Validation of food diary method for assessment of dietary energy and macronutrient intake in infants and children aged 6 ± 24 months

Transcription

1 (2001) 55, 124±129 ß 2001 Nature Publishing Group All rights reserved 0954±3007/01 $ Validation of food diary method for assessment of dietary energy and macronutrient intake in infants and children aged 6 ± 24 months JA Lanigan 1 *, JCK Wells 1, MS Lawson 1 and A Lucas 1 1 Institute of Child Health, London, UK Objectives: To compare the estimated food diary record (ED) method against weighed intake record method (WI) for assessing dietary intake in infants and children aged 6 ± 24 months; additionally, to compare WI with metabolisable energy intake (ME) measured by doubly labelled water (DLW) in infants aged 6 ± 12 months. Design: Cross-over study of 5 day WI vs 5 day ED. Subjects: Seventy-two children aged 6 ± 24 months. Methods: Subjects were randomly assigned to one method during week 1 crossing over to the alternative method in week 2. Data were coded and translated into daily nutrient intakes using COMP-EAT version 5 nutritional analysis software. The analysis compared energy, protein, fat and carbohydrate. Twenty-one infants were dosed with DLW for measurement of total energy expenditure (TEE) and ME. Results: Mean energy intake calculated from WI and ED was 3782 and 3920 kj=day, respectively. There was no signi cant difference between these values. Using WI as a reference, ED showed a mean bias of 138 kj=day, equivalent to 3.6% of mean energy intake. Limits of agreement ( 2 s.d. of the bias) were wide at 1385 kj=day. There were no signi cant differences between methods for any of the nutrient sub-classes. Using DLW as a reference, WI showed a mean bias of 243 kj=day, equivalent to 7.3% of mean energy intake, limits of agreement were wide at 1686 kj=day. Conclusion: There is no evidence from the present analysis that ED is less accurate than WI for assessing energy and nutrient sub-class intakes in groups of this age but this good agreement between methods in groups does not extend to individuals. Sponsorship: Procter & Gamble Company, Cincinnati, USA. Descriptors: dietary assessment; doubly labelled water; food records; intake (2001) 55, 124±129 Introduction Dietary assessment forms the basis of most studies investigating relationships between dietary intake and health outcomes. Assessing dietary intake is an expensive and time-consuming process and a considerable degree of commitment is required from study participants. It is recommended that, in clinical investigations, weighed intake (WI) records of up to 14 days are obtained, while *Correspondence: JA Lanigan, Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK. Guarantor: Professor Alan Lucas. Contributors: JL coordinated the study, conducted the eld work and data analysis, prepared the original manuscript and edited subsequent versions. JW participated in the eld work including the doubly labelled water analysis and co-wrote the manuscript. ML assisted in the study design, participated in eld work, and co-wrote the manuscript. AL conceived the original study concept and assisted in the design and appraisal of the ndings. Received 9 May 2000; revised 16 October 2000; accepted 16 October 2000 in epidemiological studies it has been suggested that 3 days of records are needed for assessing energy intake (Bingham, 1987). Certain hypotheses dictate a need to assess intake of speci c nutrients, and studies of this nature may require greater numbers of food records. The estimated diary food record (ED) method represents a simpler, less expensive method than the WI method for assessing dietary intake in young children, but its accuracy in groups and individuals has not previously been assessed. The ED method must therefore be validated against an alternative method of measuring energy intake, which must itself have been proven to be valid. Accuracy of the WI method has been considered in preschool children aged 1.5 ± 4.5 y using doubly-labelled water as a reference (Davies et al, 1994). These authors concluded that the WI method was acceptable for assessing dietary intake in groups but not in individuals. The difference between energy intake estimated from weighed food records and energy expenditure as measured by the doublylabelled water method averaged 3% for the entire sample, but reached 6% in the younger children aged 1.5 ± 2.5 y. No

2 equivalent data have been reported previously for infants younger than 1.5 y. The aims of the present study were therefore (1) to assess the validity of the criterion WI method in younger infants, and (2) to validate the ED method against the WI method in a larger sample of infants aged 0.5 ± 2 y. The comparison of dietary intake methods was also extended to include macronutrient sub-classes. Subjects and methods Subjects Ethical permission for the study was sought and obtained for researchers to approach parents of infants and children who attended clinics at GP practices and health centres. Parents were interviewed initially as a screening procedure with the assistance of an appointed commercial agency. The agency used (GfK Ltd) worked under the guidance of Institute of Child Health researchers and administered a questionnaire to assess suitability for participation in the study. If parents met inclusion criteria (Table 1), they were then invited to attend a study brie ng session. Characteristics of subjects completing the study are shown in Table 2. The age distribution within both age groups showed a degree of skew towards younger subjects. There were no signi cant differences in either subject number or age between the two age groups. In terms of body size, the sample was representative of this age group with a mean standard deviation score for body weight of 0.26 (s.d. 1.21) (Freeman et al, 1995). Table 1 Inclusion criteria Criteria for study inclusion Exclusion criteria Mothers with singleton births Mothers with multiple births One child aged between Additional child under 2 y of age 6 ± 24 months One main carer Child cared for by minder=nursery Child eating a western-type diet Child eating ethnic diet Daily telephone contact possible Unable to contact main carer daily Carer able to give 2 weeks to study Unable to give 2 weeks to study Subjects in good health Chronic illness, developmental delay Table 2 Characteristics of study subjects 6 ± 12 months 12 ± 24 months Number Percentage male Mean s.d. Mean s.d. Age (months) Bodyweight (kg) Bodyweight s.d. score Study design At the brie ng session, researchers presented an overview of the study to interested parents who were then invited to attend a training session to be held the following week at the same venue. The subjects' expenses were reimbursed and an offer made to provide nutritional analysis of each child's diet. Three training sessions were held per phase of study and each study phase spanned a 2 week period. A total of four phases was needed to complete the study. Upon arrival at the training centre each child's body weight was measured and recorded. Children were then accommodated in a creâche provided at the study site, to allow training of parents. The study took the form of a simple cross-over design. At the rst training session, parents were randomly assigned to record their child's dietary intake with either WI or ED. Parents were trained in small groups (maximum ve people) in one method of recording dietary intake. Dietitians taught parents how to estimate their child's food intake either by describing amounts relative to standard household measures or by weighing amounts of food and drink offered, using Soehnle electronic kitchen scales, accurate to 1 g. Parents were also asked to provide weights or estimates of quantities of food not consumed. These could be foods left by the child in the feeding vessel or lost as wastage. Food diaries designed speci cally for this study, as described below, were provided. Estimated food intake records The estimated food intake record comprised a 5-day food diary with printed instructions. A four-column table was provided for each day with provision for recording meal type and time offered, food and drink items offered, amount of food and drink offered and amount of food and drink not consumed. Parents were asked to choose 5 days when the child was eating his or her usual diet. On these days recording was directed to commence at midnight to incorporate episodes of night feeding. Where infant formulae were reconstituted, parents were asked to record the total amount after addition of powder to water and to record the volume not consumed after feeding. Mothers were asked to record the duration of each breast-feeding episode. Estimation of breast milk consumption was based on Medical Research Council data (Paul et al, 1988 in Mills & Tyler, 1992) to be approximately 135 g for infants aged 6 ± 7 months and 100 g for those aged 8 ± 12 months, where a feed of 10 or more minutes was taken as equivalent to a full feed. Where feeding duration was less than 10 min, consumption was calculated proportionally. When describing foods, the diary instructions asked that parents gave a full description of foods offered with examples given. These should include brand name, container type and size etc. Where possible parents were asked to provide food labels from manufactured produce. In the case of home-prepared food, parents were asked to provide 125

3 126 recipes, complete with quantities, detailed descriptions of ingredients used and cooking methods. A page was included describing standard household measures including capacities. For example parents were asked to estimate the quantity of food offered in standard household spoons, describing these as level, rounded or heaped. Leftovers were instructed to be measured in the same way with estimates made for wastage. The capacity of standard tea, dessert and tablespoons were given as 5, 10 and 15 ml, respectively, and pictorial representations were included in the diary. Parents were asked to check their utensils for compliance with these quantities. A simple ruler was printed on the food diary and parents were asked to use this to indicate the size of foods offered and left uneaten. A printed example page was provided with the diary. Weighed food intake record The weighed food record took the same format as the estimated food record, differing in the instructions and examples given. At the training session parents were shown how to record food items individually and cumulatively using the scales. When dried foods were reconstituted parents were asked to weigh the dry powder rst and record the quantity and to then weigh the nal quantity offered to the child after addition of uid. For the estimation of wastage, instructions were given to collect all visible leftovers=spillage and reweigh these on a clean plate recording the total quantity. The contribution of individual foods to the total wasted was then requested to be estimated proportionally. Parent support According to the study design, parents were instructed in the alternative recording method at the second training session. In between training sessions regular telephone contact was maintained with parents by the dietitian coordinating the study. During these calls parents were encouraged to discuss any problems they were having with the food intake records. Where dif culties were reported, problems were resolved either by discussion or, where necessary, home visits were made. Doubly labelled water In a sub-set of 21 infants aged 6 ± 12 months, total energy expenditure was measured using doubly labelled water (DLW) as described by Davies et al (1989) and metabolizable energy intake was calculated as described previously (Wells & Davies, 1998). The isotope was offered diluted in the infant's usual drink. Wherever possible, the infant's regular feeding vessel was used. The parents of these children were instructed individually in collection of 7 day urine samples. Data sheets were provided for parents to record the collection times of the daily urine samples. Statistical methods Intakes of energy and other nutrients were calculated from the food records using COMP-EAT version 5 nutritional analysis software. These data were then analysed using the Statistical Package for Social Sciences (SPSS) version 8.0. Sample means were calculated and compared via paired t-test. The data were also analysed using the Bland and Altman (1986) technique. This method of analysis assesses agreement of methods in groups, which is de ned as the bias, and also assesses agreement in individuals, de ned as the limits of agreement ( 2 s.d. of the bias). For total energy and nutrient sub-class intakes, the relative bias was calculated as the mean difference between the estimated and weighed intake records. Additional analyses were conducted to determine whether agreement between methods varied in relation to the age of the infant. Firstly, the correlation between age and absolute between-method difference was calculated for energy and the three macro-nutrient subclasses. Secondly, the between-method difference was expressed as a percentage of the intake of the relevant nutrient, with the value of the intake being calculated as the mean by the two methods. The correlation between age and between-method difference adjusted for intake was then calculated. Results A description of the sample can be seen in Table 2. Results for mean metabolizable energy intake, which includes total energy expenditure and energy stored in new tissue from growth, in 21 infants aged 6 ± 12 months are presented in Table 3. The mean error for the doubly labelled water analysis was 6%. When estimating energy expenditure from records of food intake, both the weighed record and food diary produced a 7.3% overestimation compared with metabolizable energy intake (Table 4). The mean energy intakes for all subjects calculated from estimated and weighed records are presented in Table 5. There was no signi cant difference between these two mean values. Mean values for intakes of energy-yielding nutrients were also not signi cantly different. ED gave higher estimates but not signi cantly so, for dietary intake of energy and nutrient sub-classes protein, fat and carbohydrate, the difference being < 4% for all nutrients (Table 5). Table 6 gives the bias and limits of agreement of ED for energy and the three nutrient sub-classes. The table also Table 3 Results of double labelled water measurements in 21 infants aged 6 ± 12 months Mean Age (months) Wt (kg) TEE a (kj=day) TEE=kg (kj=kg=day) Metabolisable energy intake (kj=day) Weighed intake (kj=day) Diary (kj=day) a Total energy expenditure as measured by doubly labelled water. s.d.

4 gives the correlation coef cients for the relationship between the bias and the mean intake for each nutrient. There was a signi cant effect of age on correlation between methods. With respect to total energy, protein and fat, absolute differences between methods increased signi cantly with age of subject. In the case of carbohydrate intake there was no signi cant association with age (Table 6). However, when differences were expressed as percentage of the mean value obtained by the two methods per individual, the correlations for protein and energy decreased below the signi cance level chosen (P < 0.05), whereas the correlation for fat remained signi cant. Thus agreement between methods was poorer in older children, even after taking the effect of body size into account. Discussion Assessment of dietary intake is generally recognized to be dif cult to accomplish. Many studies investigating dietary intake in children have therefore focused on less timeconsuming and costly methods, the most popular choices being the diet history, 24 h recall, or food frequency questionnaire. It is also dif cult to assess the validity of Table 4 Relative bias and limits of agreement between two methods of dietary intake assessment and energy expenditure measured by DLW Bias (kj=day) Limits of agreement (kj=day) Weighed intake ± DLW 243 = Diary ± DLW a 238 = a Metabolisable energy intake as measured by doubly labelled water. such methods because they rely on subjective reports provided either by the subjects themselves, or in the case of preschool children, on proxy reports provided by parents or carers. The presence of non-random errors within methods assessing dietary intake would confound results of dietary surveys, so it is essential to evaluate the accuracy of such methods in groups and individuals. There is a dearth of research evaluating methods of dietary intake in infants and young children. We are aware of no other studies which have attempted to validate estimated food records using a diary against weighed food records in groups of children aged 6 ± 24 months. The main aim of our study was to validate a method of recording food intake using another food intake method as a reference. The study further attempted to assess the validity of this reference method, using another reference in which energy intake was calculated from measurements of energy expenditure and weight gain as opposed to measurements of food intake. Provided that growth is taken into account, the doubly labelled water technique provides a precise measure of metabolisable energy intake among free-living individuals. Although good agreement between measured metabolizable intake and observed energy intake calculated from dietary intake assessment methods has been demonstrated (Davies et al, 1994), several other studies have reported underestimations of habitual intake (Bandini et al, 1990; Livingstone et al, 1990, 1992; Schoeller et al, 1990; Black et al, 1997). Work by Davies et al (1994) and the present study indicate a moderate but not signi cant bias of weighed food records against doubly labelled water in infants and young children. However, there are a number of reasons why absolute agreement between these methods is not to be expected. 127 Table 5 Comparison of sample means for intake of energy and energy-yielding nutrients in 72 subjects Estimated record Weighed record Difference Dietary component Mean s.d. Range Mean s.d. Range (% mean) Energy (kj=day) ± ± (kcal=day) (937.) (205.) (521 ± 1469) (904.) (206.) (482 ± 1479) Protein (g=day) ± ± Fat (g=day) ± ± Carbohydrate (g=day) ± ± Table 6 The relative bias, limits of agreement and age-related correlation coef cients of the measurements of energy, fat, protein and carbohydrate intake by estimated food diary record method and by weighed food intake method in 72 children Energy (kj=day) Fat (g=day) Protein (g=day) Carbohydrate (g=day) Bias Limits of agreement Bias as percentage mean intake Correlation of age and absolute difference between methods 0.37* 0.38** 0.31*** 0.11 Correlation of age and between-method difference expressed as a percentage of the intake a { *P ˆ 0.002; **P ˆ 0.001; ***P ˆ 0.009; { P ˆ a Intake calculated as the mean by the two methods.

5 128 Firstly, metabolisable energy intake will only correspond to observed energy intake if gut function is fully mature, and an appropriate approach accounting for gross energy losses is incorporated into the calculations. Secondly the successive rather than simultaneous application of isotopic and observational methods in our study may have included actual differences in food intake during the 2-week-long measurement periods. Thirdly, breast-milk intake was predicted from feeding times and not directly measured by test weighing. In a study among younger infants in whom breast-milk intake may have accounted for a large proportion of total dietary intake, this would have been a greater cause for concern. However, in the current study breast-milk accounted for only 6% of the total group energy intake. These factors may therefore have contributed to the mean bias of 6 ± 7% between metabolisable intake and observed intake found both by Davies and colleagues and ourselves. However, subject to this relatively small bias, these ndings support the use of the WI method for measuring food intake in groups of younger children. In the comparison between observed intake methods, we found no evidence that ED overestimated mean intakes relative to the WI method. The bias for both energy and macronutrient sub-classes was minimal and not statistically signi cant. However, this good agreement between methods in groups does not extend to individuals. Other studies comparing methods of dietary assessment in children have reported signi cant differences between methods recording dietary intake. One study among Indo-Asian children aged 4 ± 40 months comparing weighed records with diet history found that the latter method gave signi cantly higher estimates of 7%, 9% and 9% for energy, protein and iron intake, respectively. Fat and vitamin C intake were also overestimated by 3 and 6%, respectively, although these differences did not reach signi cance (Harbottle & Duggan, 1993). The Leiden Pre-School Children Study (Horst et al, 1988) also demonstrated systematic bias, albeit small at less than 5%, between methods when comparing 24 h recall with duplicate food portions. Energy and nutrient intake of Dutch children aged 4 ± 28 months was overestimated by the recall method. However, in the present study these overestimations were minimal and did not exceed 4% for any of the categories of nutrients. Measurement of food intake by weighing, report or recall methods is notoriously dif cult in all age groups. For obvious reasons, assessment of dietary intake in preschool children must rely on parental reports. These reports may be distorted by a number of factors, including the altering of habitual behaviour while participating in the study (the Hawthorne effect), genuine forgetfulness, lack of compliance, and being economical with the truth either consciously or unconsciously. Foods play non-nutritive roles in early life and may be used as bribes or rewards (Wells, 1998), but parents may be unwilling to admit such behaviour patterns to researchers. However, despite these dif culties, other factors are likely to promote accuracy in this age group. As parents largely have control over the dietary intake of children and because this age-group mainly eats at home with limited access to supplementary sources of foods, a good degree of accuracy is at least possible. Furthermore, parents might be expected to be highly motivated to report their child's diet accurately, since a report on the child's intake comprised a reward for participating in the study. Reliability of parental report is an issue which few studies have addressed. In one study accuracy of mothers' recall when reporting dietary intake among children aged 3 ± 5 y was investigated. When estimating intake using the 24 h recall method, mothers were more likely to under than over-report foods consumed (Baranowski et al, 1991). For intake of nutrients, however, group mean differences between mothers' recall and investigators' observations were quite small. Standard deviations of the difference scores were quite large, suggesting that substantial disagreement exists between mothers' reports for individuals and those of observers. In our study, we assume that such inaccuracies would have equivalent effects in the two methods but this may not be true and the issue is worthy of further research. Age effects Our study has considered agreement between methods not only for total energy but also fat, carbohydrate and protein intakes. Correlation analysis, investigating the relationship between difference between methods and age, found no signi cant differences with respect to total energy, protein and carbohydrate. However, for fat we found a low but signi cant correlation between age and difference between methods. There was no evidence that this difference consisted of a consistent bias of one technique compared to the other, only that as age of the child increased, the techniques gave increasingly different values for fat intake. This could be an effect of the change in diet which occurs during later infancy as the proportional intake of milk decreases. Over 50% of the energy in breast-milk (Department of Health and Social Security 1977) or whole cow's milk (Holland et al, 1991) is derived from fat and therefore when a high proportion of an individual infant's dietary intake is as milk, reported intake from fat is likely to agree more closely. Conclusion There is no evidence from the present analysis that estimated food records are less accurate than weighed records for assessing energy and nutrient sub-class intakes in groups in this age group, but this good agreement between methods in groups does not extend to individuals. These results may have signi cance for large-scale studies where use of weighed intake is impractical due to greater time required by researchers for training and by subjects for recording. Acknowledgements Ð We are grateful to the Procter & Gamble Company, Cincinnati, USA who provided funding for this project.

6 References Bandini LG, Scoeller DA, Cyr HN & Dietz WH (1990): Validity of reported energy intake in obese and nonobese adolescents. Am. J. Clin. Nutr. 52, 421 ± 425. Baranowski T, Sprague D, Baranowski JH & Harrison JA (1991): Accuracy of maternal dietary recall for preschool children. J. Am. Diet Assoc. 91, 669 ± 674. Bingham SA (1987): The dietary assessment of individuals; methods, accuracy, new techniques and recommendations. Nutr. Abstr. Rev. 57, 705 ± 742. Black AE, Bingham SA, Johansson G & Coward WA (1997): Validation of dietary intakes of protein and energy against 24 hour urinary N and DLW energy expenditure in middle-aged women, retired men and postobese subjects: comparisons with validation against presumed energy requirements. Eur. J. Clin. Nutr. 51, 405 ± 413. Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurement. Lancet i, 307 ± 310. Davies PSW, Ewing G & Lucas A (1989): Energy expenditure in early infancy. Br. J. Nutr. 62, 621 ± 629. Davies PSW, Coward WA, Gregory J, White A & Mills A (1994): Total energy expenditure and energy intake in the pre-school child: a comparison. Br. J. Nutr. 72, 13 ± 20. Department of Health and Social Security (1977): The Composition of Mature Human Milk. Report on Health and Social Subjects, no. 12. London: HMSO. Freeman JV, Cole TJ, Chinn S, Jones PRM, White EM & Preece MA (1995): Cross sectional stature and weight reference curves for the UK, Arch. Dis. Child. 73, 17 ± 24. Harbottle L & Duggan MB (1993): Dietary assessment in Asian children Ð a comparison of the weighed inventory and diet history methods. Eur. J. Clin. Nutr. 47, 666 ± 672. Holland B, Welch AA, Unwin ID, Buss DH, Paul AA, & Southgate DAT (1991): McCance and Widdowson's The Composition of Foods, 5th edn. Cambridge: Royal Society of Chemistry. Horst CH, Obermann-De-Boer GL & Kromhout D (1988): Validity of the 24-hour recall method in infancy: the Leiden Pre-School Children Study. Int. J. Epidemiol. 17, 217 ± 221. Livingstone MBE, Prentice AM, Strain JJ, Coward WA, Black AE, Barker, ME, Mckenna PG & Whitehead RG (1990): Accuracy of weighed dietary records in studies of diet and health. Br. Med. J. 300, 708 ± 712. Livingstone MBE, Prentice AM, Coward WA, Strain JJ, Black AE, Davies PSW, Stewart C, Mckenna PG & Whitehead RG (1992): Validation of estimates of energy intake by weighed record and diet history in children and adolescents. Am. J. Clin. Nutr. 56, 29 ± 35. Mills A & Tyler H (1992): Food and Nutrient Intakes of British Infants Aged 6 ± 12 months. London: HMSO. Schoeller DA, Bandini LG & Dietz WH (1990): Inaccuracies in selfreported intake identi ed by comparison with the doubly labelled water method. Can. J. Physiol. Pharmac. 68, 941 ± 949. Wells JCK (1998): Child distress and parental response: implications for energy intake. Ann. Hum. Biol. 25, 392 ± 393. Wells JCK & Davies PSW (1998): Estimation of the energy cost of physical activity in infancy. Arch. Dis. Child. 78, 131 ±