Detection of Underestimated Energy Intake in Young Adults

Transcription

1 International Journal of Epidemiology O International Epldemiologlcal Association 1994 Vol. 23, No. 3 Printed In Great Britain Detection of Underestimated Energy Intake in Young Adults J6HANNA HARALDSD6TTIR AND BRITTMARIE SANDSTRdM Haraldsddttir J (Research Department of Human Nutrition, Royal Veterinary and Agricultural University, Rolighedsvej 25, DK-1958 Frederiksberg C, Copenhagen, Denmark) and Sandstrom B. Detection of underestimated energy intake in young adults. International Journal of Epktemiotogy1994; 23: Background Critical evaluation of energy intake data from dietary studies is difficult but important A simple approach has been suggested using minimum cutoff limits for the ratio of energy intake to basal metabolic rate as predicted from body weight (EI/BMR«t)- This study investigates the potential of this particular approach in young subjects. Methods. Energy intake during strictly controlled dietary intervention studies (2-6 weeks or 8 months) was compared with reported intake before the study started (7 days' weighed food records). Participants were 48 men and 22 women, years. Results. Mean EI/BMRest values during the intervention study were 1.92 for men and 1.77 for women. Reported intakes before the study were only slightly lower (women 8%, P< 0.05, men 5%, P> 0.05). Two EI/BMR»»t. limits were tested. Smaller underestimations could be detected by applying calculated cutoff limits based on average EI/BMFUSL values from the intervention studies instead of published limits assuming a low physical activity level (6% versus 22% underestimation of group intake). Conclusions. These results suggest that for the evaluation of energy intake of young adults it is more meaningful to use EI/BMRi»t cutoff limits which assume a moderate, instead of a low, average physical activity level. The need for critical evaluation of energy intake data obtained in dietary studies has been clearly demonstrated by recent research. Particular concern has been raised by reports of widespread, systematic and differential underestimation in studies using the weighed food record method; the method that has been regarded as the gold standard. Jt Direct validation of reported energy intake from freeliving individuals can be made by the doubly-labelled water technique which determines total energy expenditure. 5 However, that method is too costly to be routinely included in epidemiological studies and alternative ways of critically evaluating reported intakes are necessary. One alternative, suggested recently by Goldberg el al. 6 is to express energy intake as a ratio of the predicted basal metabolic rate (EI/BMRe«.) and compare it with estimates of physiologically plausible energy expenditure/basal metabolic rate ratios (EE/BMR). For this purpose Goldberg et al. constructed tables with lower limits for EI/BMRen. values based on recent expenditure studies and FAO/WHO's energy requirement estimates. 7 These tables only aim at identifying underestimations which result in implausibly low energy intakes, incompatible with weight maintenance and minimal physical activity. However, the full potential of this approach may be broader. The main objective of our study was to investigate the potential of EI/BMRa,. limits for detecting underestimation of energy intake in young adults. A second objective was to measure the validity of 7-day weighed food records in a group of highly motivated young volunteers. SUBJECTS AND METHODS Subjects This paper is based on data from 70 weight-stable volunteers (48 men and 22 women) from a series of controlled dietary intervention studies at the Department. 8 " 10 Inclusion criteria in the intervention studies were: age years, body mass index <25 kg/m 2, light or moderate level of physical activity (maximum 2 hours per week of intensive physical training), tobacco consumption <10 cigarettes per day. All subjects had an initial medical examination. No subject participated in more than one feeding study. The intervention studies were approved by the regional ethical committee of Frederiksberg. The 70 subjects came from a larger group of 87 (61 men, 26 women) of which 17 did not fulfill the criterion of weight stability (see Body-weight). Research Department of Human Nutrition, RoyaJ Veterinary and Agricultural University, Rolighedsvej 25, DK-1958 Frederiksberg C, Copenhjgen, Denmark. Study Design and Diet Dietary intakes were strictly controlled during the intervention studies. All food items were weighed on precision 577

2 578 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY scales and had been prepared by kitchen staff in the Department. Subjects had their two main meals at the Department every day during the week, and were supplied with weighed, prepacked meals for the next breakfast, evening snacks and for the weekend. The duration of the feeding studies was 2, 3 or 6 weeks, or 8 months. For subjects from the 8-month study (n = 22) only intake data from 8 weeks (2x4 weeks during February-April) were used in this paper. These 8 weeks were selected as they did not include major holidays where a certain degree of freedom was allowed. 8 The intervention diets were all low-fat (30-3 5E% fat) but varied in respect to the content of dietary fibre and the fatty acid composition. In the long-term study the amount of food taken was decided by the subjects but weighed on a balance connected to a computer for immediate identification and entry. In the short-term studies the majority of intake was preweighed in individual portions. Left-overs were weighed and subtracted, and individual wishes for additional intake were met by snacks and sandwiches with the same composition as the average diet. For a more detailed description of the studies see previous publications. 8 " 1 Energy intakes were calculated using DANKOST, a program based on the official Danish food table.'' Calculated values compared to analysed values for fat, protein and dietary fibre have been published from one of the intervention studies. Body-weight Body-weight (fasting, wearing only underwear, no shoes) was measured on an electronic scale once a week (once a month in the 8-month study). Subjects from the shortterm studies were defined as weight-stable if the difference between their weight at the beginning and the end of the intervention period did not exceed 1.0 kg. Subjects from the long-term study (8 months) were defined as weight-stable if their monthly weights during February-May (the period of the dietary data included in this paper) did not deviate more than 1.0 kg from their mean weight for this period. Reported Habitual Intake Subjects reported their dietary intake before entering the studies by 7-day weighed food records using an electronic scale (Soehnle, 1 g intervals). Energy intakes were calculated in the same way as for the intervention diets. Prediction of Basal Metabolic Rate Basal metabolic rate (BMRa.) was predicted from bodyweight according to the FAO/WHO equations for the age group years: 7 Men: BMIW (MJ/d) = ( x body-weight) Women: BMR«. (MJ/d) = ( x body-weight) Derivation of Cutoff Limits for EIIBMR Cutoff limits for physiologically plausible energy intakes at weight-maintenance, expressed as EI/BMR, were calculated from the equations developed by Goldberg et al. 6 These equations are based on a 'true' mean total energy expenditure value (TEE/BMR) for the population group in question (e.g. the WHO reference value for that sex and age group). At energy balance intake (EI/BMR) should equal expenditure. The lower cutoff limit is then calculated as two (or three) standard deviations below the mean TEE/BMR, thus taking into account the inter-individual variation in TEE/BMR values. The calculations also include allowances for the day-to-day variation in individual energy intakes, the error in predicting BMR instead of measuring it (when that is so), the number of days on which the intake is based, and the number of individuals in the group. Based on these assumptions Goldberg et al. arrived at the following equation for a physiologically plausible EI/BMR value: EIIBMR > TEEIBMR x expf where the term S includes the variations, the prediction error, and the number of days mentioned above, n is the number of individuals in the group, and SDmm is the number of standard deviations corresponding to the requested confidence interval (= -2 for a 95% confidence interval). Assuming that the coefficients of variation are 12.5% for inter-individual TEE/BMR, 23% for intra-individual energy intake, and 8% for the prediction error of BMR, the equation for S becomes: For a detailed explanation of the derivation of these equations and the assumptions behind them, the reader is referred to Goldberg et al's paper. 6 That paper also includes extensive tables of cutoff values for various number of days and for various group sizes. These tabulated values were based on an average TEE/BMR value of 1.55, corresponding to FAO/WHO's estimated average energy requirement for low activity. 7 In this study new cutoff limits were calculated, for men and women separately, using the mean weightmaintenance EI/BMRjn. results from the feeding studies as TEE/BMR values, instead of 1.55, in the above equation.

3 DETECTION OF UNDERESTIMATED ENERGY INTAKE 579 TABLE 1 Energy intake during controlled intervention studies expressed as MJId and as a multiple of the estimated basal metabolic rate Men Group A b (n = 13) Group ff(n = 29) All men (n = 42). Women Group A b (n = 9) Group B c (n= 19) All women (n = 28) MJ/d x±sd ± ± ± ± ±0.70 EI/BMR«.' x ± SD min.-max ± ± ± ± ± ± Z Z 'Energy intake divided by the predicted basal metabolic rate, O, calculated from body-weight according to the FAO/WHO equations. 7 b Group A: From 2x4 weeks of an 8 monthj' intervention study. c Group B: From 2 weeks', 3 weeks' or 6 weeks' intervention studies. Statistics Data were analysed using the SPSS statistical program. Group means for reported versus determined energy intakes were compared using a paired t test. Pearson's correlation coefficient was used for correlating reported and determined intakes. RESULTS Actual EI/BMRe«. ratios at weight-maintenance, as reflected in intakes during the intervention studies, are presented in Table 1. Similar values were obtained from the long-term study (group A) and the shorter studies (group B). When the two groups were combined the average values were 1.92 for the men and 1.77 for the women. Minimum values were 1.50 (women) and 1.65(men) and maximum values were 2.07 (women) and 2.28 (men). In Table 2 the reported energy intakes before the intervention studies started are presented and compared with the intakes during the studies. The mean reported values were slightly lower (5% and 8% lower for men and women respectively). This difference was statistically significant for the total group and for the women, but not for the men. The range of the reported intakes was larger than during the intervention with minimum EI/BMRot. values as low as 0.90 (women) and 1.15 (men). The correlation between the reported intakes and intakes during the intervention was moderately strong (Pearson's r = 0.46) and significant at the 0.01% level. Investigation of the individual differences showed that underestimation was more common than o verestimation (67% versus 31% of the subjects). Energy intakes were compared with two different EI/BMIW limits (Table 3): tabulated cutoff values from the literature 6 and calculated cutoff values derived from the same equations as the tabulated values but based on the average EI/BMRe«. from the feeding studies. The calculated cutoffs were higher than the tabulated values (23% higher for men, 14% higher for women). These two sets of limits were used to evaluate the reported energy intake before the studies started. Using the tabulated limits the average group intakes were classified as acceptable and only one individual intake as underestimated. If instead, the calculated limits were used the average intake of the women was classified as slightly underestimated, and four individual intakes as underestimations. The theoretical ability of the two sets of cutoff limits to detect underestimation was tested on the actual intake from the feeding studies (Table 3). The actual group averages (men 1.92, women 1.77) would have to be underestimated by approximately 20% (men 24%, women 18%) before they were detected by the tabulated cutoff values presented by Goldberg et al., 6 but only by 6% to be detected by the cutoff values calculated from the intervention studies (men 1.82, women 1.67). The individual participants would have to underestimate TABLE 2 Comparison of reported energy intake before start of controlled intervention studies with actual intake during the studies (mean ± SD) Reported* intake before study MJ/d Actual intake b during study MJ/d Difference MJ Reported 1 intake before study El/BMR^ Actual intake' during study El/BMRot Men(n = 42) 13.49± ± Women(n = 28) 9.66± O± ^ -8.0 All(n = 70) 11.96± ± ^ -f>.\ 1.82 ± ± ± ± ± ±0.14 "From 7 days' weighed food records before start of controlled intervention studies. b From controlled intervention studies lasting 2-6 weeks or 8 months (see note to Table 1) 'Energy intake expressed as a multiple of the predicted basal metabolic rate, BMRtn., calculated from body weight according to the FAO/WHO equations. 7 d Difference statistically significant (paired t test), P < 0.05.

4 580 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 3 Energy intakes compared with two cutoff limits for EJ/BMRat* Reported intake' before intervention EI/BMRc*. Mean (range) Actual intake 0 during intervention nudes EI/BMRo,. Mean (range) Tabulated* 1 cutoff EI/BMR«. Group f mean Individual* values Group r mean Calculated 0 cutoff EI/BMR«. Individual* values Men(n = 42) Women (n «28) 1.82( ) 1.62( ) 1.92( ) 1.77( ) J "Energy intake expressed as a multiple of the predicted basal metabolic rate, BMR«u., calculated from body-weight according to the FAOAVHO equation!. 1 T^rom 7 days' weighed food records before beginning of intervention studies. 'From controlled intervention studies lasting 2-6 weeks or 8 months (see note to Table 1). d Cutoff values from Goldberg el al's table 6 based on an average EI/BMR of 'Calculated according to Goldberg el al's equations' but based on average E1/BMR«values from the present intervention studies (men 1.92, women 1.77). r Cutoff for group mean: That value below which the average EI/BMRot value of the group may be considered implausibly low (P < 0 05), based on 7 days intake and the actual group size. Cutoff for individual values: That value below which an individual's EI/BMRt*. may be considered implausibly low (/»< 0.05), based on 7 days intake. their actual intake (EI/BMR«range ) by 28-53% before they were detected by the tabulated individual cutoff, as compared to 18-41% before being detected by the calculated cutoffs derived from the intervention studies (1.36 men, 1.25 women). Although all subjects included in this study fulfilled the criterion of weight stability a slight negative trend in body-weight was observed during the intervention period, mean weight change (±SD) = (±0.45) kg. A similar negative trend was observed for the 17 subjects who were not included in this study, due to lack of weight stability; mean weight change (±SD) = (±1.28) kg. DISCUSSION Validation of Reported Intakes Before the Studies Against Intake During the Intervention Studies This study showed that the energy intake before the feeding studies started was slightly underestimated by the 7-day weighed food records. However, in this group of carefully selected, normal-weight, young volunteers who had been thoroughly instructed, this underestimation only amounted to an average of 6% (0.8 MJ/d) for the total group. The present results thus support validation studies using the doubly-labelled water method, which indicate that underestimation is a general problem with weighed food records. 1 ' 2 ' 4 ' 12 '' 3 The largest errors were found in groups of obese or weight-conscious individuals, where average underestimations up to 33% have been reported, 2 ' 4 whereas lean volunteers seem to perform better. 212 The underestimation observed in this study is smaller than found in some comparable studies reported in the literature. Based on a similar comparison between reported energy intake before the study (food records) and weight-maintenance energy intakes in feeding trials Mertz et al. M observed an average underestimation of 18% for 266 males and females (21-64 years), Lissner et al. 15 observed 23% underestimation in a group of 63 females (19-67 years) and Hallfrisch et al} 6 observed 37% underestimation for 12 females and 19% for 12 males (21-51 years). Whereas Lissner et al.'s study included both lean and obese subjects neither Mertz et al. or Hallfrisch et al. specified this. In contrast to the present study the food records used for reporting habitual intake in two of these studies' 5 ' 16 were not weighed records but estimated records using household measures instead of weighing, and the third study 14 used a combination of weighing, measuring cups and measuring spoons. The better agreement found in the present study might be explained by differences in subject characteristics (adiposity, weight-consciousness, motivation, age) and in the precision of the food recording method. The observed underestimation of energy intake in the present study could be caused by underreporting, with omission of foods, but is more likely a reflection of reduced consumption during the recording period due to an increased awareness of eating habits. It is also possible that the controlled intervention study with well balanced regular meals or the experimental diet in itself imposed metabolic changes of energy utilization. Actual EI/BMResi. Ratios During the Intervention Studies The average EI/BMR«t ratios during the controlled intervention studies were 8% higher than FAOAVHO's estimated requirements for moderately active subjects

5 DETECTION OF UNDERESTIMATED ENERGY INTAKE 581 (1.78 for men and 1.64 for women, years.) 7 This difference was not expected as the physical activity level of the subjects had been classified as light to moderate, based on their information on sport activities and workload. The FAO/WHO requirement estimates are not based on measurements of total 24-hour energy expenditure in free-living subjects however, but on factorial calculations summing the energy costs per minute of different types of activities, allocating certain amounts of time for each type of activity during a 24-hour period. 7 Compared with measurements of total energy expenditure in free-living subjects, using the doubly-labelled water method, the present EI/BMRen results were 8% higher than average expenditure values published for a broader age group (1.78 for men and 1.62 for women, n = 105). 6 However, the results were close to measured values for normal-weight subgroups of the same age (20-30 years) in the study of Livingstone et al. (1.94 for seven men and 1.78 for five women), 1 and slightly lower than results from Schulz et al.'s study on six subjects of the same age group. 12 In contrast, lower energy expenditures (1.41) have been reported for seven normal-weight women of similar age. 2 Comparisons between energy intake and energy expenditure values are based on two assumptions: that the energy balance is zero and that the calculated intake reflects the 'available' energy content of that diet. In this study only subjects who met the criteria for weight stability were included, and although the intervention period was relatively short in some studies (2-3 weeks) the similar average EI/BMRot. values obtained from the long-term and the short-term studies support the validity of the short-term results. Corrections for a reduced energy availabilityresultingfrom the relatively highfibrecontent in the intervention diets (s*3 g/mj), by 2.5%, as suggested by FAO/WHO, 7 reduced the average EI/BMIW ratio to 1.87 for men and 1.73 for women. However, the values were still higher, approximately 6%, than the FAO/ WHO requirement estimates for moderate activity. In theory the relatively high EI/BMRd. ratios observed in this study could be explained by a systematic underestimation of the BMR, even if this is not a likely explanation. The basis for the algorithms used for predicting the BMR values 7, measured BMR values and body-weights from 4726 men and women (18-60 years), 17 and a recent comparison between measured BMR values (whole body calorimetry) and values predicted from the age-specific equations, for 233 subjects, did not indicate any systematic underestimation. 6 The present results and the published data on total energy expenditure discussed above suggest that the average activity level of this age group, years, may typically be slightly above moderate, compared to FAO/WHO estimates, even in the absence of heavy work or vigorous sport activities. This could be due to the length of time spent on daily activities such as walking, cycling and moving around, as well as occasional disco evenings and sports activities. Thus the observed energy intakes might not be valid for young adults in other countries with a more sedentary lifestyle. Differences in body mass index and smoking habits are other factors that have to be accounted for. The results therefore need to be confirmed by additional data on total energy expenditure in free-living subjects of this age. Use ofeiibmrcn. Limits to Evaluate Energy Intakes A general advantage of expressing energy intake in relation to BMR is that it gives an immediate impression of the physiological plausibility of the intake. By defining cutoff limits for such ratios this concept can be used as an evaluation instrument. The present results demonstrate that the sensitivity of this approach depends upon the reference point used for calculating the cutoff limits, and that for this group of year old men and women the basis of the published cutoff limits, an average EI/BMR of 1.55, 6 seems to be too low. The decision to use FAO/WHO's estimated requirement for light activity (1.55) as a reference point instead of the average energy expenditure values from the doubly-labelled water studies on free-living subjects (1.78 for men, 1.62 for women), 6 was in accordance with the main aim of Goldberg et al.; to identify only the obvious underestimations resulting in implausibly low energy intakes. This resulted in an instrument with maximal specificity, but was at the expense of its sensitivity. The present results suggest that for young adults it would be more appropriate to base the cutoff limits on the estimated energy requirements for moderate, instead of light, physical activity (1.78 for men and 1.64 for women, years). 7 Cutoff limits for EI/BMRtn. ratios are particularly useful for evaluating mean intakes of groups and subgroups. The sensitivity of the method is higher on the group level, and the potential error in predicting BMR instead of measuring it has a larger impact on the individual level. The prediction error has been estimated to a coefficient of variation of 8% when FAO/WHO's equations are applied. 6 Inclusion of height in the prediction equations does not markedly improve the prediction. 7 Future studies including reliable measurements of lean body mass, instead of only weight and height, may be able to predict BMR more accurately than today. 18 The general utility of EI/BMRt*, limits for identifying underestimation does not manifest itself clearly in the present study on highly motivated, normal-weight subjects where underestimation of intake was much less of a problem than reported for other groups. A better

6 582 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY demonstration of the value of this approach was provided by Black et al. 19 who found that mean energy intake was obviously underestimated in 46 out of 68 subgroups from 37 published studies. The cutoff limits discussed in this paper aim at evaluating the actual energy intake rather than the habitual intake, and therefore take into account the day-to-day variation in individual intake and the number of days on which the intake is based (see the Methods section). Consequently the limit based on 1 day's intake is lower than if based on 4 or 14 days, in particular for a small group or an individual, as illustrated by tabulated values. 6 The sensitivity of the method is therefore considerably improved if energy intake is based on a larger number of days, as it better reflects habitual intake. Goldberg et al. expressly refrained from defining upper limits for plausible EI/BMRot, ratios. 6 One way of evaluating high energy intakes could be by comparison with FAO/WHO's estimated average requirements for heavy activity. Thus, group means exceeding these values (men 2.10, women 1.82) 7 would be improbable for most groups and indicate overestimation. Further, for individual intakes exceeding 2.63 (men) and 2.28 (women) (i.e. two SD above the average requirements for heavy activity, assuming a coefficient of variation of 12.5%) the probability of overestimation would be high. When EI/BMRot. values of these magnitudes are encountered there is reason to question their validity. Valuable information about the sensitivity and the specificity of different EI/BMRm cutoff limits may be obtained through doubly-labelled water studies and studies including biomarkers of intake, and such studies should therefore be encouraged. In conclusion, EI/BMRot limits are useful for detecting obvious underestimations of energy intake, and the present results indicate that there is scope for extending and improving their use. For young adults the sensitivity of the method may be considerably improved by using calculated cutoff limits allowing for the higher activity level in this age group. The results suggest that for this age group it would be more appropriate to base the cutoff limits on estimated energy requirements for moderate, instead of light, average physical activity. ACKNOWLEDGEM ENTS We are indebted to Hanne Jensen for dietetic and computing assistance. This study was partly supported by grants from the Danish Agricultural and Veterinary Research Council (grant no ) and from the E)anish Ministry of Agriculture (grant no. LME-KVL-8). REFERENCES 'Livingstone MBE, Prentice A M, Strain i i el al Accuracy of weighed dietary records in studies of diet and health. Br Med ; 300: Prentice A M, Black A E, Coward W A et al High levels of energy expenditure in obese women. BrMcdJ 1986, 291: Bingham S A, Welch A, Cassidy A, Runswkk S. The use of 24h urine nitrogen to detect bias in the reported habitual food intake of individuals assessed from weighed dietary records. Proc Nutr Soc 1991; 50: 32A(abs.). 4 Black A E, Jebb S A, Bingham S A. Validation of energy and protein intakes assessed by diet history and weighed records against energy expenditure and 24h urinary nitrogen excretion Proc Nutr Soc 1991; 50: 108A(abs.). 'Coward W A. The doubly-labelled water ( 2 H 2 "O) method: principles and practice. Proc Nutr Soc \9iS; 41: 'Goldberg G R, Black A E, Jebb SAetaL Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. Derivation of cut-off limits to identify under-recording. Eur J Oin Nuir 1991; 45: FAO/WHO/UNU. Energy and Protein Requirements. Report of a joint expert consultation. WHO Tech Rep Series No. 724 Geneva: WHO, 'SandstrSm B, Marckmann P, Bindslev N. An eight-month controlled study of a low-fat high-fibre diet: effects on blood lipids and blood pressure m healthy young subjects. Eur J Clin Nutr 1992; 46: 'Marckmann P, Sandstrom B, Jespersen J. Effects of total fat content and fatty acid composition in diet on factor VII coagulant activity and blood lipids. Atherosclerosis 1990; 80: l0 Marclunann P, Jespenen J, SandstrSm B. Dietary n-3 and n-6 polyunsaturated fatty acids affect the fibrinolytk system differently. In. Sinclair A, Gibson R (eds). Essential Fatty Acids and Eicosanoids: Invited Papers from the Third Internationa} Congress. Champaign, IL American Oil Chemists'Society, 1993, pp '' Mailer A. Lemedsmtddeltabeller(Food composition tables) Copenhagen: Danish National Food Agency, l2 Schulz S, Westerterp K R, Brack K. Comparison of energy expenditure by the doubly labeled water technique with energy intake, heart rate, and activity recording in man. Am J Clin Nutr 1989; 49: ''Livingstone MBE, Davies P S W, Prentice AM et al Comparison of simultaneous measures of energy intake and expenditure m children and adolescents. Proc Nutr Soc 1991; 50: 15A(abs.). l4 Mert2 W, Tsui J C, Judd JTetal What are people really eating? The relation between energy intake derived from estimated diet records and intake determined to maintain body weight. Am J Clin Nutr \99\; 54: l3 Lissner L, Habicht J P, Strupp B J, Levitsky D A, Haas J D, Roe D A. Body composition and energy intake: do overweight women overeat and underreport? Am J Clin Nutr 1989; 49: "Hallfrisch J, Steele P, Cohen L. Comparison of seven-day diet record with measured food intake of twenty-four subjects. Nutr Res 1982;! l7 Schofield W N, Schofiekl C, James W P T. Basal metabolic rate. Hum Nutr Cltn Nutr m5;39c(sappliy "Astrup A, Thorbek G, Lind J, Isaksson B. Prediction of 24-h energy expenditure and its components from physical characteristics and body composition in normal-weight humans. AmJ Clin Nutr 1990; 54: "Black A E, Goldberg G R, Jebb S A, Livingstone MBE, Cole T J, Prentice A M. Critical evaluation of energy intake data using fundamental principles of energy physiology: 2. Evaluating the results of published surveys. Eur J Clin Nutr 1991; 45: (Revised version received December 1993)