ORIGINAL COMMUNICATION

Transcription

1 (2003) 57, & 2003 Nature Publishing Group All rights reserved /03 $ ORIGINAL COMMUNICATION Energy intake in Swedish adolescents: validation of diet history with doubly labelled water A Sjöberg 1 *, F Slinde 1, D Arvidsson 1, L Ellegård 1, E Gramatkovski 1, L Hallberg 1 and L Hulthén 1 1 Department of Clinical Nutrition, Sahlgrenska Academy at Göteborg University, Göteborg, Sweden Objective: To compare habitual energy intake (EI) estimated from diet history (DH) with total energy expenditure (TEE) measured with doubly labelled water (DLW) in adolescents. Design: DH included a detailed questionnaire and an interview. TEE was measured during a 14-day period. Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR Setting: Participants were recruited from grade 9 in a compulsory school in Göteborg, Sweden. All data were collected at school and DLW dosages were distributed at Sahlgrenska University Hospital. Subjects: A total of 35 adolescents (18 boys, 17 girls), 15.7 (0.4) y. Results: EI was 11.0 (3.6) MJ and TEE was 11.4 (2.1) MJ (P¼0.42). DH was able to rank EI compared to TEE (Spearman s r¼0.59, Pp0.001). For girls, EI was 18% lower (P¼0.0067) and for boys, EI was 7% higher (P¼0.26) compared to TEE. The 95% limits of agreement for difference between TEE and EI were 5.6 to 6.5 MJ. In total, 20 subjects were defined as AR (57%), nine as UR (26%) and six as OR (17%). Energy from in-between meals was 33% lower (P¼0.0043) in UR girls and 57% higher (P¼0.026) in OR boys, compared to adequate reporting girls and boys, respectively. In UR girls, energy-adjusted intake (10 MJ) of specific foods did not differ significantly, fat was lower and carbohydrate and vitamin C were higher compared to AR girls (all Po0.05). OR boys had no significant differences in food and nutrient intake in 10 MJ compared to AR boys. Conclusion: The diet history was able to capture EI for the group and to rank subjects. There was a wide individual range in reporting- accuracy related to gender. Sponsorship: The Ingabritt and Arne Lundberg Foundation, The Wilhelm and Martina Lundgren Foundation. (2003) 57, doi: /sj.ejcn Keywords: energy intake; energy expenditure; doubly labelled water; adolescent nutrition; food habits Introduction Valid dietary assessment methods are important to identify dietary effects on health. Probably, it is the habitual diet rather than intake during a short time period that is decisive *Correspondence: A Sjöberg, Department of Clinical Nutrition, Sahlgrenska Academy at Göteborg University, Box 459, SE Göteborg, Sweden. agneta.sjoberg@nutrition.gu.se Guarantor: L Hulthén Contributors: L Hallberg and L Hulthén initiated the study. L Hulthén, FS and AS were responsible for study design. FS was responsible for distribution of DLW dosages and DA for urine collections. EG and LE performed analysis and calculations of doubly labelled water measurements. AS was responsible for dietary interviews, analysis of dietary intake, statistical analysis and wrote the paper. All investigators were involved in discussion and interpretation of results, drafting and revision of the paper. Received 17 June 2003; revised 13 August 2003; accepted 17 September 2003 for relationships between diet and disease. Diet history is a retrospective method used to measure food habits. For validation of habitual energy intake (EI), total energy expenditure (TEE) measured with doubly labelled water (DLW) during a 2 week period is an appropriate method if habitual activity levels and weight stability are assumed. Few studies validating the diet history (DH) method using DLW have been conducted according to a recent review (Livingstone & Black, 2003). Usually, intakes from dietary assessments representing shorter periods, such as food records and 24-h recalls, have been used to assess dietary habits and the EI has been validated with TEE using overlapping measurement periods. In adolescents as well as in adults, assessment of EI for the group is often biased towards underestimation (Livingstone & Robson, 2000), partly related to the dietary method used. In adolescents, DH captured EI for the group better than food record when validated simultaneously with DLW (Livingstone et al, 1992).

2 1644 Dietary assessments in adolescent populations may have difficulties beyond that in studies in adults. Adolescents are expected to be able to recall and estimate their habitual intake, but body image and lack of motivation may impair reporting. Further, unstructured eating and high-energy requirement could also make it more difficult to report intake adequately (Livingstone & Robson, 2000; Livingstone & Black, 2003). A high body mass index (BMI) has been related to a higher degree of under-reporting for adolescents as well as for adults (Bandini et al, 1990; Bratteby et al, 1998; Johansson et al, 1998). Social desirability and social approval may be important for difference in dietary reporting between gender. Social desirability has been associated with lower reported intake of energy and fat in a study among women (Hebert et al, 2002). For men there was an association between social approval and higher reported intake of energy and fat (Hebert et al, 1997). A higher degree of underreporting in female populations compared to male has been shown when under-reporting was identified using the ratio of EI to basal metabolic rate (EI/BMR) (Johansson et al, 1998). On the other hand, comparison of proportions of under-, valid- and over-reporting men and women in a review of 43 DLW studies showed no significant differences according to gender (Livingstone & Black, 2003). In the current study, the aim was to validate the DH method designed to capture habitual intake in adolescents. EI calculated from DH was compared with TEE measured with DLW in a group of adolescent boys and girls comparing results of the group and individual measurements. A second aim was to describe characteristics of under- and overreporters and how reported intake differed, with respect to meal pattern, food choice and nutrient intake, in relation to reporting accuracy. Subjects and methods Study design The dietary method validated in this study was designed to study food habits in adolescents in two cross-sectional studies in Göteborg in the years 1994 and The first study included 1245 and the second 1025 adolescents. The aim was to study iron status and food habits in Swedish adolescents when the general iron fortification of white wheat flour was removed. All data were collected in schools and all students in grade 9 in 13 schools in Göteborg were invited to participate. To achieve a representative sample of adolescents of Göteborg, great effort was made in selection of schools using socioeconomic area index (high, medium and low status areas) (Sjöberg et al, 2003). In the current study, performed in 2002, the school-based design was also used and students in grade 9 in one of the schools (medium status area) included in the two cross-sectional studies were invited to participate. The study consisted of DH, medical survey and DLW measurement. DH included a detailed dietary questionnaire and an individual interview with a nutritionist. Dosages of DLW were distributed at the Sahlgrenska University Hospital and urine samples were collected at the school. Subjects A total of 83 students (43 boys and 40 girls), aged y, were invited to participate in the study and completed the dietary questionnaire. In total, 29 subjects declined before medical examination or dietary interview. In total, 54 students (27 boys and 27 girls) participated in the medical examination and dietary interview, and were eligible for DLW measurement. DLW dosages were given to 37 students, 17 subjects declined at the stage of DLW dosage and urine collection. All subjects and their parents gave written informed consent, and the study was approved by the Ethics Committee of the Medical Faculty at the University of Göteborg. Anthropometry A specially trained research nurse performed the clinical examination. Standing height was measured with a standardised wall measuring stick to the nearest 0.5 cm. Body weight in light clothing was measured, with digital weighing equipment (Seca, Germany), to the nearest 0.1 kg. The first weighing was conducted at the clinical examination visit and the second weighing at day 15 of DLW measurement. To define overweight and obesity, the International Obesity Task Force (IOTF) reference was used with age- and sexspecific cut points defined to pass through BMI 25 kg/m 2 at age 18 y (Cole et al, 2000). The cut points for age 15.5 y were used; for overweight boys BMI kg/m 2 and girls BMI kg/m 2, for obesity boys BMI kg/m 2 and girls BMI kg/m 2. Doubly labelled water Energy expenditure by the DLW method was measured over a period of 14 days. Sample analysis and calculation procedures have been described elsewhere (Slinde et al, 2003). First, a baseline urine sample was collected for the determination of background isotope enrichment (day 1). Then the subject ingested a weighed mixture of deuteriated and oxygenated water, corresponding to 0.05 g of deuterium oxide ( 2 H 2 O) and 0.10 g of oxygen-18-water (H 2 18 O)/kg body weight. Seven urine samples from each participant (days 2, 3, 4, 8, 13, 14 and 15) were collected in school by a member in the research group and exact voiding time was registered. Samples were analysed in triplicates on a Finnigan MAT Delta Plus Isotope-Ratio Mass Spectrometer (ThermoFinnigan, Uppsala, Sweden). TEE was calculated by the multipoint method by linear regression from the difference between elimination constants of deuterium and oxygen-18, with assumptions for fractionating and a respiratory quotient of 0.85 according to published recommendations from International Atomic Energy Agency (IAEA) (IDECG, 1990). The relationship between total body water estimated by deuterium (N D ) and total body water estimated by oxygen-18 (N O ), accepted range between and 1.060, was used as a

3 quality measurement of the DLW analysis as proposed by IAEA (IDECG, 1990). Diet history The questionnaire was introduced and completed in the classroom under supervision by our staff. The questionnaire elicited habitual meal pattern and consumption, frequency and amounts of common foods in household measures (Sjöberg et al, 2003). It was divided into sections to capture intake during different time periods of the school-day: intake in the morning before school, during lunch break and dinner in the afternoon/evening. Intake of foods, beverages, sweets and snacks that were eaten in other time periods than in main meals were defined as in-between meals. Subjects were asked to specify intake with fixed questions of frequency and amounts of common foods in household measures. There were open-ended questions to capture foods not specified in the questionnaire and for intake in in-between meals and during weekends. In the interview, the nutritionist asked for additional information about food choice, serving sizes and about frequencies of dishes for dinner. For evaluation of main meals, the quality of the intake was characterised. Intake before school was defined as breakfast, if it contained a cereal component and at least one milk product or one fruit/juice or a meat/fish/egg product. Lunch could be a prepared meal or bread meal in the school restaurant, at home or in a cafeteria. Dinner was a prepared warm or cold meal or a bread meal in the afternoon after school or in the evening. Sweet baked goods, cookies, cakes, ice cream and desserts, jam and sugar, sweets, nuts, potato crisps, popcorn, cheese doodles and soft drinks were defined as snack-food. We used the term snack-food to assess the quality of the intake. Snack-food could be consumed with inbetween meals as well as with main meals. In Sweden, school children are offered free lunch every school day. The school menu and information from the dining-hall matron were used to determine the composition of lunch dishes and common serving sizes. The intake of energy and nutrients was calculated using The Food Composition Table Energy and Nutrients (Swedish National Food Administration, 1996). Daily intake of foods, energy and nutrients, and the intake from energy in the different meals were calculated. Weight satisfaction The dietary questionnaire included a question with a fivelevel scale about perception of own body weight (too heavy, a little too heavy, satisfactory, a little too light, too light). The answers were divided into three categories, satisfactory, too heavy and too light. Calculations and statistics Basal metabolic rate (BMR) (Schofield, 1985) and BMI (kg/ m 2 ) were calculated from measured weight (weighing 1), height and for BMR also gender and age (10 18 y). The change of body weight was calculated from weighing 1 to weighing 2. Body weight, body height, BMI, changes in body weight and weight satisfaction (in five-level scale) were correlated with the ratio of EI TEE (EI/TEE) to investigate the relation to reporting accuracy. Subjects were identified as adequate- (AR), under- (UR) and over-reporters (OR) based on 95% confidence limits of the expected ratio EI/TEE, The 95% confidence limits between the two measurements were calculated from the equation (Black & Cole, 2000): 95% CL ¼2 p ½ðCV 2 EI =dþþðcv TEEÞ 2 Š CV EI is the coefficient of variation for daily EI (23%) (Bingham, 1987), but this factor disappeared since the number of days (d) are taken as infinity for methods aimed to capture habitual intake. The coefficient of variation, CV TEE, for repeated measurements of energy expenditure by the DLW, 8.2% (Black & Cole, 2000) was used 95% CL ¼2 p ½ðCV TEE Þ 2 Š¼16:4 According to this calculation subjects were defined as AR, UR and OR depending on how much EI/TEE deviated from the expected ratio AR was defined if the ratio EI/TEE was within the 95% confidence limits ( ), UR if below the lower 95% confidence limit (o0.84) and OR if above the upper 95% confidence limit (41.16). Individual and group data, mean and standard deviations (s.d.), were used to describe anthropometric, energy measures and for dietary intakes group data. Mean differences between EI and TEE and change of body weight were tested using the Student s t-test for paired samples. Mean differences in intake of energy from meals and energy-adjusted intake of foods as well as nutrients (in 10 MJ) between AR and UR or OR were analysed using the Student s t-test for independent samples. The Spearman rank-correlation coefficients were calculated for EI and TEE to analyse the ability of diet history to rank subjects according to energy needs. The Spearman rank-correlation coefficients were also calculated for the ratio EI/TEE and body weight, height, BMI, change in body weight and weight satisfaction to find out if reporting accuracy was related to these factors. The difference between TEE and EI was plotted against the mean of TEE and EI, as proposed by Bland and Altman (1986) to study individual agreement and the Pearson correlation coefficient was used for correlation between the mean and the difference. The tests were considered significant at Po0.05. Statistical calculations were performed using the SPSS for Windows (version 10.0) software programme. Results A total of 35 adolescents, 18 boys and 17 girls, age (0.4) y, completed the study with dietary assessment and DLW measurements. Two subjects were excluded due to incomplete urine sampling. The quality measurement of the 1645

4 1646 DLW analysis, the relationship between N D and N O, fell well into the accepted range. Time between dietary interview and the first day of DLW measurements was 12 (10) days. Mean EI was 11.0 (3.6) MJ and mean TEE was 11.4 (2.1) MJ, mean ratio EI/TEE was 0.96 (0.26). This 4% difference was nonsignificant (P¼0.42). The Spearman rank-correlation coefficient between reported intake and expenditure was r¼0.59 (Po0.001). For the girls, the group EI was 18% lower than the TEE (P¼0.0067) and for the boys, EI was 7% higher than TEE (P¼0.26). Of the subjects, 20 were identified as ARs, nine as URs (seven girls) and six as ORs (five boys). Detailed results for boys and girls are given in Tables 1 and 2. Mean BMR was 7.7 (0.6) MJ for the boys and 6.0 (0.4) MJ for the girls. BMI ranged from 17.3 to 27.2 kg/m 2, none of the adolescents were classified as obese. Three boys and three girls were classified as overweight according to the IOTF cut points. The Bland and Altman plot with each individual s difference between TEE and EI plotted against the mean of the methods, Figure 1, illustrates that the group means were close, EI was 0.41 MJ lower than TEE. However, both underand over-reporting occurred and for three individuals there were wide discrepancies. There was a trend (Pearson s correlation coefficient, r¼ 0.58, Po0.001) of higher degree of over-reporting for higher means of TEE and EI. The 95% limits of agreement for the difference between TEE and EI were 5.63 to 6.45 MJ. The ratio EI/TEE was not significantly correlated to any anthropometric measurement when all the subjects were analysed (Table 3). For the girls, body height was negatively correlated to the ratio EI/TEE. For the ratio EI/TEE and changes of body weight there was a significant correlation, mainly in the girls. There was no significant correlation between the ratio EI/TEE and BMI or weight satisfaction. However, BMI differed significantly in OR boys (P¼0.035) and UR girls (P¼0.010) compared to the AR boys and girls, respectively (Tables 1 and 2). In UR girls weight differed (P¼0.003), but not in OR boys. Change in body weight was þ 0.45 (1.26) kg (range 1.9 to þ 2.5 kg) (P¼0.042) between weighing 1 and weighing 2 over a period of 31.3 (12.2) days sexes combined. For boys there was a change in body weight of þ 0.82(1.39) kg (P¼0.022) and for girls there was no change in body weight (0.01(1.01) kg, P¼0.81). Every second participant found body weight satisfactory (n¼17) and onefourth considered body weight too heavy (n¼9) or too light (n¼9), respectively. Six AR boys (55%) and four AR girls (44%) were satisfied with their body weight. Reported food habits and reporting accuracy In total, 11 boys (61%) and six girls (35%) had breakfast, lunch and dinner every day. Irregular consumption of breakfast, lunch or dinner and three main meals on a daily basis in relation to reporting accuracy is shown in Figure 2 Table 1 Anthropometric and energy measures in boys, individual data and mean (s.d.) Subject EI/TEE Height (cm) Weight (kg) BMI (kg/m 2 ) TEE (MJ) EI (MJ) TEE/BMR a EI/BMR a Under-reporters b Adequate reporters b Over-reporters b All boys 1.08 (0.26) (9.0) 64.1 (8.7) 20.8 (2.5) 12.7 (1.7) 13.5 (3.2) 1.69 (0.21) 1.80 (0.48) Adequate reporters b 0.96 (0.08) (10.3) 61.8 (9.3) 19.8 (1.7) 12.9 (1.7) 12.4 (2.2) 1.75 (0.20) 1.69 (0.27) Over-reporters b 1.43 (0.22) (7.3) 66.8 (7.9) 22.7 (3.4) 12.0 (1.7) 17.1(2.9) 1.56 (0.21) 2.24 (0.45) a BMR, basal metabolic rate was calculated from weight, height, age and gender (Schofield, 1985). b Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR

5 Table 2 Anthropometric and energy measures in girls, individual data and mean (s.d.) 1647 Subject EI/TEE Height (cm) Weight (kg) BMI (kg/m 2 ) TEE (MJ) EI (MJ) TEE/BMR a EI/BMR a Under-reporters b Adequate reporters b Over-reporter b All girls 0.85 (0.20) (6.2) 56.4 (9.1) 20.8 (2.7) 10.1 (1.5) 8.3 (1.4) 1.67 (0.20) 1.39 (0.26) Adequate reporters b 0.95 (0.08) (5.8) 51.1 (5.4) 19.3 (2.3) 9.3 (1.0) 8.8 (0.9) 1.60 (0.15) 1.51 (0.13) Under-reporters b 0.66 (0.14) (6.3) 63.7 (8.8) 22.7 (2.2) 11.2 (1.4) 7.4 (1.3) 1.78 (0.21) 1.17 (0.21) a BMR, basal metabolic rate was calculated from weight, height, age and gender (Schofield, 1985). b Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR Figure 1 Difference between energy expenditure measured with DLW (TEE) and EI calculated from the diet history plotted against the mean of the two methods (n¼35), &¼boys (n¼18), m¼girls (n¼17). (girls) and Figure 3 (boys). The girl with the lowest ratio EI/ TEE regularly had lunch and dinner and she had 1 2 inbetween meals, but only had breakfast once a week. All boys and most of the girls always had lunch during schooldays, but during weekends fewer participants regularly had lunch (more pronounced in girls). Boys had dinner every day while three girls omitted dinner once a week. Girls omitted main meals more often than boys and UR girls more often omitted main meals than AR girls. The number of in-between meals was also lower in UR girls, 1.67 (0.49) times a day vs 1.81 (0.43) times a day in AR girls. OR boys had in-between meals 2.66 (0.53) times a day vs 2.02 (0.65) in AR boys. The difference of energy intake according to reporting accuracy was not evenly spread between the meals (Table 4). UR girls had 33% lower intake of energy from in-between meals compared to AR girls and OR boys had 57% higher intake of energy in in-between meals compared to AR boys. In percent of energy intake there were no significant differences of energy from meals according to reporting accuracy. For UR girls intake of bread, milk, potatoes, rice and pasta were lower, but the intake of snack-food, including sweets, was similar compared to AR girls (Table 5). The intake of juice, fruit and soft drinks were higher in the UR girls. Intake of protein, fat and carbohydrate was lower, vitamin C was higher, but sucrose, fibre, calcium, iron and zinc were similar in the UR compared to AR girls (Table 6). The OR boys had higher intakes of milk, bread, meat and snack-food, including soft drinks compared to AR boys (Table 5). The intake of potatoes, rice and pasta or fruit did not differ, but juice was lower. OR boys had higher values of all nutrients except for vitamin C (Table 6).

6 1648 Table 3 Relation of reporting accuracy a to anthropometric variables and weight satisfaction All subjects Boys Girls Variable r b P-value r b P-value r b P-value Body weight (kg) Body height (cm) Body mass index (kg/m 2 ) Weight change (kg) Weight satisfaction c a The ratio, EI/TEE. b Spearman s rank-correlation coefficients. c A five-level scale about perception of own body weight (1¼too heavy; 2¼a little too heavy; 3¼satisfactory, 4¼a little too light; 5¼too light). Intake of foods in 10 MJ in UR girls did not differ significantly (Table 5), but intake of fat was lower and intake of carbohydrate and vitamin C was higher compared to AR girls (Table 6). For OR boys, the only significant difference of foods in 10 MJ was intake of soft-drinks (Table 5). For nutrients in 10 MJ, there was no significant difference in the quality, but the values for sucrose were higher and vitamin C and iron lower in OR boys compared to AR boys (Table 6). Figure 2 Percent of girls with irregular intake (omitting the meal at least once a week) of breakfast, lunch or dinner and less than three main meals on a daily basis. URs (white) and ARs (checked). Figure 3 Percent of boys with irregular intake (omitting the meal at least once a week) of breakfast, lunch or dinner and less than three main meals on a daily basis. ORs (grey) and ARs (checked). Discussion The DH validated in this study was used to assess food habits in two school-based cross-sectional studies, including about 2000 adolescent boys and girls in Göteborg, Sweden. In the current study, habitual EI estimated with the DH method was validated with TEE measured with DLW in a smaller group of boys and girls. Mean level of estimated EI was not significantly different from mean TEE as measured by DLW, and the DH was able to rank subjects EI compared to TEE. The initial design of the two cross-sectional studies with dietary questionnaire, medical examination and dietary interview was identical in the current study, but the participants in the current study, were informed of all parts of the study including visiting the hospital and urine sampling at study initiation. In the current study the demands were larger on the participants and this probably affected the participation rate. The mean EI and TEE were close, but comparison of individual measures showed that the limits of agreement of the differences between energy measures were wide ( 5.6 to 6.6 MJ). In the current study, measurement periods were not overlapping, and this might widen the limits of agreement. Substantial variability has been reported in studies using overlapping measurement periods. In FFQ limits of agreement were 5 to 3 MJ (Andersen et al, 2003). In a British study, 16 adult women were included in a validation of EI by both DH and weighed record with the DLW method (Black et al, 2000). In the DH, the limits of agreement were 5.23 to 4.37 MJ and in the weighed record limits of agreement were 4.65 to 2.35 MJ. In an American study, validating a food frequency questionnaire aimed at youth

7 Table 4 Mean (s.d.) intake of energy and percentage of total energy intake from meals according to reporting accuracy 1649 Boys Girls Adequate reporters a (n¼11) Over-reporters a (n¼5) Student s t-test P-value Adequate reporters a (n¼9) Under-reporters a (n¼7) Student s t-test P-value Intake of energy (MJ) Breakfast 2.37 (0.93) 3.91 (2.10) (0.57) 1.17 (0.63) 0.61 Lunch 2.81 (0.58) 3.35 (0.67) (0.42) 1.73 (0.50) 0.62 Dinner 3.63 (1.00) 4.13 (1.06) (0.33) 2.35 (0.44) 0.62 In-between meals 3.63 (1.73) 5.70 (0.95) (0.57) 2.11 (0.64) Percentage of energy intake (%) Breakfast 19.0 (6.6) 22.0 (9.2) (5.7) 15.2 (6.4) 0.93 Lunch 23.0 (5.3) 20.0 (5.0) (4.5) 23.6 (6.3) 0.37 Dinner 29.2 (6.7) 24.3 (5.1) (3.5) 28.3 (5.0) 0.11 In-between meals 28.8 (12.4) 33.7 (5.0) (3.7) 29.1 (9.7) a Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR Table 5 Intake of foods in UR- and ORs compared to adequate reporters Boys Girls Adequate reporters a Over-reporters a Student s t-test Adequate reporters a Under-reporters a Student s t-test Mean (s.d.) (n¼11) Mean (s.d.) (n¼5) P-value Mean (s.d.) (n¼9) Mean (s.d.) (n¼7) P-value Foods (g) Milk 747 (558) 905 (462) 352 (249) 326 (306) Bread 153 (52) 269 (121) 123 (39) 100 (25) Potatoes, rice and pasta 300 (106) 307 (79) 210 (62) 170 (57) Meat 145 (41) 198 (68) 96 (56) 75 (29) Snack-food 418 (272) 989 (419) 363 (170) 402 (267) Sweets 39 (39) 74 (32) 42 (21) 39 (20) Soft drinks 293 (264) 802 (354) 238 (146) 294 (264) Fruit 91 (83) 117 (108) 72 (45) 98 (66) Juice 244 (239) 109 (149) 75 (131) 159 (219) Energy adjusted intake (10 MJ): Milk 600 (422) 543 (302) (268) 413 (338) 0.95 Bread 121 (31) 154 (56) (39) 139 (40) 0.97 Potatoes, rice and pasta 239 (55) 182 (41) (81) 237 (86) 0.91 Meat 117 (26) 114 (27) (60) 103 (42) 0.89 Snack-food 337 (213) 589 (273) (172) 563 (416) 0.33 Sweets 34 (36) 45 (23) (24) 57 (32) 0.53 Soft drinks 232 (200) 475 (222) (151) 419 (410) 0.32 Fruit 71 (64) 70 (67) (47) 145 (119) 0.17 Juice 178 (158) 61 (85) (143) 205 (282) 0.28 a Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR studies in which food habits covered the last year, the limits of agreement were 6.3 to 6.7 MJ (Perks et al, 2000). Invalid estimate of habitual EI is one source of variation for the individual comparison, but also the DLW measurements include individual variability. It has been shown for 7-day records that during the food-record period the main part of the under-reporting was due to under-eating (Goris & Westerterp, 1999). During the DLW measurement, consciousness of being under observation might also have an effect on physical activity. A group mean EI in level with TEE combined with an ability to rank the subjects as we have shown in this study is what can be expected from comparisons of this kind of measurements, although a perfect agreement for all subjects would be desirable. Both

8 1650 Table 6 Intake of nutrients in UR- and OR compared to adequate reporters Boys Girls Adequate reporters a Over-reporters a Student s t-test Adequate reporters a Under-reporters a Student s t-test Mean (s.d.) (n¼11) Mean (s.d.) (n¼5) P-value Mean (s.d.) (n¼9) Mean (s.d.) (n¼7) P-value Nutrients Protein (g) 117 (25) 146 (31) 75 (12) 60 (14) Fat (g) 103 (18) 151 (39) 72 (10) 52 (16) Carbohydrate (g) 388 (81) 526 (67) 282 (29) 258 (38) Sucrose (g) 73 (17) 128 (34) 63 (25) 66 (17) Fibre (g) 22 (6) 28 (12) 15 (4) 14 (2) Vitamin C (mg) 120 (74) 97 (60) 59 (33) 98 (52) Calcium (mg) 1630 (600) 2020 (560) 970 (290) 840 (390) Iron (mg) 14.4 (3.6) 17.5 (4.0) 9.3 (1.6) 8.2 (2.1) Zinc (mg) 16.0 (3.4) 20.6 (4.6) 10.5 (2.5) 8.3 (1.8) Energy adjusted intake (10 MJ) Protein (g) 94 (10) 86 (10) (9) 81 (11) 0.47 Fat (g) 83 (9) 87 (9) (8) 70 (11) Carbohydrate (g) 311 (16) 310 (16) (13) 354 (32) Sucrose (g) 60 (16) 77 (26) (27) 91 (29) 0.20 Fibre (g) 17 (2) 16 (6) (3) 20 (4) 0.14 Vitamin C (mg) 91 (47) 56 (35) (36) 135 (68) Calcium (mg) 1310 (410) 1190 (350) (340) 1120 (400) 0.96 Iron (mg) 11.6 (1.9) 10.2 (1.9) (1.4) 11.1 (2.2) 0.57 Zinc (mg) 12.8 (1.0) 12.0 (1.5) (2.1) 11.2 (1.5) 0.48 a Adequate (AR), under- (UR) and over-reporters (OR) were defined from the ratio EI/TEE: AR , UR o0.84 and OR energy measures include variation and perfect agreement cannot be expected even for valid data (Livingstone & Black, 2003). BMI was not correlated to the ratio EI/TEE. Both OR boys and UR girls had higher BMI compared to AR boys and girls, respectively. An association between under-reporting and BMI has been found in numerous other studies on adults (Stallone et al, 1997; Braam et al, 1998; Kretsch et al, 1999). In normal weight adolescent populations, negative associations were found between energy reporting and BMI (Livingstone et al, 1992; Bratteby et al, 1998). The distribution of BMI in the current study with low prevalence of overweight and no obese subjects could explain the lack of correlation between the ratio EI/BMR to BMI. However, distribution of BMI and prevalence of overweight in the current study agreed well with the two cross-sectional studies, prevalence of overweight was 10 13% in both populations and prevalence of obesity was about 1 2% (Hulthén et al, 2003; Sjöberg et al, 2003). Our finding that OR boys also had higher BMI than AR boys is interesting. Few studies presenting data about over-reporting have been published, but adult male overreporters, identified with a ratio of EI/BMR 42.4, had lower BMI compared to adequate reporting men in a Norwegian study (Johansson et al, 1998). It has been proposed that questions to identify weight-conscious individuals could be used for evaluation of reported EI (Goldberg & Black, 1998). The answers concerning perception of body weight were, however, not conclusive for that purpose in the current study. The change in body weight during the study period occurred mainly among the boys. Correlation between the ratio EI/TEE and change in body weight occurred among the girls. This correlation was accentuated by the fact that five of the UR girls had lost weight and the OR girl had gained weight during the study period. If we assume that the difference in body weight entirely was due to change in fat mass, it would correspond to 0.44 (1.43) MJ per day (Westerterp et al, 1995). However, variation of body weight is expected depending on circadian changes in water balance and, in the girls, due to hormonal variations during the menstrual cycle. Since no measurements of body composition were conducted, it would be inappropriate to correct for changes in body weight. Reported food habits and reporting accuracy Misreporting could be due to inadequately reported number of meals or frequencies or amounts of foods. In the current study, there were small differences in frequency of main meals according to reporting accuracy. UR girls more often omitted main meals, and also the same subject omitted more than one type of main meal, compared to AR girls (Figure 2) and compared to the boys (Figure 3). Comparing energy from meals, according to reporting accuracy, showed that misreporting differed between the meals. The 16% lower EI in UR girls compared to AR girls, ranged from 4% lower intake with dinner to 33% lower with in-between meals. For OR boys the 38% higher EI varied between 14% higher

9 intake with dinner to 65% higher intake with breakfast. It was only the EI with in-between meals that differed significantly also reported from a study among young women (Poppitt et al, 1998). This was important methodologically since under- and over-reporting were less pronounced for lunch and dinner. Lunch and dinner should be the most important meals for the primary objective for The Göteborg Adolescence Studies, that is, intake of iron and how it affected the iron situation. Under-reporting is common in female populations ranging from 10% in a food frequency questionnaire in healthy young Norwegian students (Andersen et al., 2003) to 21 to 31% when food records were used in young women (Bratteby et al, 1998; Hill & Davies, 1999; Weber et al, 2001). For the girl who reported the lowest energy intake, fewer meals could not be the only explanation for the low EI/ BMR and EI/TEE, she must also have reported fewer foods or smaller portions in the reported meals. This girl would have been identified as an UR also with the Goldberg cutoff limits (Goldberg et al, 1991), but her high TEE accentuated the under-reporting. UR girls reported a healthier diet than AR girls, which has also been observed in population studies where under-reporting was defined from the ratio EI/BMR. In a Finnish study of adults, under-reporters reported a significantly higher proportion of vegetables, fish, meat, potatoes, fruit and less fat (Hirvonen et al, 1997). In a Canadian study, under-reporters reported less fat, more fruit and vegetables and fewer sweets (Pomerleau et al, 1999). One study reported a relation between social desirability and intake of vegetables and fruit, but negative correlation with snack-food in adult women (Worsley et al, 1984). The UR girls in the current study had similarities with these observations since the quality of the intake was lower in fat and higher in vitamin C; however, intake of sweets and soft-drinks was not lower. Adult male over reporters, identified with a ratio of EI/BMR 42.4, had a higher percentage of energy from fat and sugar, and in 10 MJ a lower intake of bread and potatoes and higher intake of soft drinks compared to adequate reporters in a Norwegian study (Johansson et al, 1998). In the current study, only the higher soft-drink intake in 10 MJ in OR boys had similarities with this observation. To the best of our knowledge no validation study has provided information about meals, foods and nutrients in relation to reporting accuracy. Since validation studies with DLW have often included few subjects, power might be too low to detect differences between UR and OR to AR. To show that OR boys in energy-adjusted intake (10 MJ) had significantly lower intake of potatoes, rice and pasta and higher intake of snack-food at the 80% level would require twice as many participants. To show that UR girls had a significantly higher proportion of sucrose in their diet based on the values shown in Table 5 would require 28 UR and 36 AR subjects to reach a power of 76%. DLW measurement was assumed to capture habitual TEE and since the DH was expected to capture habitual EI there was no overlapping of measurement periods for EI and TEE. The CV for TEE of 8.2 used for calculation of the 95% confidence limits to identify URs- and ORs include analytic and biologic variation when measures are performed simultaneously, but no variation due to changed activity level. Time span between DH and DLW measurement was for some subjects a month. Regression equations of repeated DLW measurements indicate that the CV for the TEE measurement increases with longer time span and for 4 weeks the CV TEE would instead be 8.6 (Black & Cole, 2000). However, the use of CV TEE 8.6 for DLW would not change the classification of under- and over-reporters in the current study. This study showed that for dietary assessment in adolescents with a structured school-day, the DH method worked well. At the age of 16 y, the adolescents are expected to be able to report food habits. To use information about the school-lunch facilitated data collection and resulted in lower demands on the participants. The DH was time-consuming, but since all information was collected face to face with the subject, it had many advantages for this age group compared to food records or repeated 24-h recall. Food records require homework and often result in change of eating behaviour during the registration period and 24-h recalls require several contacts. These factors limit the usefulness of these dietary methods on adolescents. In this study, EI and TEE were compared based on the assumption that both assessed the habitual energy levels. We showed that DH as performed in this study was valid to assess habitual EI for the group and for ranking of subjects. For a few individuals there were large differences between the two measurements. Reporting accuracy was related to gender and comparison of UR- and ORs with ARs showed that there was a difference in the quality of the reported intake. Acknowledgements We thank all subjects for their cooperation and Mrs Annica Alklind for performing the anthropometric measurements. References Andersen LF, Tomten H, Haggarty P, Lovo A & Hustvedt BE (2003): Validation of energy intake estimated from a food frequency questionnaire: a doubly labelled water study. Eur. J. Clin. Nutr. 57, Bandini LG, Schoeller DA, Cyr HN & Dietz WH (1990): Validity of reported energy intake in obese and nonobese adolescents. Am. J. Clin. Nutr. 52, Bingham S (1987): The dietary assessment of individuals: methods, accuracy, new techniques and recommendations. Nutr. Abst. Rev. 57, Black AE & Cole TJ (2000): Within- and between-subject variation in energy expenditure measured by the doubly-labelled water technique: implications for validating reported dietary energy intake. Eur. J. Clin. Nutr. 54, Black AE, Welch AA & Bingham SA (2000): Validation of dietary intakes measured by diet history against 24 h urinary nitrogen excretion and energy expenditure measured by the doubly- 1651

10 1652 labelled water method in middle-aged women. Br. J. Nutr. 83, Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, Braam LA, Ocke MC, Bueno-de-Mesquita HB & Seidell JC (1998): Determinants of obesity-related underreporting of energy intake. Am. J. Epidemiol. 147, Bratteby LE, Sandhagen B, Fan H, Enghardt H & Samuelson G (1998): Total energy expenditure and physical activity as assessed by the doubly labeled water method in Swedish adolescents in whom energy intake was underestimated by 7-d diet records. Am. J. Clin. Nutr. 67, Cole TJ, Bellizzi MC, Flegal KM & Dietz WH (2000): Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 320, Goldberg GR & Black AE (1998): Assessment of the validity of reported energy intakes - review and recent developments. Scand. J. Nutr./Näringsforskning 42, 6 9. Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR, Coward WA & Prentice AM (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, Goris AH & Westerterp KR (1999): Underreporting of habitual food intake is explained by undereating in highly motivated lean women. J. Nutr. 129, Hebert JR, Ebbeling CB, Matthews CE, Hurley TG, Ma Y, Druker S & Clemow L (2002): Systematic errors in middle-aged women s estimates of energy intake: comparing three self-report measures to total energy expenditure from doubly labeled water. Ann. Epidemiol. 12, Hebert JR, Ma Y, Clemow L, Ockene IS, Saperia G, Stanek III EJ, Merriam PA & Ockene JK (1997): Gender differences in social desirability and social approval bias in dietary self-report. Am. J. Epidemiol. 146, Hill RJ & Davies PS (1999): The validity of a four day weighed food record for measuring energy intake in female classical ballet dancers. Eur. J. Clin. Nutr. 53, Hirvonen T, Mannisto S, Roos E & Pietinen P (1997): Increasing prevalence of underreporting does not necessarily distort dietary surveys. Eur. J. Clin. Nutr. 51, Hulthén L, Hallberg L, Albertsson-Wikland K, Bondestam M, Ekblom B, Höglund D, Lindstedt G, Sjöberg A & Slinde F (2003): Height and weight in two representative samples of Swedish adolescents. Eur. J. Clin. Nutr. (in press). IDECG (1990): The Doubly-labelled Water Method for Measuring energy Expenditure. Technical Recommendations for use in Humans. Vienna: International Atomic Energy Agency (IAEA). Johansson L, Solvoll K, Bjorneboe GE & Drevon CA (1998): Under- and overreporting of energy intake related to weight status and lifestyle in a nationwide sample. Am. J. Clin. Nutr. 68, Kretsch MJ, Fong AK & Green MW (1999): Behavioral and body size correlates of energy intake underreporting by obese and normalweight women. J. Am. Diet. Assoc. 99, : quiz Livingstone MB & Black AE (2003): Markers of the validity of reported energy intake. J. Nutr. 133(Suppl 3), 895S 920S. Livingstone MB, Prentice AM, Coward WA, Strain JJ, Black AE, Davies PS, Stewart CM, McKenna PG & Whitehead RG (1992): Validation of estimates of energy intake by weighed dietary record and diet history in children and adolescents. Am. J. Clin. Nutr. 56, Livingstone MB & Robson PJ (2000): Measurement of dietary intake in children. Proc. Nutr. Soc. 59, Perks SM, Roemmich JN, Sandow-Pajewski M, Clark PA, Thomas E, Weltman A, Patrie J & Rogol AD (2000): Alterations in growth and body composition during puberty. IV. Energy intake estimated by the youth-adolescent food-frequency questionnaire: validation by the doubly labeled water method. Am. J. Clin. Nutr. 72, Pomerleau J, Ostbye T & Bright-See E (1999): Potential underreporting of energy intake in the Ontario Health Survey and its relationship with nutrient and food intakes. Eur. J. Epidemiol. 15, Poppitt SD, Swann D, Black AE & Prentice AM (1998): Assessment of selective under-reporting of food intake by both obese and nonobese women in a metabolic facility. Int. J. Obes. Relat. Metab. Disord. 22, Schofield WN (1985): Predicting basal metabolic rate, new standards and review of previous work. Hum. Nutr. Clin. Nutr. 39, Sjöberg A, Hallberg L, Höglund D & Hulthén L (2003): Meal pattern, food choice, nutrient intake and lifestyle factors in The Göteborg Adolescence Study. Eur. J. Clin. Nutr. 57, Slinde F, Ellegård L, Grönberg AM, Larsson S & Rossander-Hulthén L (2003): Total energy expenditure in underweight patients with severe chronic obstructive pulmonary disease living at home. Clin. Nutr. 22, Stallone DD, Brunner EJ, Bingham SA & Marmot MG (1997): Dietary assessment in Whitehall II: the influence of reporting bias on apparent socioeconomic variation in nutrient intakes. Eur. J. Clin. Nutr. 51, Swedish National Food Administration (1996): The Food Composition Table - Energy and Nutrients. Uppsala, Sweden: Swedish National Food Administration. Weber JL, Reid PM, Greaves KA, DeLany JP, Stanford VA, Going SB, Howell WH & Houtkooper LB (2001): Validity of self-reported energy intake in lean and obese young women, using two nutrient databases, compared with total energy expenditure assessed by doubly labeled water. Eur. J. Clin. Nutr. 55, Westerterp KR, Donkers JH, Fredrix EW & Boekhoudt P (1995): Energy intake, physical activity and body weight: a simulation model. Br. J. Nutr. 73, Worsley A, Baghurst KI & Leitch DR (1984): Social desirability response bias and dietary inventory responses. Hum. Nutr. Appl. Nutr. 38,