngest and oldest pre-schoolers, however, consistent differences in variances and mean intakes were found between these two age groups.
Variance components for energy and macronutrients
Mean intakes of the youngest pre-schoolers were for all macronutrients lower than mean intakes of the oldest age group (table 2). Mean intakes of girls were also lower for all macronutrients than mean intakes of boys (table 3). Ratios of within- to between-person variances were generally >1 for macronutrient intakes, with the exception of carbohydrates and water intakes. Between-person variability was for most macronutrients highest in the youngest children, what mostly resulted in lower variance ratios, in comparison with the oldest age group. Also for macronutrients expressed as percentages of energy intake, between-person variability was highest in the youngest age group (table 4). Although the differences for between-person variability among boys and girls were not as consistent as for the two age groups, the between-person variability for most of the macronutrient intakes was highest in girls (tables 3 & 5).
|
|
|
|
Variance components for micronutrients
Overall, micronutrient intakes had smaller variance ratios than macronutrient intakes. Ratios of within- to between-person variances were generally >1 for children between 4 and 6.5 years old (table 6). However, for micronutrient intakes in the youngest children, variance ratios were generally <1, except for zinc and sodium intakes. Higher between-person variability was mainly responsible for these lower variance ratios in the youngest children. No consistent differences in variances were found between boys and girls (table 7).
|
|
Days required for calculating habitual intakes
For most intake distributions not significantly different from normal, D based on loge-transformed data was within 1 day of the untransformed values. D-values for nutrients with non-normally distributed intakes were, as expected, much lower when the loge-transformed data were used (except for vitamin D and vitamin C, for which loge-transformed data gave D-values higher than the untransformed data). For cholesterol intakes in the youngest age group, for instance, the number of days required decreased from 38 to 21 days after loge transformation.
Days required for energy and macronutrient intakes
For assessment of energy and macronutrient intakes (except from fat, fatty acids and cholesterol intakes), 7-day diet records would be adequate to achieve r 0.9 if all dietary components reported here required simultaneous assessment (table 2). However, specific macronutrients (like carbohydrates) may require fewer days of recording, while fat, fatty acid and cholesterol intakes require much more replicates (up to 18 for cholesterol in the oldest age group, when using loge-transformed data). For most macronutrients, more record days are required for the oldest children, in comparison with the youngest age group (except for complex carbohydrates).
Days required for micronutrient intakes
For children aged 4–6.5 years, at least 4 days are required for estimating micronutrient intakes, while for children aged 2.5–3 years, 4 days should be sufficient for most micronutrients (except for sodium and zinc intakes) (table 6).
Overall, 7-day dietary records were found to be sufficient for accurately estimating 16 of the 23 nutrients reported in both age groups and for 15 nutrients in both sexes.
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Main results
Overall, micronutrient intakes had lower variance ratios and days required (D) than macronutrient intakes and except from carbohydrate, water and different micronutrient intakes, variance ratios were >1. When comparing the youngest pre-school children with the older ones, in general, the youngest group had lower variance ratios than the oldest group of children for almost all nutrients under study. The highest within- and between-variance ratios (and accordingly also the number of days required) were found for cholesterol intakes, followed by fat and fatty acid intakes, zinc and sodium intakes.
Methodological issues and limitations
Like any dietary assessment method, EDRs are prone to a degree of misreporting what may have influenced our mean intakes and variance components. In addition, parents/caregivers were used as a proxy for recording children's diet intakes. However, as described previously by Huybrechts and De Henauw30, great efforts were made in this study in order to avoid misreporting as much as possible. In addition, the percentage of under-reporters in our study sample was very low when using Goldberg's cut-offs adapted for children (0% for youngest group; <2% for oldest group).30
Although all days of the week were almost equally represented, autumn and winter were the only two seasons included in the study. No data were found about potential seasonal influences on nutrient intakes in this population group in Belgium. However, from the Belgian National Food Consumption Survey in 2004, it could be concluded that seasonal variations were only limited for nutrient intakes in our Belgian population at least 15 years old,36 possibly due to the widespread availability of most foods year round. The representativeness of our study was described in more detail in a previous paper,30 showing good geographical representativeness. Though some selection bias can be expected for what concerns socioeconomic status of the parents, since the group of parents participating in the study was higher educated than the general population aged 25–35 years living in Flanders in 2002.30
Furthermore, comparison with a larger sample of Flemish pre-schoolers who did not complete the 3-day EDR revealed that the proportion of children consuming a hot lunch at school was much lower in this sample of children included in the analyses of 3-day EDR (only 16% compared with 24% in larger sample). These results might point to the difficulties that the parents had to report the lunches consumed at school with sufficient accuracy. The fact that children who consumed a lunch at school were underrepresented in our study sample, might indicate some selection bias as well since their eating habits might be different compared with children who do not consume their lunch at school.
It is noteworthy that the cluster design with schools as primary sampling units and classes as secondary sampling units could have an effect on the between-subject variability since children from a same school and/or class might be eating more homogeneous diet than children from different schools/classes (e.g. some children could be using the same lunch in a school canteen). Therefore, the authors corrected for the cluster design in the analysis.
Lastly, it should be underlined that nutrient intakes derived from dietary supplements were not considered in these analyses. Thus, the results presented in this article reflect only the within- and between-person variation derived from foods and drinks consumed by the Flemish pre-schoolers.
Comparison with other studies
When comparing our Flemish pre-schoolers’ variance ratios with those of toddlers (1–4 years old) and children (5–17 years old) participating in a study in Cambridge in 1983 and reported by Nelson et al.10, important similarities were found. Like in our study, Nelson et al. found variance ratios generally >1, with the exception of some micronutrient intakes and their variance ratios were also lower for toddlers (1–4 years old) than for older children.10 Although these lower variance ratios found in younger children could imply a more monotonous diet of preferred foods in this youngest group, when looking at CVb, these lower ratios were more attributable to important between-person variability in these younger children. This pattern suggests mainly a greater heterogeneity in this youngest group, which may be attributable to higher parental influences in these younger children just starting pre-school classes, than in older children who are already more influenced by their friends instead of the parents and so creating a more homogeneous diet in this older age group.
When comparing variance ratios relatively over the different nutrients, Nelson et al. also found the highest variance ratios and number of days required for cholesterol intakes, followed by fat and fatty acid intakes among toddlers and children.10 However, when considering absolute intakes, variance ratios, and consequently also the number of days required, reported by Nelson for fat, fatty acids, cholesterol, zinc and vitamin C intakes in toddlers and children were lower than those reported in our study.10 For energy, carbohydrates, fibre, potassium, thiamine and riboflavin intakes, the absolute variance ratios reported by Nelson et al. were higher than those reported in our study.10 The large difference in age range between the study by Nelson et al. (1–4 years and 5–17 years) and our study (2.5–3 years and 4–6.5 years) might be partly responsible for these absolute differences. Also changes in dietary patterns during the last decades could be partly responsible for these absolute differences.
Our results confirmed the finding reported earlier by Nelson et al. that in general, variance ratios are lowest for nutrients appearing regularly in diets of some subjects, but not in others (e.g. sugars and vitamins: CVw low, while CVb high) and highest for nutrients appearing in large amounts only occasionally in almost all subjects’ diets (e.g. cholesterol: CVw high, while CVb low).10 If we consider fruits, for instance, which are typically high in vitamin concentrations and often consumed regularly by some people while only rarely by other people, it is clear that this high between-person variability and rather low within-person variability in fruit intake will result in, respectively high and low between- and within-person variability for vitamin intakes as well. Furthermore, can the extreme large variance ratios for cholesterol intakes, which are found in most studies investigating variance ratios (no matter what study population) possibly be explained by the high cholesterol concentration of food products that are not consumed regularly (or not on a daily basis) such as eggs and organs.
It is noteworthy that the exceptional finding of lower D-values for untransformed data in comparison with transformed data for vitamin C was also in line with the findings from Nelson et al.10
When comparing our variance ratios with those reported by Roma-Giannikou et al.24 for children 2–14 years old, one could notice that for energy and for most macronutrient intakes, the ratios reported by Roma-Giannikou et al. were lower (except for fatty acids and cholesterol even <1) than those reported in our article and by Nelson et al. A possible explanation for these lower variance ratios reported by Roma-Giannikou et al. could be that the large age range in their study (2–14 years old) was responsible for much larger between-person variability, mainly due to important differences in eating habits between toddlers and pre-school children.
When comparing our results with previous work in adults,2,3,10 in general, fewer days of records are required to assess nutrient intakes in pre-school children (<7 days for most nutrients, except fat and fatty acid intakes, cholesterol, zinc and sodium intakes) than in adults (>7 days), which was in line with the previous findings.10,20 This effect may be due to the lack of extreme variability in young children's diets, which are based on fewer foods than are consumed in adulthood.
Relevance and implications of our findings
Since most variance ratios are 1 for children at least 4 years old, it may be more difficult to show changes or find epidemiological associations with health outcomes among these older children because associations are difficult to detect when the ratio exceeds unity.37 In addition, for nutrients with high variance ratios (like cholesterol and sodium), attenuated linear regression estimates of diet–health associations in children could hide important associations when the number of replicates is only limited.
Potential errors caused by using an insufficient number of days has been studied in more detail by Liu et al.38 Given the inherent variation in intake and the large range of studies only using a limited number of replicates without appropriate adjustments, it is not surprising that many investigations have failed to report diet–disease associations. In light of the present results, one should be sceptical of such negative conclusions and even more when it concerns intakes of nutrients like cholesterol, fatty acids and sodium.
In addition, Dietary Reference Intakes, used to evaluate group intakes require usual intake distributions, which cannot be calculated when only a single food record or 24 h recall is collected. If an unadjusted distribution is used, bias can be considerable in the estimate of the prevalence of inadequate intakes. Jahns et al.4,39,40 have discussed these potential errors in detail and indicated that bias in prevalence estimates of inadequate intakes could be considerable. For example, among 14- to 18-year old girls from the Continuing Survey of Food Intakes by Individuals (CSFII), the estimated prevalence of inadequate intakes using 1 day of data was 35 and 54% for thiamine and vitamin C, respectively. When the adjusted intake distribution was used in the calculation, however, those estimates decreased to 14 and 31%, respectively, indicating that the bias resulting from using only one 24 h recall without appropriate adjustments was 150% in the case of thiamine and 69% in the case of vitamin C.4,39,40 Though, less than optimal, variance estimates can be borrowed from another population, when only one record/recall day is available and study-specific variance estimates (e.g. from a sub-sample) are lacking.4 However, our findings suggest that one should be cautious when attempting to do so. This mainly, since the results of our current study demonstrate important differences for variance components between age categories. Like the use of unadjusted distributions, adjustments, using variance estimates that are too high or too low may contribute to biased prevalence estimates as well. So, when only one record day is available and variance estimates need to be borrowed from another population, authors are recommended to search for studies reporting variance estimates from a population that fits as good as possible their own study population (e.g. in age group, gender and culture).
|
Conclusions |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
The number of record days required for reliably classifying pre-school children raises with increasing age for most nutrients and varies from 3 or 4 days for some nutrients like carbohydrates to 2 or 3 weeks for others like dietary cholesterol or monounsaturated fatty acids. For studies aiming to evaluate the populations’ usual diet intakes in comparison with recommendations using one record day only, variance estimates should be borrowed from a population that fits as good as possible their own study population. Therefore, the variances reported in this article can be used by other dietary studies in pre-school children where study-specific day-to-day variability is lacking.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
We thank Mia Bellemans and Mieke De Maeyer, the dieticians of our team, for the data input. We are also extremely grateful to all the parents and teachers who participated into this project and generously volunteered their time and knowledge. Funding for this project was provided by the Belgian Nutrition Information Center.
Conflicts of interest: None declared.
Key points
|
|
References |
|---|
|
TopAbstractIntroductionSubjects and methodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
1 Kohlmeier L. Future of dietary exposure assessment. Am J Clin Nutr (1995) 61(3 Suppl):702S–9S.[Medline]
2 Beaton GH, Milner J, Corey P, et al. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr (1979) 32:2546–59.
3 Beaton GH, Milner J, McGuire V, et al. 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 (1983) 37:986–95.
4 Jahns L, Carriquiry A, Arab L, et al. Within- and between-person variation in nutrient intakes of Russian and U.S. children differs by sex and age. J Nutr (2004) 134:3114–20.
5 Carriquiry AL. Estimation of usual intake distributions of nutrients and foods. J Nutr (2003) 133:601S–8S.
6 Hoffmann K, Boeing H, Dufour A, et al. Estimating the distribution of usual dietary intake by short-term measurements. Eur J Clin Nutr (2002) 56(Suppl 2):S53–62.[Web of Science][Medline]
7 Chang HY, Suchindran CM, Pan WH. Using the overdispersed exponential family to estimate the distribution of usual daily intakes of people aged between 18 and 28 in Taiwan. Stat Med (2001) 20:2337–50.[CrossRef][Web of Science][Medline]
8 Borrelli R, Cole TJ, Di Biase G, Contaldo F. Some statistical considerations on dietary assessment methods. Eur J Clin Nutr (1989) 43:453–63.[Web of Science][Medline]
9 Mennen LI, Bertrais S, Galan P, et al. The use of computerised 24 h dietary recalls in the French SU.VI.MAX Study: number of recalls required. Eur J Clin Nutr (2002) 56:659–65.[CrossRef][Web of Science][Medline]
10 Nelson M, Black AE, Morris JA, Cole TJ. Between- and within-subject variation in nutrient intake from infancy to old age: estimating the number of days required to rank dietary intakes with desired precision. Am J Clin Nutr (1989) 50:155–67.
11 Neuhaus JM, Murphy SP, Davis MA. Age and sex differences in variation of nutrient intakes among U.S. adults. Epidemiology (1991) 2:447–50.[Web of Science][Medline]
12 Nyambose J, Koski KG, Tucker KL. High intra/interindividual variance ratios for energy and nutrient intakes of pregnant women in rural Malawi show that many days are required to estimate usual intake. J Nutr (2002) 132:1313–18.
13 Ogawa K, Tsubono Y, Nishino Y, et al. Dietary sources of nutrient consumption in a rural Japanese population. J Epidemiol (2002) 12:1–8.[Web of Science][Medline]
14 Oh SY, Hong MH. Within- and between-person variation of nutrient intakes of older people in Korea. Eur J Clin Nutr (1999) 53:625–29.[CrossRef][Web of Science][Medline]
15 Palaniappan U, Cue RI, Payette H, Gray-Donald K. Implications of day-to-day variability on measurements of usual food and nutrient intakes. J Nutr (2003) 133:232–35.
16 Persson V, Winkvist A, Ninuk T, et al. Variability in nutrient intakes among pregnant women in Indonesia: implications for the design of epidemiological studies using the 24-h recall method. J Nutr (2001) 131:325–30.
17 Tarasuk V, Beaton GH. Day-to-day variation in energy and nutrient intake: evidence of individuality in eating behaviour? Appetite (1992) 18:43–54.[CrossRef][Web of Science][Medline]
18 Tokudome Y, Imaeda N, Nagaya T, et al. Daily, weekly, seasonal, within- and between-individual variation in nutrient intake according to four season consecutive 7 day weighed diet records in Japanese female dietitians. J Epidemiol (2002) 12:85–92.[Medline]
19 Willett WC. Nutritional epidemiology (1998) 2nd edn. New York, NY: Oxford University Press.
20 Lanigan JA, Wells JC, Lawson MS, et al. Number of days needed to assess energy and nutrient intake in infants and young children between 6 months and 2 years of age. Eur J Clin Nutr (2004) 58:745–50.[CrossRef][Web of Science][Medline]
21 Miller JZ, Kimes T, Hui S, et al. Nutrient intake variability in a pediatric population: implications for study design. J Nutr (1991) 121:265–74.
22 Black AE, Cole TJ, Wiles SJ, White F. Daily variation in food intake of infants from 2 to 18 months. Hum Nutr Appl Nutr (1983) 37:448–58.[Web of Science][Medline]
23 Field AE, Peterson KE, Gortmaker SL, et al. Reproducibility and validity of a food frequency questionnaire among fourth to seventh grade inner-city school children: implications of age and day-to-day variation in dietary intake. Public Health Nutr (1999) 2:293–300.[Medline]
24 Roma-Giannikou E, Adamidis D, Gianniou M, Nikolara R, Matsaniotis N. Nutritional survey in Greek children: nutrient intake. Eur J Clin Nutr (1997) 51:273–85.[CrossRef][Web of Science][Medline]
25 Serdula MK, Alexander MP, Scanlon KS, Bowman BA. What are preschool children eating? A review of dietary assessment. Annu Rev Nutr (2001) 21:475–98.[CrossRef][Web of Science][Medline]
26 CEC. Nutrient and energy intakes for the European Community. (Thirty-first series). Reports of the Scientific Committee for Food. Commission of the European Communities, Food-science and Techniques, Luxembourg, 1993.
27 Huybrechts I, De Henauw S. Energy and nutrient intakes by pre-school children in Flanders-Belgium. Br J Nutr (2007) 98:600–10.[CrossRef][Web of Science][Medline]
28 Gezondheidsraad Hoge. Dietary recommendations for Belgium, revised version 2003 (Voedingsaanbevelingen voor België: herziene versie 2003). (2003) Brussels: Hoge Gezondheidsraad.
29 Iowa State University. C-side. (2006) http://www.cssm.iastate.edu/software/cside.html. (Accessed 22 March 2008).
30 Huybrechts I, De Henauw S. Energy and nutrient intakes by pre-school children in Flanders-Belgium. Br J Nutr (2007) 98:600–10.[CrossRef][Web of Science][Medline]
31 De Backer G. Standard measures and weights. Manual for standardised quantification of foods in Belgium, revision January 2005. Brussels: The Superior Health Council, Federal Public Service Health, Food Chain Safety and Environment.
32 Unilever. Becel Nutrient Calculation Program. (1995) Rotterdam: Nederlandse Unilever Bedrijven B.V.
33 Steel R, Torrie JH. Principles and procedures of statistics: a biometrical approach (1980) 2nd edn. New York, NY: McGraw-Hill.
34 Guenther PM, Kott PS, Carriquiry AL. Development of an approach for estimating usual nutrient intake distributions at the population level. J Nutr (1997) 127:1106–12.
35 Huybrechts I, Matthys C, Pynaert I, et al. Flanders preschool dietary survey: rationale, aims, design, methodology and population characteristics. Accepted in Archives of Public Health (in press).
36 De Vriese S, Huybrechts I, Moreau M, Van Oyen H. The Belgian Food Consumption Survey 1 - 2004: Report (Enquête de consommation alimentaire Belge 1 - 2004: Rapport). (2006) D/2006/2505/16.
37 Paeratakul S, Popkin BM, Kohlmeier L, et al. Measurement error in dietary data: implications for the epidemiologic study of the diet-disease relationship. Eur J Clin Nutr (1998) 52:722–27.[CrossRef][Web of Science][Medline]
38 Liu K, Stamler J, Dyer A, et al. Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol. J Chronic Dis (1978) 31:399–418.[CrossRef][Web of Science][Medline]
39 Jahns L, Arab L, Carriquiry A. Dietary reference intakes still used incorrectly in journal articles. J Am Diet Assoc (2003) 103:1292–93.[CrossRef][Web of Science][Medline]
40 Jahns L, Arab L, Carriquiry A, Popkin BM. The use of external within-person variance estimates to adjust nutrient intake distributions over time and across populations. Public Health Nutr (2005) 8:69–76.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||