Previous - Next
The most important criteria in choosing a method for collecting food intake data in children and adolescents are: (a) the technique should not interfere with the subject's dietary pattern; (b) the data should be representative of usual or habitual intake and (c) the technique should be suitable for application in large study groups.
The methods most frequently used in childhood and adolescent population groups are similar to those applied in adult studies, namely:
(1) Retrospective or food recall methods, which depend on dietary information given from memory by the child/ adolescent and/or parent/child carer. Several specific types of data collection fall within this category, including those aimed at quantifying actual intake for a precise time (usually the previous day, or 24-h recall) and those designed to elicit information about usual consumption patterns for a longer, less precisely defined time period (diet history or food frequency methods). More than one 24-h recall should be made on different days of the week, especially when there are cultural cyclic changes in food intake (e.g. weekdays compared with weekends). Recalls of more than 24 h are sometimes performed but the accuracy with which subjects and/or parents can remember food consumption is debatable, particularly if food intake patterns are highly unstructured or unstable. In the food frequency method, subjects and/or parents/child carers report by interview or self-administered questionnaire, the frequency of consumption of particular foods during a specified time span (week, month, year). A quantitative component is added by including the size and number of portions most frequently consumed for each food.
(2) Prospective or food record methods, which require that all food items consumed be recorded at the time of consumption. Intakes are quantified by direct weighing of the food, by estimates using, household measures or by collection of duplicate diets. Quantitative assessment of usual food intake can be obtained by increasing the number of measurement days. Seven days are generally assumed to represent a good compromise between precision, subject/parental cooperation, cultural dietary patterns and investigator workload.
Each of these methods has advantages and drawbacks when applied to children and adolescents. Ultimately, all survey methods are dependent on the motivation, compliance and ability of subjects and/or parents/child carers to report accurately habitual food intake.
Food intake data must then be converted into energy equivalents. This is often done disaggregating recipes into their food components and calculating their metabolizable energy as reported in food composition tables. Care must be taken to make all necessary conversions for the proper use of food composition data. A common error is applying to 'cooked' or 'wet' weight of foods the energy values for 'raw' or 'dry' foods that appear in composition tables, without applying adequate conversion factors.
A more accurate approach is to perform chemical or calorimetric analyses of samples of foods that are ready to be eaten. This is particularly useful to calculate the energy provided by food recipes that are unlikely to appear in food composition tables or that may be subject to variations. When the energy content of food is measured by bomb calorimetry, appropriate corrections must be made to calculate metabolizable energy.
Validity of energy intakes in children and adolescents
Most dietary intake studies in children assume that the data obtained are representative of habitual food consumption, and many recent studies concluded that energy intakes (EI) have declined in industrialized countries and more privileged groups in developing countries in response to a secular trend towards lower levels of activity in children and adolescents. However, studies in adults using doubly-labeled water (DLW) measurements of total energy expenditure (TEE) to validate EI have demonstrated that intake data may underestimate habitual food intake to a greater extent than has been appreciated (Prentice et al, 1986; Livingstone et al, 1990b; Schoeller, 1990). It is conceivable, therefore, that the reportedly low intakes of children may be artifacts of dietary survey methodology, rather than indicative of a diminution in energy expenditure.
Validation studies have been reported to assess the accuracy of EI in children and adolescents, using DLW measurements of TEE. These include studies of EI by 4-day weighted dietary record (WDR) in 1.5-4.5 year olds (n = 81) (Davies et al, 1994), by 7-day WDR in 7, 9, 12, 15 and 18 year olds (n = 58), by diet history (DH) in 3, 5, 7, 9, 12, 15 and 18 year olds (Livingstone et al, 1992b) and by 14 day estimated food records in non-obese and obese adolescents (n = 55) (Bandini et al, 1990a).
The results shown in Figures 8-10 indicate that bias in dietary reporting does not operate uniformly across age groups and that it is influenced by the particular methodology used.
In children aged 1.5-4.5 years, mean El calculated by 4-day WDR were not significantly different from mean TEE (+3%) (Figure 8a). Similarly, the mean EI by 7-day WDR of 7 and 9 year olds were in close correspondence with simultaneous measurements of TEE ( + 2%) (Figure 10a), but in adolescents and young adults there was increasing divergence between EI and TEE as age increased: mean EI were significantly lower than TEE in 12 year olds ( - 14%) and in 15 and 18 year olds (-24%, P<0.01) (Figure 10a). Using 14-day estimated intake records, Bandini et al (1990a) also showed a substantial underestimation of EI by adolescents, with the negative bias being most apparent in obese subjects (Figure 8b). After adjustment for changes in body composition, mean estimated EI were 80 ± 23% (non-obese) and 54 ± 32% (obese) of TEE values (P < 0.001).
The age-related discrepancy differed in the study to validate EI by diet history in 3-18 year olds. There was a bias towards overestimation of EI in the younger children by this technique: as age increased, mean differences were + 12%, + 9%, + 11% and - 1% (Figures 9 and 10b)
These validation studies can be criticized because they only involved a small number of subjects in various age groups. However, all of them indicate that a bias in dietary reporting is highly probable. Thus, considerable caution needs to be applied when interpreting energy intake data sets as a basis for deriving energy requirements. Moreover, the magnitude and direction of the errors in children's EI are likely to be different from those found in adults. These biases are highly relevant to the problem of determining appropriate energy intakes for nitrogen balance studies (see Appendix).
Age is an important variable that affects compliance in dietary reporting. The results presented suggest that the mean EI assessed by weighed dietary records are more likely to represent usual food intake in younger than in older subjects. This could be due to the fact that in young children overall control of food intake and responsibility for dietary reporting are shared by parents and other adults concerned with child caring. Younger children also have less unsupervised access to food in- and out-of-home. On the other hand, by early adolescence the responsibility for reporting shifts more to the subjects themselves. Consequently, their greater food requirements in combination with unstructured eating patterns and a significant degree of out-of-home eating suggest that under-reporting (by WDR) may be partly due to forgetfulness and lack of compliance with a demanding protocol.
Obesity is another important factor. In common with obese adults (Prentice et al, 1986), obese adolescents have been found to under-report EI significantly more than their non-obese counterparts (Bandini et al, 1990a). Preoccupation with body weight and image, which may lead to real or apparent dietary restraint, seems to be well developed in girls with normal and low weight by the age of 12 years. Similar, although less marked trends, have been observed in adolescent boys (Livingstone et al, 1992b).
The method used to assess EI also may influence the results. Validation studies with various EI methods across the entire age range of childhood and adolescence are lacking. Only one study has validated simultaneously EI by WDR and DH with TEE (Livingstone et al, 1992b). Although EI by DH were biased towards overestimation in most age groups and individual measurements lacked precision, mean intakes assessed by DH seemed more representative of habitual EI across the age range than WDR. The apparent superiority of DH in overcoming an age-related bias in dietary reporting is contrary to expectations and needs to be evaluated carefully. Since DH is not a standardized instrument and it only measures memory and perception of usual diet, it is subjective and children may tend to exaggerate the intake of 'good' foods and under-estimate 'bad' foods. Accuracy in reporting is also dependent on motivation, intelligence, an adequately developed concept of time, ability to recognize foods, the complexity and stability of food patterns and the age at which children can reliably report their own food intake without control or supervision of adults.
Other factors which are likely to influence reporting accuracy and about which little is known, include social class and educational background.
In addition to the credibility of food intake reports, assessment of EI can be distorted by the use of inadequate food composition tables and/or overlooking the conversion of cooked and processed foods into their raw ingredients2.
2 The world-wide food composition data network being developed by INFOODS offers electronic access to information on prepared and processed foods often not available in local food composition tables (for information: http ://www.crop.cri.nz/crop/infoods/infoods.html).
Dietary energy intake data of children and adolescents
A selection of dietary intake studies reported in the literature from about 1980 onwards are reviewed here since earlier studies were evaluated extensively by Ferro-Luzzi & Durnin (1981), as the basis for the 1985 FAO/WHO/UNU estimated requirements. Since 1980, a vast number of dietary intake studies on children and adolescents have been reported and the studies cited in this review are by no means an exhaustive compilation. Many studies were excluded based on the following criteria:
(1) When energy intakes were reported for wide age bands (e.g. 11-16 years) and the mean age was not recorded.
(2) When energy intakes were reported combined for boys and girls over 10 years old.
(3) When data were presented in a format which could not be readily interpreted for the purposes of this review (e.g., in graphs). Unfortunately, many studies in developing countries were excluded for this reason.
(4) When the children studied were generally malnourished or obese, and their mean weight-for-height differed from the NCHS/WHO standards by more than 2 s.d. Many reports were based on representative study populations and therefore included children with a range of body weights.
(5) Only studies of healthy children were included, since many disease states are likely to affect energy intakes and requirements.
Tables 23 and 24 give details of the studies that were reviewed. Forty-eight involved children approximately 110 years old, and 41 studies included children and adolescents approximately 10-18 years old.
Tables 25-30 show the energy intakes of the children, by ascending age. Boys and girls under 5 years are listed together in Table 25, as many studies did not separate the results for each sex. The same is true of the six studies in Table 30. When body weights were not reported, median weights (NCHS) at the mid-point of the age range were assumed and, in Tables 25 and 30, averaged for boys and girls. Energy intake data are presented as absolute values, in relation to body weight, and as multiples of the estimated BMR. The latter were calculated from the mean weights using the equations proposed by Schofield to FAO/WHO/UNU (1985).
Comparison with total energy expenditure and dietary recommendations
When energy intakes are used to assess requirements or to estimate whether the mean intake satisfies a population's dietary recommendations, the possibility of bias must be acknowledged and the data should be analyzed and interpreted accordingly. Information that is incompatible with fundamental principles of energy physiology should not be accepted, as it cannot represent long-term usual intake or is due to methodological bias or inadequate reporting. Goldberg et al (1991) and Black et al (1991) suggested a screening of EI data of adult populations, calculating them as multiples of BMR. For example, a value below 1.27 × BMR, considered as the survival requirement for adults (FAO/ WHO/UNU, 1985), is unacceptable as representative of habitual intake.
Following that logic, we used the PALs shown in Table 21 to establish reasonable limits to evaluate dietary energy surveys among children and adolescents. Mean results lower than two times the coefficient of variation (i.e. 12%) below the PAL corresponding to light habitual activity, or higher than two times the CV above the PAL for heavy habitual activity were considered unlikely to represent the usual intake of healthy children. Since the PALs for boys or girls 6-13 and 14-18 years old in Table 21 are reasonably close, the acceptable limits for those age groups were averaged to simplify the evaluation of the results in Tables 25-30. Further corrections for the energy needs for growth were not made, as they are only about 3% at age 1 and less than 1 % in late adolescence.
Thus, Tables 31-33 were prepared from the data in Tables 25-30 that were between 1.28 and 1.79 × BMR for children 1-5 years, between 1.39 and 2.24 × BMR for boys 6-18, and between 1.30 and 2.10 × BMR for girls 6-18. Mean energy intakes expressed as MJ/d, kJ/kg/d and × BMR, were weighted for the number of children in each study. When a study included more than 500 or 1000 children of a given age and sex, only 30% or 20% of the number, respectively, were used to calculate the weighted means to avoid an extreme bias toward the results of that study.
As Table 31 and Figure 11 show, energy intake per unit of body weight is fairly constant for both boys and girls between 3 and 7 years of age, after which it decreases gradually until age 15 (girls) or 16 (boys).
Compared with total energy expenditure assessed with doubly-labeled water and heart rate monitoring, energy intake tends to overestimate requirements under 8-10 years and to underestimate them after that age. Those trends also apply to the 1985 FAO/WHO/UNU energy recommendations, but the overestimation is markedly higher under 6 years of age. This is partly due to the 5% additional dietary energy recommended in 1985 for children 1-10 years old to accommodate 'a desirable level of physical activity'.
The reported EI of children 1-5 years old is about 13% lower than FAO/WHO/UNU requirements (Figure 11, Table 31). Although the wide range between data sets could reflect real differences in intake, unrepresentative study samples, or artifacts in dietary survey methodology, mean intakes fell short of FAO/WHO/ UNU requirements in about 80% of the data sets.
The influence of sex on dietary energy intake is illustrated in Figure 12 and Tables 31-33. Girls have lower EI than boys, whether expressed in absolute terms or relative to their body weights or their estimated BMR, and the difference becomes greater in adolescence. These findings are consistent with their lower total energy expenditure (Tables 2-7 and 20, and Figure 5).
Conclusions
Recent trends in EI of children and adolescents suggest that if the groups studied are representative of their age and sex, and the EI data are valid measures of habitual intake, then:
(a) Habitual energy intakes of 1-6 year old children are lower than current recommendations. Increasing reported energy intakes by 5% to accommodate a 'desirable level of physical activity' may be unrealistic.
(b) Energy requirements for physical activity may be more variable in adolescent males but lower in the adolescent females, than has been assumed when deriving factorially estimated energy requirements.
For methodological and economic reasons it seems inevitable that we will continue to rely partly on reported EI data as a basis of estimating energy requirements for most populations. However, it is clear that these data can no longer be tacitly accepted as representative of usual intake. Therefore, the following recommendations need to be considered:
(a) At present there are too few studies in which energy intake and energy expenditure have been studied in the same population to know the nature and extent of bias involved in these measurements. This will require more extensive validation studies of energy and nutrient intakes that take into account differences in methodology, social status, education, age, and geographical region in both developing and industrialized countries. From these studies guidelines may emerge for detecting patterns of bias and the characteristics of individuals contributing to it.
(b) Variation among individuals within the same population can be appropriately characterized by a mean and standard deviation whose validity will depend upon the adequacy of the sample. However, the nature and extent of differences in mean values among different populations make it unlikely that they can be appropriately characterized by a single mean and standard deviation, no matter how many populations are sampled. It may be better to express a range of mean values for this purpose.
(c) Research must be done to find ways of minimizing the psychological basis of under- and over-reporting in these age groups.
(d) Appropriate 'cut-off' values based on fundamental principles of energy physiology should be used to determine the acceptance of energy intake results. This will require an extensive data base of basal and total daily energy expenditures (BMR and TEE) in association with objective measures of physical activity. In the meantime, the following estimates of multiples of BMR are suggested as provisional cut-off points: 1-5 years (boys and girls): 1.28-1.79 × BMR; 6-18 years: 1.39-2.24 × BMR (boys) and 1.30-2.10 × BMR (girls).
These recommendations will not guarantee valid data and cannot eliminate the considerable differences among populations, but may lead to the design of more effective instruments for assessing energy intake and requirements of children and adolescents.