Hostname: page-component-77c89778f8-m42fx Total loading time: 0 Render date: 2024-07-22T10:01:15.919Z Has data issue: false hasContentIssue false

Prevalence and determinants of misreporting among European children in proxy-reported 24 h dietary recalls

Published online by Cambridge University Press:  06 August 2012

C. Börnhorst
Affiliation:
BIPS – Institute for Epidemiology and Prevention Research, Achterstraße 30, 28359Bremen, Germany
I. Huybrechts
Affiliation:
Department of Public Health, Ghent University, 2 BlokA De Pintelaan 185, 9000Ghent, Belgium Dietary Exposure Assessment Groups, International Agency for Research on Cancer, Lyon, France
W. Ahrens
Affiliation:
BIPS – Institute for Epidemiology and Prevention Research, Achterstraße 30, 28359Bremen, Germany
G. Eiben
Affiliation:
Department of Public Health and Community Medicine, University of Gothenburg, Box 454, 40530Gothenburg, Sweden
N. Michels
Affiliation:
Department of Public Health, Ghent University, 2 BlokA De Pintelaan 185, 9000Ghent, Belgium
V. Pala
Affiliation:
Nutritional Epidemiology Unit, Department of Preventive and Predictive Medicine, Fondazione IRCSS Istituto Nazionale dei Tumori, Via Venezian 1, 20133Milan, Italy
D. Molnár
Affiliation:
Department of Pediatrics, Medical Faculty, University of Pécs, József A.u.7 H-7623, Pécs, Hungary
P. Russo
Affiliation:
Institute of Food Sciences, CNR, Via Roma 64, 83100Avellino, Italy
G. Barba
Affiliation:
Institute of Food Sciences, CNR, Via Roma 64, 83100Avellino, Italy
S. Bel-Serrat
Affiliation:
GENUD (Growth, Exercise, Nutrition and Development) Research Group, Escuela Universitaria de Ciencias de la Salud, Universidad de Zaragoza, Corona de Aragón 42, 2nd floor, 50009Zaragoza, Spain
L. A. Moreno
Affiliation:
GENUD (Growth, Exercise, Nutrition and Development) Research Group, Escuela Universitaria de Ciencias de la Salud, Universidad de Zaragoza, Corona de Aragón 42, 2nd floor, 50009Zaragoza, Spain
S. Papoutsou
Affiliation:
Research and Education Institute of Child Health, 8 Attikis Street, 2027Strovolos, Cyprus
T. Veidebaum
Affiliation:
Center of Health and Behavioral Science, National Institute for Health Development, Hiiu 42, 11619Tallinn, Estonia
H.-M. Loit
Affiliation:
Center of Health and Behavioral Science, National Institute for Health Development, Hiiu 42, 11619Tallinn, Estonia
L. Lissner
Affiliation:
Dietary Exposure Assessment Groups, International Agency for Research on Cancer, Lyon, France
I. Pigeot*
Affiliation:
BIPS – Institute for Epidemiology and Prevention Research, Achterstraße 30, 28359Bremen, Germany
*
*Corresponding author: I. Pigeot, fax +49 421 218 56 941, E-mail: pigeot@bips.uni-bremen.de
Rights & Permissions [Opens in a new window]

Abstract

Dietary assessment is strongly affected by misreporting (both under- and over-reporting), which results in measurement error. Knowledge about misreporting is essential to correctly interpret potentially biased associations between diet and health outcomes. In young children, dietary data mainly rely on proxy respondents but little is known about determinants of misreporting here. The present analysis was conducted within the framework of the multi-centre IDEFICS (Identification and prevention of dietary- and lifestyle-induced health effects in children and infants) study and is based on 6101 children aged 2–9 years with 24 h dietary recall (24-HDR) and complete covariate information. Adapted Goldberg cut-offs were applied to classify the 24-HDR as ‘over-report’, ‘plausible report’ or ‘under-report’. Backward elimination in the course of multi-level logistic regression analyses was conducted to identify factors significantly related to under- and over-reporting. Next to characteristics of the children and parents, social factors and parental concerns/perceptions concerning their child's weight status were considered. Further selective misreporting was addressed, investigating food group intakes commonly perceived as more or less socially desirable. Proportions of under-, plausible and over-reports were 8·0, 88·6 and 3·4 %, respectively. The risk of under-reporting increased with age (OR 1·19, 95 % CI 1·05, 1·83), BMI z-score of the child (OR 1·23, 95 % CI 1·10, 1·37) and household size (OR 1·12, 95 % CI 1·01, 1·25), and was higher in low/medium income groups (OR 1·45, 95 % CI 1·13, 1·86). Over-reporting was negatively associated with BMI z-scores of the child (OR 0·78, 95 % CI 0·69, 0·88) and higher in girls (OR 1·70, 95 % CI 1·27, 2·28). Further social desirability and parental concerns/perceptions seemed to influence the reporting behaviour. Future studies should involve these determinants of misreporting when investigating diet–disease relationships in children to correct for the differential reporting bias.

Type
Full Papers
Copyright
Copyright © The Authors 2012

Due to its low respondent burden and easy application, the 24 h dietary recall (24-HDR) is often the method of choice for short-term assessment of dietary intakes in large epidemiological studies. However, numerous sources of measurement error have been encountered when operating with 24-HDR data. Memory of consumption, estimation of portion sizes, decompositions of mixed dishes (unknown recipes), supplement use as well as instrument-based biases are common problems that researchers are confronted with(Reference Arab and Akbar1). As young children lack the cognitive skills to complete dietary assessments(Reference McPherson, Hoelscher and Alexander2), 24-HDR data in children younger than 7 years old usually rely on proxy reporters, mainly the parents(Reference Livingstone and Robson3). Here additional problems emerge from meals that are not under parental control (e.g. school meals), leading to unintentional misreporting(Reference Eck, Klesges and Hanson4Reference Basch, Shea and Arliss6).

Among these difficulties, biased assessment of energy intake (EI) is often a consequence of intentional under- or over-reporting attributed to specific groups. Anthropometry, for example the actual weight status of the study subject, is a well-known determinant of misreporting(Reference Black and Cole7, Reference Poppitt, Swann and Black8). Age, sex, socio-economic status, psychosocial and behavioural characteristics are further factors that were found to be related to misreporting(Reference Livingstone and Black9Reference Lioret, Touvier and Balin11). The validity of proxy-reported EI might additionally be affected by parental characteristics as well as by psychological factors such as parental perception of their child's weight status(Reference Livingstone and Robson3, Reference McGloin, Livingstone and Greene12Reference Bandini, Cyr and Must14). Further social desirability may result in over-reporting of food items perceived to be healthy while unhealthy/energy-dense food items might be under-reported at the same time(Reference Tooze, Subar and Thompson15, Reference Lissner16). Intentional misreporting introduces differential error that may attenuate or even hide associations between dietary factors and health outcomes, whereas non-differential error may distort such associations in any direction.

Recent validation studies based on the doubly labelled water technique in children have revealed inconsistent results concerning misreporting (under-reporting from 19 to 41 %; over-reporting from 7 to 11 % of reported EI)(Reference Burrows, Martin and Collins17) where data relied mainly on self-reports – partially with parental assistance.

Whether the accuracy of proxy reports is comparable to that of self-reports and whether determinants of misreporting coincide for self- v. proxy-reports is yet unknown. Several studies have only addressed under-reporting; the nature and extent of over-reporting have rarely been addressed in young populations(Reference Maurer, Taren and Teixeira10, Reference Forrestal18, Reference Poslusna, Ruprich and de Vries19). Knowing the degree and direction of misreporting is essential for the assessment of diet–disease relationships as well as for the evaluation of dietary guidelines and nutrition policies. Therefore, the present study aimed to investigate the prevalence and determinants of misreporting (including under- and over-reporting) in a large sample of European children.

Methods

Study population

The IDEFICS (Identification and prevention of dietary- and lifestyle-induced health effects in children and infants) study is a multi-centre setting-based study aiming to prevent and investigate the causes of diet- and lifestyle-related diseases such as overweight and obesity in 2–9-year-old European children. The baseline survey was conducted from September 2007 until June 2008; more than 31 500 children were invited, out of whom, finally, 16 220 participated and fulfilled the inclusion criteria of the IDEFICS study. Details on the design and objectives of the study have been given elsewhere(Reference Ahrens, Bammann and de Henauw20Reference Ahrens, Bammann and Siani22). Briefly, children were recruited through schools/kindergartens. Interviews with parents concerning lifestyle habits and dietary intakes as well as anthropometric measurements and examinations of the children were included in the survey. Biomarker information was collected via blood, urine and saliva samples. All measurements were conducted using standardised procedures by all eight centres participating in the study (Italy, Estonia, Cyprus, Belgium, Sweden, Germany, Hungary and Spain).

The present study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the local ethics committees in each participating country. Parents provided written informed consent for all examinations and for the collection of blood, urine, saliva samples as well as subsequent analysis and storage. Each child was informed orally about the modules by fieldworkers and asked for its consent immediately before examination(Reference Ahrens, Bammann and Siani22). Verbal consent was witnessed and formally recorded.

Dietary data

Dietary data were assessed using the computerised 24-HDR ‘SACINA’ (Self-Administered Children and Infants Nutrition Assessment) based on the previously designed and validated YANA-C (‘Young Adolescents’ Nutrition Assessment on Computer’) developed for Flemish adolescents and further adapted to European adolescents in the HELENA (Healthy Lifestyle in Europe by Nutrition in Adolescence) study(Reference Vereecken, Covents and Sichert-Hellert23, Reference Vereecken, Covents and Matthys24). The SACINA is structured according to six meal occasions (breakfast, morning snack, lunch, afternoon snack, dinner and evening snack) embedded in questions related to a range of chronological daily activities. Proxies, mainly the parents, completed the 24-HDR under the supervision of fieldwork personnel in about 20–30 min. Except for Cyprus, where school ends at 13.00 hours, school meals were additionally assessed by means of direct observation. Teachers and school kitchen staff were interviewed by trained survey personnel and data were documented using special documentation sheets including portion sizes. School meal data were merged with the parentally reported 24-HDR data to enhance completeness of dietary intakes. The 24-HDR were assessed on non-consecutive days over the whole week and over the complete IDEFICS assessment period. The assessment procedure in Hungary slightly differed from the other study centres. Here dietary information was recorded on documentation sheets and entered into the SACINA program afterwards.

The uniquely coded food items were linked to country-specific food composition tables. Missing quantities for single food items as well as obviously implausible data entries were imputed by country, food group and age-specific median intakes (0·01 % of the entries) to avoid excessive record exclusions. Although up to six repeated 24-HDR were carried out in a smaller sample, only the first recall day was included in the present analysis (including weekdays and weekend days) to obtain an equal number of 24-HDR for each child and to achieve an adequate statistical power for a cross-country analysis. Incomplete interviews were excluded, for example if the proxy did not know about at least one main meal or in case of missing school meal information. Further, intakes of more than 16 736 kJ/d (4000 kcal/d) were excluded (n 10).

Anthropometry

Height (cm) of the children was measured to the nearest 0·1 cm with a calibrated stadiometer (model: telescopic height measuring instrument SECA 225 Stadiometer; SECA). Body weight (kg) was measured in the fasting state in light underwear on a calibrated scale accurate to 0·1 kg (model: electronic scale TANITA BC 420 SMA with an adapter; Tarita Europe GmbH).

Covariables

A set of covariables previously found to be related to misreporting in adults or expected to be relevant in children(Reference Livingstone and Black9, Reference Lioret, Touvier and Balin11, Reference Vereecken, Covents and Sichert-Hellert23, Reference Vereecken, Covents and Matthys24) was defined to explore the determinants of misreporting in this young study population: age, sex, BMI z-scores according to Cole et al. (Reference Cole, Freeman and Preece25, Reference Cole, Freeman and Preece26) and average audio-visual media time (h/week) of the child, age, sex and self-reported BMI of the proxy, educational level (maximum of both parents, dummy: high v. medium/low) according to the International Standard Classification of Education (ISCED 1997, UNESCO Institute for Statistics: Montreal, 2006), net household income (dummy: high v. medium/low) and number of persons below 18 years of age in the household as indicators for socio-economic status, the interview day (dummy: weekday v. Saturday/Sunday), assessment of a school meal (dummy: yes v. no) and the use of a day-care service or babysitter (dummy: yes v. no) were considered.

The following information on parental concerns/perceptions of their child's weight status was included where the questions were obtained from previously validated questionnaires(Reference Birch, Fisher and Grimm-Thomas27, Reference Mensink, Kleiser and Richter28) and slightly modified for use in IDEFICS: ‘How concerned are you about your child’ – (1) ‘eating too much when you are not around him/her?’; (2) ‘having to diet to maintain desirable weight?’; (3) ‘becoming overweight?’; (4) ‘becoming underweight?’ (answer categories: ‘Unconcerned’, ‘A little concerned’, ‘Concerned’ or ‘Very concerned’); ‘Do you think your child is’ – (1) ‘Much too underweight?’; (2) ‘Slightly too underweight?’; (3) ‘Proper weight?’; (4) ‘Slightly too overweight’; (5) ‘Much too overweight?’ (answer categories: ‘Yes’ or ‘No’). The rationale behind this was the assumption of parental concerns/perceptions being associated with misreporting. Furthermore, the question ‘Do you sit down with your child when he/she eats meals?’ (answer categories: ‘Never’, ‘Rarely’, ‘Sometimes’, ‘Often’ or ‘Always’) was included as an indicator for family meal behaviours.

To investigate the degree to which given answers were influenced by social desirability, intakes of the following food items commonly perceived to be healthy/unhealthy were included as predictors for misreporting in a second step: chocolate products; other sugary products (e.g. cakes, biscuits, ice cream); carbonated soft drinks; fruits/vegetables; milk (all as the percentage from total EI per d); water (g/d).

Statistical methods

BMR was estimated using the equations published by Schofield(Reference Schofield29) and recommended by the FAO/WHO/UNU (1985) taking into account age, sex, body height and weight. Goldberg et al. (Reference Goldberg, Black and Jebb30) defined cut-off values to classify the 24-HDR in under-reports (UNR), plausible reports (PLR) and over-reports (OVR), respectively. The cut-offs make allowance for the errors associated with the duration of dietary assessment (number of recall days), the sample size as well as the variation in BMR, physical activity (PA) level (PAL) and EI. Minimum/maximum plausible levels of EI are defined as multiples of BMR. Since these cut-offs were developed for adults without considering differences in EI due to age and sex, adaptations are required for application in children.

Upper and lower cut-off values to identify plausible/implausible reports of EI were calculated substituting Goldberg's single cut-off 2(Reference Goldberg, Black and Jebb30) by age- and sex-specific cut-offs for children, as suggested previously(Reference Lioret, Touvier and Balin11, Reference Sichert-Hellert, Kersting and Schoch31), using the following formula:

$$\begin{eqnarray} Cut\hyphen off = PAL\times exp\left [\pm 1\cdot 96\times \frac {( S /100)}{\sqrt { n }}\right ]. \end{eqnarray}$$

where

$$\begin{eqnarray} S = \sqrt {\frac { CV _{wEI}^{2}}{ d } + CV _{wBMR}^{2} + CV _{PA}^{2}.} \end{eqnarray}$$

The within-subject CV for EI (CVwEI), the within-subject CV for BMR (CVwBMR) and the CV for PA (CVPA) were replaced by age- and sex-specific reference values as given in Nelson et al. (Reference Nelson, Black and Morris32) and Black(Reference Black33). The number of days (d) was set to one as the analysis is based on one 24-HDR per child. Goldberg's overall level of 1·55 for PA was substituted by age- and sex-dependent levels of light PA (2–5 years: 1·45; 6–10 years: males 1·55, females 1·50) according to Torun et al. (Reference Torun, Davies and Livingstone34). All reference values used are summarised in Table 1. The resulting age- and sex-specific cut-off values to define UNR, PLR and OVR are given in Table 2. Records were classified as UNR, PLR and OVR according to the recalculated cut-off values.

Table 1 Reference values to recalculate the Goldberg cut-offs for application in children

EI, energy intake; PAL, physical activity level.

* Within-subject CV of energy intake; values obtained from Nelson et al. (Reference Nelson, Black and Morris32).

Within-subject CV of BMR; values obtained from Black(Reference Black33).

PAL; values obtained from Torun et al. (Reference Torun, Davies and Livingstone34).

§ CV of PAL; values obtained from Black(Reference Black33).

Table 2 Lower and upper cut-off limits to classify 1 d 24 h dietary recalls (24-HDR) as under-, plausible and over-reports based on the ratio of energy intake (EI*):BMR

* EI estimated from 24-HDR.

BMR estimated from Schofield equations(Reference Schofield29).

Multi-level logistic regression analysis was conducted to identify factors statistically significantly associated with misreporting. Determinants for UNR and OVR were investigated in separate models (model 1a: outcome UNR, reference PLR; model 2a: outcome OVR, reference PLR). In the model addressing UNR, records classified as OVR were excluded and the other way around. All covariables mentioned earlier were entered into the two models except for the dietary variables and the backward selection procedure was applied to screen out the relevant factors. Under this approach, one starts fitting a model that contains all covariables. The least significant one is dropped except if it is significant at the critical level of 0·05. The reduced models are successively refitted applying the same rule until all the remaining variables are statistically significant.

In a next step, the dietary variables were added to the resulting models (including only the relevant covariables now) to investigate their predictive power for misreporting (model 1b: outcome UNR; model 2b: outcome OVR). Random effects for the study centre and setting (schools/kindergartens) were entered in all models to account for the clustered study design.

The present analysis only includes children with 24-HDR and complete covariable information (n 6101).

All analyses were performed using the statistical software package SAS (version 9.1; SAS Institute).

Results

Both the prevalence of UNR (1·2–16·4 %) and OVR (1·5–5·4 %) strongly differed between the study centres (Table 3). UNR was highest in the Hungarian study centre, OVR in the Italian one. UNR and OVR were higher in girls and UNR was higher in older children. Regarding the total study group, 8·0 % of the reports were classified as UNR and 3·4 % as OVR.

Table 3 Prevalence of misreporting by study centre, sex and age group (Total numbers and percentages)

Descriptive statistics of all covariables can be obtained from Tables 4 and 5 stratified by reporting group (UNR, PLR and OVR). The mean BMI of children and their proxies were highest in UNR, whereas the percentage of proxies with a high income or educational level was highest in PLR. In UNR, a higher percentage of proxies were male and the use of day-care services was less frequent. The percentage of recalls assessed on weekends was highest in OVR. Furthermore, proxies of UNR were more likely to perceive their child as overweight/obese and stated more often to be concerned about their child becoming overweight, whereas proxies of OVR were more concerned about their child becoming underweight. Percentages of daily EI from chocolate products and sugary products increased with reporting group (lowest in UNR and highest in OVR), whereas percentages of EI from fruits/vegetables decreased with reporting group (Table 5).

Table 4 Descriptive analysis of categorical covariates by reporting group (Percentages and total numbers)

ISCED, International Standard Classification of Education.

* Low/medium education is defined as ISCED levels 1–3; high education is defined as ISCED levels 4 and 5 (ISCED 1997, UNESCO Institute for Statistics: Montreal, 2006).

Days without school meals relate either to weekend days or to working days where the child had no lunch or lunch at home.

Table 5 Descriptive analysis of continuous covariates and dietary intakes by reporting group (Mean values and standard deviations)

EI, energy intake.

Application of the backward selection procedure including all covariables except the dietary ones revealed that different factors were significantly associated with UNR compared with the model addressing OVR (models 1a and 2a; Table 6). The risk of UNR increased with age (OR 1·19, 95 % CI 1·11, 1·27), BMI z-score of the child (OR 1·23, 95 % CI 1·10, 1·37), the number of persons below 18 years of age in the household (OR 1·12, 95 % CI 1·01, 1·25) and was higher in the low/medium income group (OR 1·45, 95 % CI 1·13, 1·86; reference: high income group) as well as on interview days without additional school meal assessment (OR 1·58, 95 % CI 1·17, 2·13). Sitting always (OR 0·61, 95 % CI 0·43, 0·85) or often down while eating (OR 0·62, 95 % CI 0·44, 0·87; reference: sitting sometimes down while eating) turned out to be negatively associated with UNR. Proxies perceiving their child as slightly (OR 1·63, 95 % CI 1·03, 2·56) or much too overweight (OR 3·30, 95 % CI 1·51, 7·18; reference: slightly too underweight) were more likely to under-report. On the other hand, OVR was higher in female children (OR 1·70, 95 % CI 1·27, 2·28; reference: male children). BMI z-scores of children (OR 0·78, 95 % CI 0·69, 0·88) were negatively associated with OVR. Being very concerned about the child becoming overweight (OR 0·44, 95 % CI 0·23, 0·84) decreased the risk for OVR, whereas being very concerned about the child becoming underweight increased the risk (OR 1·77, 95 % CI 1·10, 2·85).

Table 6 Results of the multi-level logistic regression applying backward selection: factors significantly associated with under-reports/over-reports (models 1a and 2a) and predictive value of selected food items for misreporting (models 1b and 2b)* (Odds ratios and 95 % confidence intervals)

EI, energy intake.

* All models include random effects for study centre and setting.

Effects of continuous variables are assessed as one unit offsets from the mean.

Days without school meals relate either to weekend days or to working days where the child had no lunch or lunch at home.

Adding the dietary variables to the models showed that percentages of total EI from chocolate products, soft drinks and sugary products were negatively associated with the risk of UNR, whereas percentages of EI from fruits/vegetables were positively associated (model 1b). The OR for the other covariables changed only slightly in model 1b compared with model 1a. Inclusion of the dietary variables in model 2a revealed no significant associations between the percentages of EI and OVR for any of the considered food items except for milk (OR 0·97, 95 % CI 0·95, 0·98) and sugary products (OR 1·01, 95 % CI 1·00, 1·02).

Discussion

In general, proportions reported for UNR and OVR vary widely between publications (UNR 2–85 % and OVR 3–46 %, obtained from a current review including children and adolescents(Reference Forrestal18)) where the proportion of UNR is usually higher than that of OVR. The proportions of UNR (8·0 %) and OVR (3·4 %) found in the present study sample are difficult to compare with other studies due to differences in age groups, number of assessment days, cut-off values applied and the respondent status (self v. proxy). The relatively low proportions of UNR and OVR in the present data may be a consequence of cooperation with parents/caregivers, which has been shown to be associated with a lower risk of UNR/OVR previously(Reference Kimm, Glynn and Obarzanek35). The present data revealed a decreased risk of UNR on days with additional school meal assessment (days without school meal assessment relate either to weekend days or to working days where the child had no lunch or lunch at home). Lioret et al. (Reference Lioret, Touvier and Balin11) reported proportions for UNR and OVR of 4·9 and 1·4 %, respectively, in 3–10-year-old French children. This study was similar to the present one in terms of sample size, cut-off values and instruments applied. In the study by Murakami et al. (Reference Murakami, Miyake and Sasaki36), UNR ranged from 2·9 to 28·0 % and OVR from 3·0 to 28·1 % depending on the considered age group (children aged 6–15 years, stratified by 1-year age groups). UNR increased and OVR decreased with age, which agrees with the present results (Table 3). The increase in UNR with age may be explained by reduced parental control and a higher frequency of out-of-home meals in older children.

The notably high proportion of UNR in the Hungarian study centre (16·4 %) may be a consequence of the slightly different study protocol. As opposed to the paper-based assessment in Hungary, the computerised SACINA program used in the other study centres included reminders for certain foods and already some checks for plausibility. Further pictures with increasing portion sizes were displayed to facilitate the estimation of portion sizes. These differences in the assessment procedure may explain the discrepancy between the proportion of UNR in Hungary and the other study centres. In Cyprus, schools do not offer meals and therefore no additional information on school meals was assessed, which may explain the high percentage of UNR (16·1 %) in this study centre.

Over-reporting was found to be higher in children and adolescents compared with adults, which has been suggested to be rather a consequence of intrusion of foods that were not actually consumed than errors in portion size estimation(Reference Forrestal18, Reference Baxter, Smith and Litaker37). However, these studies relied on self-reports. In proxy reports, over-reporting could be suspected to be either a result of intrusion due to the lack of parental control or a result of overeating due to increased energy requirements during growth. The latter would result in misclassifications of records, for example OVR in spite of correct parental reports. Difficulties such as decreasing metabolic costs of movement during maturation and heavier children spending more energy at the same intensity of PA than peers may further affect classifications of UNR/OVR(Reference Freedson, Pober and Janz38). Moreover, it cannot be precluded that the 24-HDR was assessed on an exceptional day resulting in very high or low reported intakes (for example, child was ill).

Though the mean BMI of the child and of its proxy were both highest in the group of UNR in the descriptive analysis, the multivariate analysis revealed only the BMI of the child being significantly associated with misreporting. It is likely that similar dietary patterns within a family as well as shared genetic/environmental factors lead to correlations between parental and children's BMI(Reference Svensson, Jacobsson and Fredriksson39), which may explain the bivariate association between UNR and parental BMI. Previous findings on the association between parental obesity and misreporting are inconsistent in children and adolescents(Reference Livingstone and Robson3, Reference Bandini, Cyr and Must14, Reference Lanctot, Klesges and Stockton40). Nevertheless, the strong association between parental concerns/perceptions of their child's weight status and misreporting rather suggests that, actually, the BMI of the child is the determining factor. To date, no other study has examined parental concerns/perceptions in relation to misreporting of EI by proxy respondents.

UNR was higher in low/medium income groups, whereas educational level was neither found to be associated with UNR nor with OVR. Previous studies in children reported no association between misreporting and income level(Reference Lioret, Touvier and Balin11, Reference Bandini, Cyr and Must14, Reference Lanctot, Klesges and Stockton40). In adults also inconsistent results concerning socio-economic variables have been reported in a review investigating markers on the validity of reported EI(Reference Livingstone and Black9). The authors assumed that poor literacy skills in the less well-educated group and better health and diet consciousness in the better-educated group might both result in misreporting leading to contradictory results. To the authors' knowledge, the effect of household size on misreporting, which may either serve as an indicator for socio-economic status or for parental control, has rarely been addressed in children. The present data suggest a positive relationship with UNR. Opposed to the present results, Vagstrand et al. (Reference Vagstrand, Lindroos and Linne41) found UNR based on self-reports to be more likely in adolescents from one-child families. Nevertheless, in proxy reports, the impact of household size may be different as a high number of children may reduce parental control over each single child's food intakes. Parental control may also explain the effect of the ‘sitting while eating’ variable. Children sitting often or always down while eating had a reduced risk to be classified as UNR.

The present analysis revealed that OVR was higher in girls, whereas UNR was not associated with the sex of the child. It is likely that the determinants of misreporting may differ by sex and also by age group. Unfortunately, stratified analyses were not possible since corresponding models did not converge due to the high number of covariables and the comparably low number of UNR/OVR.

In a literature review mainly relying on self-reports, sex and social desirability have been reported to be consistent predictors for misreporting in adults but not in children and adolescents(Reference Forrestal18). When adding the dietary variables to our models, results pointed to intentional, selective misreporting in the UNR group reflecting socially desirable answer behaviour. Food items commonly perceived to be unhealthy were negatively associated with UNR (chocolate products, sugary products and soft drinks), whereas fruit/vegetable intake showed a positive association. Although the SACINA instrument (retrospective) does not influence the child's eating behaviour, it seems to encourage socially desirable answers of the parents.

Some studies have already applied adapted validation procedures in children substituting Goldberg's single cut-off 2 by individual limits for children(Reference Lioret, Touvier and Balin11, Reference Sichert-Hellert, Kersting and Schoch31, Reference Savage, Mitchell and Smiciklas-Wright42, Reference Sichert-Hellert, Kersting and Manz43). Sichert-Hellert et al. (Reference Sichert-Hellert, Kersting and Schoch31, Reference Sichert-Hellert, Kersting and Manz43), for example, applied recalculated Goldberg cut-offs based on three assessment days in a German sample of 695 children aged 1–18 years but only addressing under-reporting. UNR ranged from 1·2 to 19·2 % depending on the respective age group and was lower compared with the proportions obtained when applying the original Goldberg cut-offs(Reference Goldberg, Black and Jebb30) (1 d cut-off; UNR, EI:BMR ≤ 0·9; OVR, EI:BMR ≥ 2·68). Also in the present study, the recalculated cut-offs revealed a slightly lower proportion of UNR (8·0 v. 8·3 %) and a higher proportion of OVR (3·4 v. 3·1 %) compared with the original Goldberg cut-offs, as expected (data not shown in the tables).

Limitations and strengths

Only one record day per child was used in the present analysis, which does not reflect usual intakes. Black & Cole(Reference Black and Cole7) found that under-reporting is subject-specific, concluding that subjects who under-report on the first 24-HDR accordingly tend to under-report on additionally assessed 24-HDR as well. Therefore, a single 24-HDR can be considered as a reliable instrument for the identification of determinants of misreporting. Nevertheless, an additional analysis was run including only children with at least two 24-HDR. The study sample was markedly reduced (n 6101 v. 1644) and the number of study subjects strongly differed between the study centres (for example, Estonia n 3; Hungary n 828), which resulted in unstable model estimates. This corroborated the decision to include all children with at least one 24-HDR where only the first recall day was used in the analysis.

Sensitivity of the cut-off technique is limited, as it aims only to identify UNR resulting in physiologically implausible low EI(Reference Haraldsdottir and Sandstrom44). By the application of the cut-off technique, varying degrees of misreporting cannot be distinguished, for example under-reporting from a high intake level such that the ratio of EI:BMR does not fall below the cut-off will not be detected. Further cut-off values were calculated assuming light PAL for all children which may result in misclassifications. The likelihood to classify a record as UNR increases with increasing energy expenditure of the child(Reference Livingstone and Black9). As PA is a relevant determinant of energy expenditure, classification into UNR, PLR and OVR should consider individual PAL by applying different cut-off values depending on a child's PAL. Unfortunately, due to the lack of valid PAL information, this approach was not feasible. Moreover, differentiation between undereaters (EI actually lower than energy expenditure) and under-reporters is not possible so that some part of UNR may be attributed to undereaters(Reference Livingstone and Black9). The same applies analogously to overeaters. Future research should include special questions for the identification of low/high eaters.

The large study sample, the additional assessment of school meals and measured anthropometry can be considered as strengths of the present study. Further, the huge number of covariables should be highlighted, as it facilitated a comprehensive analysis of the determinants of misreporting covering various aspects.

Conclusion

Misreports differ from plausible proxy reports with respect to children's characteristics (age, sex and weight status) as well as social factors (number of persons below 18 years of age in household and net household income). Determinants for UNR and OVR only partly agree where UNR seems to be strongly affected by social desirability. Furthermore, parental concerns/perceptions of their child's weight status had a strong impact on misreporting. Researchers should bear this differential reporting bias in mind when investigating diet–disease relationships in children. Identification of influencing factors may help to improve study designs and to interpret potentially biased results.

Acknowledgements

This study was done as part of the IDEFICS study (www.idefics.eu) and is published on behalf of its European Consortium. We gratefully acknowledge the financial support of the European Community within the Sixth Research and Technological Development Framework Programme Contract no. 016181 (FOOD). The information in this paper reflects the authors' view and is provided as is. S. B.-S. is funded by a grant from the Aragón's Regional Government (Diputación General de Aragón). Each author saw and approved the contents of the submitted manuscript. All authors contributed to the conception and design, the acquisition of the data, and the analysis or interpretation of the data, and approved the final version of the manuscript. C. B. drafted the manuscript; W. A., G. E., D. M., L. A. M. and T. V. conceived the study and participated in its design and coordination; I. H. and I. P. helped with the drafting of the manuscript and the interpretation of the data; all the authors revised the article critically for important intellectual content. None of the authors has any conflicts of interest.

References

1Arab, L & Akbar, J (2002) Biomarkers and the measurement of fatty acids. Public Health Nutr 5, 865871.Google Scholar
2McPherson, RS, Hoelscher, DM, Alexander, M, et al. (2000) Dietary assessment methods among school-aged children: validity and reliability. Preventive Med 31, 1133.CrossRefGoogle Scholar
3Livingstone, MB & Robson, PJ (2000) Measurement of dietary intake in children. Proc Nutr Soc 59, 279293.Google Scholar
4Eck, LH, Klesges, RC & Hanson, CL (1989) Recall of a child's intake from one meal: are parents accurate? J Am Diet Assoc 89, 784789.CrossRefGoogle ScholarPubMed
5Baranowski, T, Sprague, D, Baranowski, JH, et al. (1991) Accuracy of maternal dietary recall for preschool children. J Am Diet Assoc 91, 669674.Google Scholar
6Basch, CE, Shea, S, Arliss, R, et al. (1990) Validation of mothers' reports of dietary intake by four to seven year-old children. Am J Public Health 80, 13141317.Google Scholar
7Black, AE & Cole, TJ (2001) Biased over- or under-reporting is characteristic of individuals whether over time or by different assessment methods. J Am Diet Assoc 101, 7080.Google Scholar
8Poppitt, SD, Swann, D, Black, AE, et al. (1998) Assessment of selective under-reporting of food intake by both obese and non-obese women in a metabolic facility. Int J Obes Relat Metab Disord 22, 303311.Google Scholar
9Livingstone, MB & Black, AE (2003) Markers of the validity of reported energy intake. J Nutr 133, Suppl. 3, 895S920S.Google Scholar
10Maurer, J, Taren, DL, Teixeira, PJ, et al. (2006) The psychosocial and behavioral characteristics related to energy misreporting. Nutr Rev 64, 5366.Google Scholar
11Lioret, S, Touvier, M, Balin, M, et al. (2011) Characteristics of energy under-reporting in children and adolescents. Br J Nutr 105, 16711680.Google Scholar
12McGloin, AF, Livingstone, MB, Greene, LC, et al. (2002) Energy and fat intake in obese and lean children at varying risk of obesity. Int J Obes Relat Metab Disord 26, 200207.CrossRefGoogle ScholarPubMed
13Johnson, RK, Driscoll, P & Goran, MI (1996) Comparison of multiple-pass 24-hour recall estimates of energy intake with total energy expenditure determined by the doubly labeled water method in young children. J Am Diet Assoc 96, 11401144.Google Scholar
14Bandini, LG, Cyr, H, Must, A, et al. (1997) Validity of reported energy intake in preadolescent girls. Am J Clin Nutr 65, 1138S1141S.CrossRefGoogle ScholarPubMed
15Tooze, JA, Subar, AF, Thompson, FE, et al. (2004) Psychosocial predictors of energy underreporting in a large doubly labeled water study. Am J Clin Nutr 79, 795804.CrossRefGoogle Scholar
16Lissner, L (2002) Measuring food intake in studies of obesity. Public Health Nutr 5, 889892.Google Scholar
17Burrows, TL, Martin, RJ & Collins, CE (2010) A systematic review of the validity of dietary assessment methods in children when compared with the method of doubly labeled water. J Am Diet Assoc 110, 15011510.Google Scholar
18Forrestal, SG (2011) Energy intake misreporting among children and adolescents: a literature review. Matern Child Nutr 7, 112127.Google Scholar
19Poslusna, K, Ruprich, J, de Vries, JH, et al. (2009) Misreporting of energy and micronutrient intake estimated by food records and 24 hour recalls, control and adjustment methods in practice. Br J Nutr 101, Suppl. 2, S73S85.Google Scholar
20Ahrens, W, Bammann, K, de Henauw, S, et al. (2006) Understanding and preventing childhood obesity and related disorders-IDEFICS: a European multilevel epidemiological approach. Nutr Metab Cardiovasc Dis 16, 302308.CrossRefGoogle ScholarPubMed
21Bammann, K, Peplies, J, Pigeot, I, et al. (2007) [IDEFICS: a multicenter European project on diet- and lifestyle-related disorders in children]. Med Klin (Munich) 102, 230235.Google Scholar
22Ahrens, W, Bammann, K, Siani, A, et al. (2011) The IDEFICS cohort: design, characteristics and participation in the baseline survey. Int J Obes (Lond) 35, Suppl. 1, S3S15.Google Scholar
23Vereecken, CA, Covents, M, Sichert-Hellert, W, et al. (2008) Development and evaluation of a self-administered computerized 24-h dietary recall method for adolescents in Europe. Int J Obes (Lond) 32, Suppl. 5, S26S34.CrossRefGoogle ScholarPubMed
24Vereecken, CA, Covents, M, Matthys, C, et al. (2005) Young Adolescents' Nutrition Assessment on Computer (YANA-C). Eur J Clin Nutr 59, 658667.Google Scholar
25Cole, TJ, Freeman, JV & Preece, MA (1998) British 1990 growth reference centiles for weight, height, body mass index and head circumference fitted by maximum penalized likelihood. Stat Med 17, 407429.Google Scholar
26Cole, TJ, Freeman, JV & Preece, MA (1995) Body mass index reference curves for the UK, 1990. Arch Dis Child 73, 2529.CrossRefGoogle ScholarPubMed
27Birch, LL, Fisher, JO, Grimm-Thomas, K, et al. (2001) Confirmatory factor analysis of the Child Feeding Questionnaire: a measure of parental attitudes, beliefs and practices about child feeding and obesity proneness. Appetite 36, 201210.CrossRefGoogle ScholarPubMed
28Mensink, GB, Kleiser, C & Richter, A (2007) [Food consumption of children and adolescents in Germany. Results of the German Health Interview and Examination Survey for Children and Adolescents (KiGGS)]. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 50, 609623.Google Scholar
29Schofield, WN (1985) Predicting basal metabolic rate, new standards and review of previous work. Hum Nutr Clin Nutr 39, Suppl. 1, 541.Google Scholar
30Goldberg, GR, Black, AE, Jebb, SA, et al. (1991) Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. Derivation of cut-off limits to identify under-recording. Eur J Clin Nutr 45, 569581.Google Scholar
31Sichert-Hellert, W, Kersting, M & Schoch, G (1998) Underreporting of energy intake in 1 to 18 year old German children and adolescents. Z Ernahrungswiss 37, 242251.Google Scholar
32Nelson, M, Black, AE, Morris, JA, et al. (1989) 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 50, 155167.Google Scholar
33Black, AE (2000) Critical evaluation of energy intake using the Goldberg cut-off for energy intake:basal metabolic rate. A practical guide to its calculation, use and limitations. Int J Obes Relat Metab Disord 24, 11191130.Google Scholar
34Torun, B, Davies, PS, Livingstone, MB, et al. (1996) Energy requirements and dietary energy recommendations for children and adolescents 1 to 18 years old. Eur J Clin Nutr 50, Suppl. 1, S37S80.Google Scholar
35Kimm, SY, Glynn, NW, Obarzanek, E, et al. (2006) Racial differences in correlates of misreporting of energy intake in adolescent females. Obesity (Silver Spring) 14, 156164.Google Scholar
36Murakami, K, Miyake, Y, Sasaki, S, et al. (2011) Characteristics of under- and over-reporters of energy intake among Japanese children and adolescents: The Ryukyus Child Health Study. Nutrition 28, 532538.Google Scholar
37Baxter, SD, Smith, AF, Litaker, MS, et al. (2004) Recency affects reporting accuracy of children's dietary recalls. Ann Epidemiol 14, 385390.Google Scholar
38Freedson, P, Pober, D & Janz, KF (2005) Calibration of accelerometer output for children. Med Sci Sports Exerc 37, S523S530.Google Scholar
39Svensson, V, Jacobsson, JA, Fredriksson, R, et al. (2011) Associations between severity of obesity in childhood and adolescence, obesity onset and parental BMI: a longitudinal cohort study. Int J Obes (Lond) 35, 4652.CrossRefGoogle ScholarPubMed
40Lanctot, JQ, Klesges, RC, Stockton, MB, et al. (2008) Prevalence and characteristics of energy underreporting in African-American girls. Obesity (Silver Spring) 16, 14071412.Google Scholar
41Vagstrand, K, Lindroos, AK & Linne, Y (2009) Characteristics of high and low energy reporting teenagers and their relationship to low energy reporting mothers. Public Health Nutr 12, 188196.Google Scholar
42Savage, JS, Mitchell, DC, Smiciklas-Wright, H, et al. (2008) Plausible reports of energy intake may predict body mass index in pre-adolescent girls. J Am Diet Assoc 108, 131135.CrossRefGoogle ScholarPubMed
43Sichert-Hellert, W, Kersting, M, Manz, F, et al. (1994) Energy intake of children and adolescents aged 1–18 years: nutrition survey versus recommendation. Bibl Nutr Dieta 4548.Google Scholar
44Haraldsdottir, J & Sandstrom, B (1994) Detection of underestimated energy intake in young adults. Int J Epidemiol 23, 577582.Google Scholar
Figure 0

Table 1 Reference values to recalculate the Goldberg cut-offs for application in children

Figure 1

Table 2 Lower and upper cut-off limits to classify 1 d 24 h dietary recalls (24-HDR) as under-, plausible and over-reports based on the ratio of energy intake (EI*):BMR†

Figure 2

Table 3 Prevalence of misreporting by study centre, sex and age group (Total numbers and percentages)

Figure 3

Table 4 Descriptive analysis of categorical covariates by reporting group (Percentages and total numbers)

Figure 4

Table 5 Descriptive analysis of continuous covariates and dietary intakes by reporting group (Mean values and standard deviations)

Figure 5

Table 6 Results of the multi-level logistic regression applying backward selection: factors significantly associated with under-reports/over-reports (models 1a and 2a) and predictive value of selected food items for misreporting (models 1b and 2b)* (Odds ratios and 95 % confidence intervals)