Relationship of High Energy Expenditure and Variation in Dietary

Home , 1100s (decade)

European Journal of Clinical Nutrition (2002) 56, 358–367 ß 2002 Nature Publishing Group All rights reserved 0954–3007/02 $25.00 www.nature.com/ejcn ORIGINAL COMMUNICATION Relationship of high energy expenditure and variation in dietary intake with reporting accuracy on 7 day food records and diet histories in a group of healthy adult volunteers

JA Barnard1*, LC Tapsell1, PSW Davies2, VL Brenninger1 and LH Storlien1

1Department of Biomedical Sciences and Smart Foods Centre, University of Wollongong, Wollongong, New South Wales, Australia; and 2School of Human Movement Studies, Faculty of Health, Queensland University of Technology, Brisbane, and Children’s Nutrition Research Centre, Department of Paediatrics and Child Health, University of Queensland, Royal Children’s Hospital, Herston, Brisbane, Queensland, Australia

Objectives: To assess the accuracy of reporting from both a diet history and food record and identify some of the characteristics of more accurate reporters in a group of healthy adult volunteers for an energy balance study. Design: Prospective measurements in free-living people. Setting: Wollongong, Australia. Subjects: Fifteen healthy volunteers (seven male, eight female; aged 22 – 59 y; body mass index (BMI) 19 – 33 kg=m2) from the local community in the city of Wollongong, Australia. Interventions: Measurement of energy intake via diet history interview and 7 day food records, total energy expenditure by the doubly labelled water technique over 14 days, physical activity by questionnaire, and body fat by dual-energy X-ray absorptiometry. Results: Increased misreporting of energy intake was associated with increased energy expenditure (r ¼ 0.90, P < 0.0001, diet history; rS ¼ 0.79, P ¼ 0.0005, food records) but was not associated with age, sex, BMI or body fat. Range in number of recorded dinner foods correlated positively with energy expenditure (rS ¼ 0.63, P ¼ 0.01) and degree of misreporting (rS ¼ 0.71, P ¼ 0.003, diet history; rS ¼ 0.63, P ¼ 0.01, food records). Variation in energy intake at dinner and over the whole day identiﬁed by the food records correlated positively with energy expenditure (r ¼ 0.58, P ¼ 0.02) and misreporting on the diet history (r ¼ 0.62, P ¼ 0.01). Conclusions: Subjects who are highly active or who have variable dietary and exercise behaviour may be less accurate in reporting dietary intake. Our ﬁndings indicate that it may be necessary to screen for these characteristics in studies where accuracy of reporting at an individual level is critical. Sponsorship: The study was supported in part by Australian Research Council funds made available through the University of Wollongong. European Journal of Clinical Nutrition (2002) 56, 358 – 367. DOI: 10.1038=ej=ejcn=1601341

Keywords: energy intake; energy metabolism; diet records; intervention studies

*Correspondence: J Barnard, Department Biomedical Sciences, University and statistical procedures on resulting data. LHS provided conceptual of Wollongong, NSW 2522, Australia. and statistical guidance. LCT prepared the initial draft of the E-mail: [email protected] manuscript, which was worked up and finished by JAB. VLB made Guarantor: L Tapsell. editorial contributions. LCT and LHS were responsible for the overall Contributors: JAB and VLB recruited subjects, analysed food record supervision of the project. and diet history data and collected urine samples for analysis. PSWD Received 9 August 2000; revised 29 August 2001; analysed the urine samples and estimated EE. JAB carried out analysis accepted 6 September 2001 Reporting accuracy on 7 day food records JA Barnard et al 359 Introduction portion sizes were noted. Dietary data were entered into the Accurate reporting of energy intake is critical in studies that Foodworks (v. 1.05) nutrient analysis software packages examine the effects of energy balance and macronutrient (Xyris software, Brisbane, Australia), containing the Austra- fuel mix on human metabolism. Whilst studies under con- lian nutrient database, NUTTAB95 (Lewis et al, 1995). trolled conditions contribute substantially, it is the long- term clinical trials in ‘free-living’ humans which provide the best evidence for dietary management. In clinical stu- Total energy expenditure dies, energy and macronutrient intakes are appropriately Total EE was assessed over 14 days using the DLW technique. assessed using the diet history method (DH) and=or food On day 0, each participant consumed an oral dose of 2 18 records (FR) which rely on self-report. The doubly labelled 0.05 g=kg body weight of H2O and 0.13 g=kg H2 O. Urine water (DLW) technique acts as a gold standard against which samples were collected on days 0, 1 and 14 for measurement reported energy intakes can be validated in weight-stable of enrichment of each isotope using isotope ratio mass conditions (Black et al, 1993). spectrometry (Hydra, Europa Scientific Ltd, UK). Since the Studies involving the DLW technique have shown that 2H leaves the body in water and the 18O in both water and misreporting of energy intakes is widespread (Black et al, exhaled carbon dioxide, the difference in the rate of disap- 1991), although a number of researchers identify overweight pearance from the body of the two isotopes in urine was used subjects as being more likely to report inaccurately (Bingham to estimate carbon dioxide production. EE was then esti- et al, 1995; Braam et al, 1998; Heitmann & Lissner, 1995; mated using indirect calorimetry equations. Changes in Lafay et al, 1997). While in the latter case psychosocial body weight were incorporated into the assessment using factors may be involved (Herbert et al, 1995; Mela & Aaron, estimated figures for weight loss and gain (FAO=WHO=UNU 1997), there will be other contextual aspects of reporting Expert Consultation, 1985). dietary intake which may affect all subgroups. The aim of the study reported here was to assess the accuracy of reporting from both a diet history and food record in a group of Body composition healthy adult volunteers and to identify some of the char- Percentage body fat (BF) was measured up to 3 weeks prior to acteristics of more accurate reporters with particular empha- day 0 by dual-energy X-ray absorptiometry with a Norland sis on variation in the lifestyle parameters of eating and XR Bone Densitometer (Norland XR-36, Norland Corpora- physical activity behaviour. tion, Fort Atkinson, WI, USA) using host software revision 2.5.2 and scanner software revision 1.3.1 (Norland Corpora- tion, Fort Atkinson, WI, USA). The resolution (pixel size) and Methods scan speed used were 6.5Â13.0 mm and 180 mm=s, respec- Subjects tively. The whole body entrance skin radiation dose at this Fifteen normal healthy adults (seven male, eight female; resolution and speed is 0.07 millrems. aged 22 – 59 y, body mass index (BMI) 19 – 33 kg=m2) from the local community in the city of Wollongong, Australia volunteered for the study. BMI was assessed as weight (kg) Assessment of reporting accuracy divided by height (m2). Digital scales (Soehnle, Germany) Reporting accuracy was determined by calculating a ratio of were used to measure weight to the nearest 0.1 kg and a energy intake (EI) to the DLW value (EEDLW). Subjects with stadiometer (SECA, model no. 220, Germany) to measure discrepancy values within the range 0.79 EI:EEDLW 1.21 height to the nearest 0.1 cm. The project was approved (Black, 1997) or with EI within 21% EEDLW using either DH by the University of Wollongong Human Research Ethics or FR method were classified as accurate reporters and the Committee. remaining subjects were classified as inaccurate reporters.

Dietary assessment Lifestyle characteristics Subjects underwent open-ended DH interviews with trained To examine the effect of variation in lifestyle parameters on dietitians at baseline (day 0) and kept a weighed FR for the reporting accuracy, FR and physical activity data were further following 7 days (days 1 – 7) during the period of energy analysed. To characterise meal-time variability, the dinner expenditure (EE) measurement (days 0 – 14). In the DH sub- meal from the FR was examined by classifying foods into jects were asked to describe their usual daily eating patterns groups. Dinner was chosen as the reference meal due to including variations (for example weekends or social events), greater variation in food choices compared to lunch and and to complete a food checklist at the end of this descrip- breakfast (Tapsell et al, 2000). To capture the additional tion. They were given full instructions for keeping food complexity of reporting mixed meals, meat-, seafood- and records and were supplied with measuring cups, spoons vegetable-based dishes were separated according to whether and kitchen scales and recording sheets. Food records were they were served as discrete portions, or components checked by the dietitian, and details regarding recipes and of mixed dishes such as stews and stir-fry meals. Major

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 360 components of each dinner meal (excluding dessert foods, of the mean of the sum of EI from the FR and DH. EE using

snack items and drinks) were then counted each day for each the MAQ (EEMAQ) was calculated by a factorial method using subject. Mixed dishes were weighted by a factor of 4 to reﬂect BMR factors for each activity type from FAO=WHO=UNU

their complexity. Expert Consultation (1985). The relationship of EEMAQ with The modifiable activity questionnaire (MAQ) adapted EEDLW, and EI from DH and FR was examined using regres- from Kriska (1997) was used to obtain type and quantity of sion analysis. leisure-time activities and occupation (including home All data were analysed using the statistical package JMP v. duties) over the previous year. The questionnaire was mod- 3.1.6.2 (SAS Institute Inc., Cary, NC, USA). Statistical signifi- ified to include leisure activities appropriate to Australian cance was accepted at the 95% confidence level (P < 0.05). conditions as indicated in the questionnaire instructions, to Data are presented as mean (s.d.). ask subjects for an estimate of intensity of leisure activities, to include more general questions to assess sedentary behaviour replacing those that assessed extremes of inactivity (eg Results surgery), and to exclude a question on lifetime involvement Subject characteristics and EI and EE data can be found in in team or individual sports. The MAQ was also used to Tables 1 and 2. Of the 15 subjects, seven were accurate

determine EE (EEMAQ) to trial it as a cheap and simple reporters (five males, two females; age 38.7 (4.9), BMI 24.7 alternative to DLW for the assessment of EE in clinical trials. (1.6), BF 29.5 (13.4) and eight were inaccurate (five females and three males; age 34.0 (3.5), BMI 25.0 (1.8), BF 29.7 (10.4); Table 3). The EE of accurate reporters was 11.49 MJ Statistics (0.83), and 20.45 MJ (1.57) for inaccurate reporters. There t-Tests were used to investigate significant differences was no significant difference between the groups for gender

between the accurate and inaccurate reporters for age, BMI, proportions, age, BMI or BF, but the difference for EEDLW was BF and EE, and for sex, chi-squared tests were applied. The signiﬁcant (P ¼ 0.0003). There was a signiﬁcant positive

relationships between reported energy intakes (DH and FR) relationship between the degree of misreporting and EEDLW

and the ‘gold standard’ for EE (EEDLW) were examined by for both the DH (r ¼ 0.90, P < 0.0001; Figure 1A) and FR regression analysis. The relationship between the degree of (rS ¼ 0.79, P ¼ 0.0005; Figure 1B). Figure 1B indicated two misreporting deﬁned as the absolute difference between EI distinct clusters, which we then re-examined on a plot of

and EEDLW, and EEDLW was investigated. The relationship real difference between EI(FR) and EEDLW against EEDLW to between the degree of misreporting with BMI, BF and with expose the distinctive difference in misreporting when

age was also investigated. EEDLW is more than 15 MJ (Figure 2). There was no A food range score for dinner foods was determined by significant relationship between degree of misreporting and calculating the difference between the minimum and max- BMI (Figure 3) or BF (Figure 4), however, there was a imum number of foods eaten at dinner over the 7 days of recording for the FR. The food range score as well as the number of leisure-time, occupational and total (leisure-time plus occupational) activities was compared to age, BMI, BF, Table 1 Subject characteristics EE , and the degree of misreporting using regression DLW Subject Age BMI %BF Weight change (kg) analysis. The comparisons with sex were made using t-tests. In addition, intra-individual variation in number of foods Women 7 eaten at dinner and energy intake over the whole day and at 1 38 24.1 52.5 1.4 2 39 33.4 43.5 0.8 dinner only (excluding dessert foods, snack items and 3 41 19.1 38.4 7 0.6 drinks) was assessed using the coefficient of variation 4 25 20.3 27.4 7 0.8 (%CV ¼ s.d.=meanÂ100%) for the 7 days of recording 5 24 19.5 26.9 1.6 6 42 28.2 41.4 2.8 intake. Subjects were divided into younger ( 28 y) and 7 51 21.9 38.8 7 1.6 older ( 36 y) age groups and the groups were compared Mean (s.d.), n ¼ 7 37.1 (9.6) 23.8 (5.3) 38.4 (9.0)a 0.1 (1.7) via t-tests for differences in %CV and number of leisure, Men occupational, and sum of leisure and occupational (total) 8 59 32.7 33.9 7 0.2 activities. 9 39 27.0 24.6 7 0.2 Multiple regression analysis was performed to determine 10 28 23.5 14.3 1.0 which factors accounted for variation in degree of misreport- 11 26 20.1 14.4 7 0.4 12 22 23.2 16.0 1.0 ing on the FR and DH. 13 49 26.9 25.0 7 1.0 For the MAQ, time spent per day in light, moderate and 14 36 24.2 26.1 7 1.2 heavy leisure and occupational activities as well as for sleep 15 24 29.2 21.0 1.8 a and inactivity and miscellaneous activities was estimated. Mean (s.d.), n ¼ 8 35.4 (13.1) 25.9 (3.9) 21.9 (6.8) 0.1 (1.1) Mean (s.d.), n ¼ 15 36.2 (11.7) 24.9 (4.6) 29.6 (11.4) 0.1 (1.3) Basal EE was estimated using standard equations (Schofield et al, 1985) and thermic effect of food was estimated as 10% aSignificantly different (P ¼ 0.0015).

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 361 Table 2 Individual energy intake and expenditure data

Energy intake Energy expenditure

Subject DH FR DLW MAQ

Women 1 5344 6632 8222 9918 2 9253 10 372 17 108 9689 3 7699 7668 16 506 7905 4 6098 7447 19 797 8802 5 8949 10 812 17 162 8500 6 7501 6719 18 242 NA 7 7522 8056 9289 NA Mean (s.d.), n ¼ 7 7480.9 (1403.7)a 8243.7 (1685.7)b 15 189.4 (4531.7) 8962.8 (836.4)c

Men 8 10 747 12 693 11 780 19 561 9 17 266 12 439 12 060 12 835 10 12 813 10 922 24 267 13 984 11 22 024 13 196 12 676 15 024 12 22 470 13 623 14 938 14 815 13 11 953 14 971 21 141 NA 14 9475 9760 11 477 NA 15 11 471 14 627 29 409 NA Mean (s.d.), n ¼ 8 14 777.3 (5140.9)a 12 778.8 (1764.4)b 17 218.5 (6857.9) 15 243.6 (2562.3)c Mean (s.d.), n ¼ 15 11 372.3 (5315.6) 10 662.4 (2873.8) 16 271.6 (5777.0) Mean (s.d.), n ¼ 10 12 103.2 (3766.5)

aSignificantly different (P ¼ 0.005). bSignificantly different (P ¼ 0.0005). cSignificantly different (P ¼ 0.001).

Table 3 Individual reporting accuracy data

Degree of mis-reportinga Percentage difference from DLW Percentage difference from MAQ

Subject DH FR DH FR DH FR

Women 1 2878 1590 7 35.00 7 19.34 7 46.12 7 33.13 2 7855 6736 7 45.92 7 39.37 7 4.50 7.05 3 8807 8838 7 53.36 7 53.54 7 2.60 7 3.00 4 13 699 12 350 7 69.20 7 62.38 7 30.72 7 15.39 5 8213 6350 7 47.86 7 37.00 5.27 27.19 6 10 741 11 523 7 58.88 7 63.17 NA NA 7 1767 1233 7 19.02 7 13.27 NA NA Mean (s.d.), n ¼ 7 7708.6 (4187.3) 6945.7 (4386.6) 7 47.0 (16.3)b 7 41.2 (19.8)c 7 15.7 (21.7) 7 3.5 (22.7)

Men 8 1033 913 7 8.77 7.75 7 45.06 7 35.11 9 5205 378 43.16 3.14 34.52 7 3.08 10 11 454 13 346 7 47.20 7 54.99 7 8.37 7 21.90 11 9348 519 73.74 4.10 46.59 7 12.17 12 7532 1315 50.43 7 8.80 51.67 7 8.04 13 9188 6170 7 43.46 7 29.18 NA NA 14 2002 1717 7 17.45 7 14.96 NA NA 15 17 938 14 782 7 60.99 7 50.26 NA NA Mean (s.d.), n ¼ 8 7962.5 (5430.6) 4892.4 (5965.4) 7 1.3 (50.1)b 7 17.9 (24.6)c 15.9 (41.5) 7 16.1 (12.7) Mean (s.d.), n ¼ 15 7844.0 (4719.9) 5850.6 (5211.9) 7 22.7 (44.3) 7 28.8 (24.8) Mean (s.d.), n ¼ 10 0.1 (35.4) 7 9.8 (18.6)

Note: accurate reports are indicated in bold. aAbsolute difference between energy intake and energy expenditure as measured by DLW in kJ. bSigniﬁcantly different (P < 0.05).

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 362 signiﬁcant relationship between age and degree of misreport- (r ¼ 0.58, P ¼ 0.02) and degree of misreporting on the DH ing for the DH (r ¼ 7 0.61, P ¼ 0.015). There was no differ- (r ¼ 0.62, P ¼ 0.01), and negatively with age (r ¼ 7 0.60, ence in degree of misreporting between males and females P ¼ 0.02). Variation in EI over the whole day correlated

for either DH or FR. positively degree of misreporting on the DH (rS ¼ 0.53, Dinner food groups used for analysis are listed in Table 4 P ¼ 0.04), and negatively with age (rS ¼ 7 0.81, P ¼ 0.0003), and variation data relating to energy intake along with but correlations did not quite reach signiﬁcance for degree of

detailed physical activity data can be found in Table 5. misreporting on the FR (rS ¼ 0.50, P ¼ 0.06) or EEDLW There was no relationship between the food range score (rS ¼ 0.44, P ¼ 0.10). When grouped by age, younger subjects and age, sex, BMI or BF, but there were positive correlations had signiﬁcantly greater variation in EI for dinner only

between the food range score and EEDLW (rS ¼ 0.63, P ¼ 0.01), (P ¼ 0.01) and over the whole day (P ¼ 0.009). There was no degree of misreporting on the FR (rS ¼ 0.63, P ¼ 0.01), and relationship between and BMI, EEDLW or degree of misreport- degree of misreporting on the DH (rS ¼ 0.71, P ¼ 0.003). ing on the FR or DH, and the number of activities (leisure,

Variation in EI at dinner correlated positively with EEDLW occupational and total) was not different between the sexes

Figure 1 Relationship between the degree of misreporting (absolute value of EI 7 EEDLW) on the diet history (A) and 7 day food record (B) with energy expenditure measured by doubly labelled water.

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 363

Figure 2 Real difference between EI(FR) plotted against EEDLW exposing distinct subgroups of reporters above and below 15 MJ. or between accurate and inaccurate reporters. There tended which is why we did not select one method over the other to to be a lower number of leisure-time activities with increas- categorise accurate reporters. As with the FR, the period of ing age (r ¼ 7 0.58, P ¼ 0.07) and BF (r ¼ 7 0.58, P ¼ 0.08). measurement of EE with DLW is assumed to be representa- Younger subjects had signiﬁcantly more leisure-time tive of usual behaviour, which may serve as a limitation since (P ¼ 0.01) and total (P ¼ 0.03) activities than older subjects in reality it measures actual behaviour. For our purposes, when grouped by age, but there was no difference in number however, the ﬁnding linking higher EE with inaccurate of occupational activities. reporting was found using both dietary assessment methods

In our multiple regression analysis, higher EEDLW, female (Figures 1 and 2). sex and higher %CV in number of dinner foods accounted Our ﬁnding that high BMI does not deﬁne misreporting for more than 96% of the variation in misreporting on the subjects is supported by work in Swedish cohorts. In these

FR. Higher EEDLW, younger age, higher %CV in number of investigations, misreporting in overweight and obese sub- dinner foods, higher food range score, and female sex jects with a wide range of BMI (25.5 – 49.5) was not found to accounted for almost 96% of the variation in misreporting increase with increasing BMI (Lindroos et al, 1999), nor was on the DH. there evidence of obesity-speciﬁc misreporting when non-

There was no correlation between EEMAQ and EEDLW, obese subjects categorised according to BMI were compared however correlations with EI from the FR (r ¼ 0.75, with obese subjects (Lindroos et al, 1997). In addition, P ¼ 0.01) and DH (r ¼ 0.58, P ¼ 0.08) were both good. analysis of data from Livingstone et al (1990) with normal weight, overweight and obese subjects showed no difference in BMI of accurate and inaccurate reporters when classiﬁed Discussion as in the present study. Classiﬁcation of individuals accord- There is a common perception in the literature that over- ing to BMI is known to be an imprecise tool, often categoris- weight subjects tend to under-report energy intakes, yet our ing those with a greater muscle mass into overweight or study found that accuracy was more associated with EE, and obese categories. Highly active subjects with greater EE can was not related to degree of overweight or obesity. Under fall into this category, and have previously been shown to be energy balance conditions, subjects with higher EE would inaccurate in reporting dietary intake (Beidleman et al, 1995; have to report a lot more. Some subjects performed better in Edwards et al, 1993; Trappe et al, 1997). one method of dietary assessment than with the other, but The lack of contribution of BMI or BF together with the then the DH assesses usual intake and relies on memory, strong contribution of EEDLW to misreporting in both modes whereas the FR assesses actual intake and is subject to an of dietary assessment indicates the involvement of a factor or intervention effect (Tapsell et al, 1993). The different mea- factors which are quite separate to weight issues but still surement contexts and skills required in recording and related to higher EE. The current study showed that dietary reporting may account for differences in performance, variation together with higher EE predicted misreporting.

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 364

Figure 3 Relationship between the degree of misreporting (absolute value of EI 7 EEDLW) on the diet history (A) and 7 day food record (B) with body mass index.

Higher energy expenders who are weight stable may achieve foods eaten at dinner over 7 days is detrimental to accuracy a matching high EI by eating greater quantities of foods in recall on the diet history. and=or eating many different foods. Difﬁculties in estimat- Interestingly, less variation at dinner was associated with ing larger portion sizes have previously been reported increasing age, as was better performance on the DH. Other (Vuckovic et al, 2000). It seems intuitive that a greater investigators have also found reduced variation in foods number of foods would also be difﬁcult to record or recall consumed by older people (Neuhaus et al, 1991; Fanelli & accurately. Our results indicate that stability in the number Stevenhagen 1985), but this was not linked to more accurate of different foods eaten at the dinner meal is important for reporting. Our older subjects had a lower number of leisure- both recording and recall accuracy and that a larger range of time and total activities than the younger ones, and older

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 365

Figure 4 Relationship between the degree of misreporting (absolute value of EI 7 EEDLW) on the diet history (A) and 7 day food record (B) with percentage body fat.

subjects would also be expected to have a lower EE. Thus it factors are involved which are inclusive but not limited to appears that people with highly active and varied lifestyles the reporting ability of the subject. may find it more difficult to report that variation accurately. The DLW method has been reported to be very accurate It is also important to point out that larger portion sizes compared to indirect calorimetry (Schoeller et al, 1986), and and a greater number of foods at any given meal may be has been described as a gold standard for validating EI difficult for an investigating dietitian to capture and so errors (Beaton et al, 1997). However some of our subjects had can also be made in the recording process for dietary assess- higher EE that was not explained by self-reported physical ments which involve an interview-based process such as the activity. This might have reflected the difference between the DH. In fact error in dietary assessment can occur anywhere assessment of habitual EE and actual EE and highlights the along the path from recording to analysis (Bingham, 1991). need to be cautious in our comparison of habitual EI and The concept of ‘misreporting’ may therefore be better char- actual EE. While neither the DH nor the FR performed well acterised by the term ‘mismeasurement’ since a range of against DLW, more subjects were able to report accurately on

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 366 Table 4 List of dinner foods obtained from the 7 day food records calculation of EE while being time-effective in small-scale clinical trials. While the there was no correlation between Food EEMAQ and EEDLW, we achieved good correlations between Beef=lamb=veal=pork=chicken — portion the two measurements of past behaviour (EEMAQ and DH) Beef=lamb=veal=pork=chicken — mixed dish which is impressive since the processes used to obtain each Seafood — portion Seafood — mixed dish set of data are quite different with a different set of associated Vegetables errors. While acknowledging the limitation of the small Potato sample size and the distinct possibility that both measures Pasta are inaccurate, further exploration of this cheap and readily Rice Sauces=gravies=pastes available method is warranted in both heterogeneous and Toppings=cheese=cream=dressings homogeneous populations. Quiche=pie=pastie Nachos=corn chips=tortilla Toast=bread=muffin=crumpet Egg Conclusions Soup Studies such as clinical trials that require volunteers to report Hot chips dietary intakes accurately under ‘free-living’ conditions may Pizza need to consider characteristics of volunteers in addition to Salad=vegetables — mixed dish medical screening. The results of our study indicate that it may be difficult to capture both usual consumption and=or the foods consumed during the study period for people with the FR. It is possible that this was a carry-over effect from the a higher EE. This may be true for subjects with either a practice provided by the preceding DH. However, again we consistently or sporadically high activity level as well as for would expect closer correspondence between FR and DLW some overweight subjects. Stability of eating behaviour as since they both attempt to measure behaviour during the well as exercise may also be important. For these reasons it study period while DH measures past behaviour. may be necessary to screen for highly active subjects, pat- Acknowledging this difference is critical; although it is terns of higher activity levels, and stability of eating and habitual diet that is important in the development of diet – exercise behaviours in studies where accuracy of measure- disease relationship and to develop treatment strategies in ment of dietary intake at an individual level is critical. clinical practice, we still have no definitive way to assess it Applying our knowledge of differences in performance accurately. In this study we trialled an alternative method of between dietary assessment methods for specific study popu- assessing EI via self-reported physical activity in a subset of lations should also be considered. Subjects with less stable subjects. Our method is detailed enough to enable factorial lifestyles, for example, younger and more active subjects may

Table 5 Lifestyle characteristics

%Variation in EIFR Number of activities %Variation in number of Subject Whole day Dinner dinner foods (FR) Food range score Leisure Occupational

Women 1 22.14 35.54 48.57 8 6 4 2 14.79 27.17 36.62 6 2 4 3 15.64 19.74 54.74 6 2 1 4 44.94 75.80 83.28 16 8 2 5 24.01 37.40 22.97 4 5 2 6 17.07 56.76 67.87 10 NA NA 7 14.88 44.04 33.75 3 NA NA Mean (s.d.) 21.9 (10.8) 42.9 (18.7) 49.7 (6.4) 7.6 (1.6) 5.0 (1.2) 2.6 (0.6)

Men 8 15.14 28.85 43.20 6 5 2 9 7.77 29.33 46.11 5 4 1 10 23.02 69.89 77.53 17 6 3 11 22.01 44.18 64.91 6 10 3 12 25.09 57.63 39.59 6 7 5 13 21.43 29.59 48.45 8 NA NA 14 23.01 56.71 45.53 4 NA NA 15 25.37 92.16 50.90 8 NA NA Mean (s.d.) 20.4 (6.0) 51.0 (22.7) 52.0 (6.0) 7.5 (1.5) 6.8 (1.2) 2.8 (1.2) Mean (s.d.) 21.1 (8.3) 47.3 (20.6) 50.9 (16.4) 7.5 (4.1) 5.9 (2.6) 2.7 (1.3)

Note: results not signiﬁcantly different between males and females.

European Journal of Clinical Nutrition Reporting accuracy on 7 day food records JA Barnard et al 367 be better suited to the food record method as opposed to the FAO=WHO=UNU Expert Consultation (1985): Energy and Protein Requirements. WHO Technical Report Series 724. Geneva: World diet history method for assessing dietary intake. In addition Health Organization. the use of cheap and relevant alternatives to assessing EE Heitmann BL & Lissner L (1995): Dietary underreporting by obese may be an option for use in clinical trials. Whilst our small individuals — is it specific or non-specific? Br. Med. J. 311, 986 – sample size limits the generalisability of our results, these 989. Herbert JR, Clemow L, Pbert L, Ockene IS & Ockene JK (1995): Social findings are worthy of consideration in clinical trials invol- desirability bias in dietary self-report may compromise the validity ving dietary assessment and warrant further investigation. of dietary intake measures. Int. J. Epidemiol. 24, 389 – 398. Kriska AM (1997): Modifiable activity questionnaire. In A Collection of Physical Activity Questionnaires for Health-Related Research,eds.MA Pereira, SJ FitzGerald, EW Gregg, ML Joswiak, WJ Ryan, RR Acknowledgements Suminski, AC Utter & JM Zmuda. Med. Sci. Sports Exerc. 29(Suppl The authors wish to express their gratitude to Ken Russell, 6), S73 – S78. Janine Higgins, Sheena McGhee, Herb Groeller and Michelle Lafay L, Basdevant A, Charles M-A, Vray M, Balkau B, Borys J-M, Gordon for their technical and organisational assistance. Eschwe`ge E & Romon M (1997): Determinants and nature of dietary underreporting in a free-living population: the Flerbaix Special thanks must go to Mia Raaschou who provided a Laventie Ville Sante´ (FLVS) study. Int. J. Obes. Relat. Metab. Disord. portion of the data for subsequent analysis for this report. 21, 567 – 573. Lewis J, Hunt A & Milligan G (1995): NUTTAB95. Canberra, Australia: Government Publishing Service. References Lindroos A-K, Lissner L, Mathiassen ME, Karlsson J, Sullivan M, Beaton GH, Burema J & Ritenbaugh C (1997): Errors in the inter- Bengtsson C & Sjo¨stro¨m L (1997): Dietary intake in relation to pretation of dietary assessments. Am. J. Clin. Nutr. 65(4 Suppl), restrained eating, disinhibition, and hunger in obese and non- 1100S – 1107S. obese Swedish women. Obes. Res. 5, 175 – 182. Beidleman B, Puhl J & De Souza M (1995): Energy balance in female Lindroos A-K, Lissner L & Sjo¨stro¨m L (1999): Does degree of obesity distance runners. Am. J. Clin. Nutr. 61, 303 – 311. influence the validity of reported energy and protein intake? Bingham SA (1991): Limitations of the various methods for collect- Results from the SOS Dietary Questionnaire. Eur. J. Clin. Nutr. 5, ing dietary intake data. Ann. Nutr. Metab. 35, 117 – 127. 375 – 378. Bingham SA, Cassidy A, Cole TJ, Welch A, Runswick SA, Black AE, Livingstone MBE, Prentice AM, Strain JJ, Coward WA, Black AE, Thurnham D, Bates C, Khaw KT, Key TJA & Day NE (1995): Barker ME, McKenna PG & Whitehead RG (1990): Accuracy of Validation of weighed records and other methods of dietary weighed dietary records in studies of diet and health. Br. Med. J. assessment using the 24 h urine nitrogen technique and other 300, 708 – 712. biological markers. Br. J. Nutr. 73, 531 – 550. Mela DJ & Aaron JI (1997): Honest but invalid: what subjects say Black AE (1997): Under-reporting of energy intake at all levels of about recording their food intake. J. Am. Diet. Assoc. 97,791– 793. energy expenditure: evidence from doubly labelled water studies. Neuhaus JM, Murphy SP & Davis MA (1991): Age and sex differences Proc. Nutr. Soc. 56, 121A. in variation of nutrient intakes among U.S. adults. Epidemiology 2, Black AE, Goldberg GR, Jebb SA, Livingstone MBE, Cole TJ & Prentice 447 – 450. AM (1991): Critical evaluation of energy intake data using funda- Schoeller DA, Ravussin E, Schutz Y, Acheson KJ, Baertschi P & Jequier mental principles of energy physiology: 2. Evaluating the results of E (1986): Energy expenditure by doubly labeled water: validation published surveys. Eur. J. Clin. Nutr. 45, 583 – 599. in humans and proposed calculation. Am. J. Physiol. 250, R823 – Black AE, Prentice AM, Goldberg GR, Jebb SA, Bingham SA, Living- R830. stone MBE & Coward WA (1993): Measurements of total energy Schofield WN, Schofield C & James WPT (1985): Basal metabolic expenditure provide insights into the validity of dietary measure- rate — review and prediction, together with an annotated biblio- ments of energy intake. J. Am. Diet. Assoc. 93, 572 – 579. graphy of source material. Hum. Nutr. Clin. Nutr. 39C(Suppl 1), 1 – Black AE, Bingham SA, Johansson G & Coward WA (1997): Validation 96. of dietary intakes of protein and energy against 24 h urinary N and Tapsell L, Bravo A & Tranter D (1993): Evaluation of a community DLW energy expenditure in middle-aged women, retired men and nutrition education program: an experience with weighed food post-obese subjects: comparisons with validation against pre- records. Aust. J. Nutr. Diet. 50,15– 18. sumed energy requirements. Eur. J. Clin. Nutr. 51, 405 – 413. Tapsell LC, Brenninger VL & Barnard JA (2000): Applying conversa- Braam LAJLM, Ocke´ MC, Bueno-de-Mesquita HB & Seidell JC (1998): tion analysis to support accurate reporting in the diet history Determinants of obesity-related underreporting of energy intake. interview. J. Am. Diet. Assoc. 100,818– 824. Am. J. Epidemiol. 147, 1081 – 1086. Trappe T, Gastaldelli A, Jozsi A, Troup J & Wolfe R (1997): Energy Edwards JE, Lindeman AK, Mikesky AE & Stager JM (1993): Energy expenditure of swimmers during high volume training. Med. Sci. balance in highly trained female endurance runners. Med. Sci. Sports Exerc. 29, 950 – 954. Sports Exerc. 25, 1398 – 1404. Vuckovic N, Ritenbaugh C, Taren DL & Tobar M (2000): A qualitative Fanelli MT & Stevenhagen KJ (1985): Characterizing consumption study of participants’ experiences with dietary assessment. J. Am. patterns by food frequency methods: core foods and variety of Diet. Assoc. 100, 1023 – 1028. foods in diets of older Americans. J. Am. Diet. Assoc. 85, 1570 – 1576.

European Journal of Clinical Nutrition