Under- and overreporting of energy intake related to weight status and lifestyle in a nationwide sample13

Lars Johansson, Kari Solvoll, Gunn-Elin Aa Bjrneboe, and Christian A Drevon

ABSTRACT Desire for weight change and level of dietary consciousness may severely bias reported food intake in dietary surveys. We evaluated to what degree under- and overreporting of energy intake (EI) was related to lifestyle, sociodemographic variables, and attitudes about body weight and diet in a nationwide dietary survey. Data were gathered by a self-administered quantitative food-frequency questionnaire distributed to a representative sample of men and women aged 1679 y in Norway, of whom 3144 subjects (63%) responded. Reported EI was related to estimated basal metabolic rate (BMR) based on self-reported body weight, age, and sex. An EI:BMR < 1.35 was considered to represent underreporting and an EI:BMR 2.4 as overreporting of EI. Fewer men than women underreported EI (38% compared with 45%). The fraction of overreporters did not differ significantly between sexes (7% of the men compared with 5% of the women). A large proportion of underreporters was obese (9%) and wanted to reduce their weight (41%). Few overreporters were obese and 12% wanted to increase their weight. Underreporters consumed fewer foods rich in fat and sugar than did the other subjects. Multiple regression analysis showed that desire for weight change and physical activity score were significantly correlated with both EI and EI:BMR when adjusted for sociodemographic and lifestyle variables. Our findings indicated that attitudes about ones own body weight influenced reported EI. These attitudes are important in the interpretation of dietary data because many of the subjects (> 30%) wanted to change their body weight. Am J Clin Nutr 1998;68:26674. KEY WORDS Energy intake, underreporting, basal metabolic rate, diet, body mass index, weight change, lifestyle, humans of under- or overreporting of EI (5). Such comparisons have shown that most dietary surveys underestimate habitual EI (1); however, little information about overestimation of EI is available. EI is typically lower when evaluated by dietary records than by dietary history (1). Thus, dietary surveys may include systematic errors in registered food consumption. Reported food and EI may also be biased by other factors such as age, sex, educational level, health consciousness, dieting (6), and degree of obesity (7, 8). Furthermore, smoking may increase energy metabolism as well as change food preferences and thereby influence body weight (9). If subjects under- or overreport their consumption of food items in proportion to their total food intake, it would be a minor problem in epidemiologic studies of diet and health. However, if study participants systematically under- or overreport consumption of selected foods, and if the degree of biased reporting is unevenly distributed within the population, the search for associations between diet and health variables may be disturbed. The increasing prevalence of obesity in many cultures (1012) and enhanced focus on the benefits of healthy dietary habits in nutritional education and marketing may aggravate the problem with biased reporting of food intake in dietary surveys. The aim of this study was to evaluate to what degree underand overreporting of EI was related to sex, age, body mass index (BMI), physical activity, smoking habits, sociodemographic variables, and attitudes about body weight and diet in a nationwide dietary survey. Furthermore, we wanted to examine whether under- or overreporters had more healthy food choices than other subjects. A major advantage with our present study compared with many other surveys was that we evaluated the effect that attitudes such as desire for weight change and dietary consciousness have on under- and overreporting of EI.

Downloaded from www.ajcn.org by guest on May 18, 2011

INTRODUCTION Bias of reporting energy intake (EI) in relation to energy expenditure is a well-known problem in dietary surveys (1). Energy expenditure is mainly determined by basal metabolic rate (BMR) and the level of physical activity (2, 3), and can be measured with good accuracy by the doubly labeled water (DLW) method (4). However, this method is costly and existing dietary survey methods suitable for use among large samples do not measure EI accurately. A comparison of reported EI with estimated BMR (EI:BMR) can be used to calculate the degree 266

1 From the National Nutrition Council, Oslo, and the Institute for Nutrition Research, University of Oslo. 2 Supported by a grant from the Ministry of Agriculture, the Ministry of Health and Social Affairs, and the Research Council of Norway. 3 Address reprint requests to L Johansson, National Nutrition Council, Box 8139 Department, N-0033 Oslo, Norway. E-mail: lars.johansson@se.dep.telemax.no. Received September 17, 1997. Accepted for publication January 9, 1998.

Am J Clin Nutr 1998;68:26674. Printed in USA. 1998 American Society for Clinical Nutrition

UNDER- AND OVERREPORTING OF ENERGY INTAKE SUBJECTS AND METHODS In a national dietary survey (NORKOST; 13), an optical-mark readable self-administered quantitative food-frequency questionnaire was distributed to a representative and randomly selected sample of Norwegians aged 1679 y, of whom 3144 subjects (63%) completed the questionnaire. The distribution of subjects in different groups of socioeconomic status, location of residence, and level of education was similar in NORKOST compared with the general population (14). Moreover, there were only small differences between the respondents and the total sample regarding age, sex, geographic distribution, and educational level (13). BMIs in our present study were also similar to measurements among 30 000 young men at military enrollment (14) and among 113 000 Norwegians examined by the National Health Screening Services (10). Of the 3144 respondents, 124 did not give information about weight or height. They were excluded and data from 3020 subjects were used in the present analysis. Five hundred one respondents reported having 1 or several special medical conditions: 130 had hypertension, 92 had hyperlipidemia, 66 had a food allergy, 32 had diabetes, 2 had anorexia, 2 had bulimia nervosa, 32 were pregnant, and 46 were lactating; 99 respondents reported having 2 of these conditions. Furthermore, 10 subjects were vegetarians and 28 did not eat a Norwegian diet because they were immigrants. These respondents were left in the sample because exclusion would not have affected the fraction of underand overreporters. The questionnaire was designed to cover the whole diet, including 180 food items. The frequency of consumption was given per day, per week, or per month, depending on the food item, and the portion sizes were units such as slices, glasses, cups, pieces, deciliters, and spoons. The portion sizes of the different food items were converted to weights on the basis of standard portions (15). Questions about weight, height, physical activity, smoking habits, meal frequency, and attitudes related to diet and body weight were also included. Statistics Norway provided information about the subjects level of education and other sociodemographic variables from their registers (14). A more detailed description of the subjects, the questionnaire, calculation of nutrients, and reported dietary habits was published previously (13, 16). The method used in our study was evaluated against 14-d weighed records (17) and 48-h recalls (18) and against the concentration of very-long-chain n 3 fatty acids in plasma phospholipids (19). These studies showed that the questionnaire could be used for assessing individual or group intakes of energy and a wide range of nutrients. Estimates of BMR were calculated from standard equations based on weight, age, and sex (20). EI:BMR was calculated for each individual and compared with cutoff values for EI:BMR of < 1.14, 1.141.34 (underreporters), 1.352.39 (normal range), and 2.4 (overreporters). The first cutoff value (1.14) was the lowest value for EI:BMR that could, within defined bounds of statistical probability, reflect actual EI over a given measured period (21). Because our questionnaire measured habitual intake during the past year, we used the value 1.14 suggested by Goldberg et al (21) for single individuals and long-term dietary records (28 d). Calculation of the lower cutoff limit allows for intraindividual variance in estimating EI, BMR, and physical activity level. In calculating this cutoff value, 1.55 BMR was used as the energy requirement for a sedentary lifestyle, against which reported intakes were evaluated. The second cutoff value

267

(1.35) was suggested as the lowest value for habitual EI of an individual that is compatible with a normal, not bedridden lifestyle (21). An EI:BMR above the range 2.02.4 is suggested as the maximum for a sustainable lifestyle (5). These cutoff values do not take into account the true energy expenditure of each individual. Body weight and height were reported by the participants to the nearest kg and cm, respectively. BMI was computed as the ratio between weight (kg) and squared height (m 2). The BMI values were divided into 4 subgroups: < 20 (lean), 2024 (normal weight), 2529 (overweight), and > 30 (obese). Attitudes related to body weight were evaluated by the question: Do you wish to change your weight? There were 3 possible answers: 1) No, 2) Yes, I want to reduce my weight, and 3) Yes, I want to increase my weight. An activity score was based on information about the subjects occupation and leisure time physical activity. The following scores were given for occupation: 1 = retired, disabled, and sick; 2 = white-collar workers (medium and highly ranked) and home-working subjects; 3 = skilled blue-collar workers, whitecollar workers (lowly ranked), pupils and students, other selfemployed workers, and occupation undefined or unknown; and 4 = farmers, fishermen, and unskilled blue-collar workers. The following scores were given for frequency of leisure time physical activity 20 min: 0 = never or less than weekly, 1 = 1 time/wk, 2 = 23 times/wk, 3 = 46 times/wk, and 4 = every day. The following scores were given for intensity of leisure time physical activity: 0 = easy exercise, not out of breath or sweating, and 2 = hard exercise, out of breath and sweating. Smoking habits were determined and subjects were divided into 3 categories: nonsmokers, light smokers (< 110 cigarettes or pipe-smoking episodes/d), and heavy smokers ( 11 cigarettes or pipe-smoking episodes/d). Dietary attention was evaluated by response to the following question: What attention do you pay to having a healthy diet? The following options were given: 1 = very low, 2 = low, 3 = medium, 4 = high, and 5 = very high attention. Location of residence was classified as rural (< 200 inhabitants), urban (3 categories: 2001999, 200019 999, or 20 00099 999 inhabitants), or cities ( 100 000 inhabitants) based on Norwegian standards (22). In this paper, dietary habits were described by percentage of dietary energy derived from fat, sugar, and alcohol, in addition to intakes of dietary fiber, vitamin C, and foods, calculated per 10 MJ. The Norwegian Data Inspectorate was notified according to standard procedures regarding the national dietary survey. Statistics Students t test and one-way analysis of variance (ANOVA) were used to test differences between groups. Bonferroni correction was made after the ANOVA. Differences in intake of foods and nutrients were tested with the Mann-Whitney U test. The chi-square test was used to test differences in proportion between categories. Multiple regression was used to test the influence of several variables on EI and EI:BMR simultaneously. In step 1 the association between the dependent variables EI and EI:BMR, respectively, and the single independent variables (desire to change weight, physical activity score, length of education, occupation, location of residence, smoking habits, dietary consciousness, and BMI) were tested with age and sex forced into the regression model. In step 2 the following independent variables were entered into the model in 1 block: desire

Downloaded from www.ajcn.org by guest on May 18, 2011

268

JOHANSSON ET AL

for weight change (desire to increase weight versus others and desire to reduce weight versus others), physical activity score (possible scores: 110), level of education (< 13 or 13 y), location of residence (rural district versus others, urban settlements with 2001999 inhabitants versus others, urban settlements with 20 00099 999 inhabitants versus others, and cities versus others), sex, age (1679 y), smoking habits (nonsmokers versus smokers), and dietary consciousness (possible scores: 15). Furthermore, BMI was related to EI but not to EI:BMR in this model because weight was included in the equation for computing both BMR and BMI. The Pearson correlation coefficient between BMI and EI:BMR was 0.28. Data were analyzed by using the program SPSS (23).

RESULTS EI and EI:BMR related to sex and age Men had a higher EI:BMR than women (1.58 compared with 1.48), and fewer men underreported EI (Table 1). Lactating women had a higher EI:BMR than other women of fertile ages (1.74 compared with 1.52, P = 0.006). Men had a higher physical activity score and BMI and paid less attention to eating a healthy diet than did women. More men than women were overTABLE 1 Energy intake, body weight status, and lifestyle variables of men and women1 Men (n =1461) Age (y) 42.7 16.12 Energy (MJ/d) 11.8 4.3 BMR (MJ/d) 7.5 0.7 EI:BMR 1.58 0.57 Height (cm) 179 7 Weight (kg) 79.2 11.9 BMI (kg/m2) 24.6 3.1 Physical activity score 5.2 2.0 Dietary consciousness 3.1 0.8 Percentage of subjects (%) EI:BMR4 <1.14 20 1.141.34 18 1.352.3 55 2.4 7 BMI4 <20 4 20-24 54 25-29 37 >30 5 Desire for weight change4 Increase 5 No change 76 Reduction 19 Smoking habits5 Nonsmokers 61 Light smokers <110 cigarettes/d 21 Heavy smokers 11 cigarettes/d 18
1 2

FIGURE 1. Percentage of underreporters of energy intake (EI) [ie, those with a ratio of EI to basal metabolic rate (EI:BMR) < 1.35] by different desires concerning change in body weight in 3 categories of BMI. Values in brackets are n values for the numbers of subjects desiring each weight change.

Downloaded from www.ajcn.org by guest on May 18, 2011

Women (n =1559) 41.6 16.7 8.4 2.93 5.8 0.53 1.48 0.543 166 63 64.8 11.23 23.4 3.83 4.9 1.93 3.4 0.73

weight and fewer men than women wanted to reduce their weight. EI and EI:BMR were highest in the youngest age groups in both sexes (Table 2), showing greater reduction with increasing age in men than in women. Underreporting of EI was more seldom and overreporting more common in the younger than in the older age groups. The physical activity score decreased with age in both sexes and the level of dietary consciousness increased with age. BMI increased with age up to 5059 y in men and up to 6069 y in women. The percentage of subjects with desire for weight reduction was lowest in the youngest and oldest age groups among men and in the oldest age group among women. Except for men in the youngest age groups, only a small fraction of men and women wanted to increase their weight. Characteristics of under- and overreporters Compared with subjects reporting EI:BMRs within the normal range, underreporters of EI were older, had lower physical activity scores, had higher BMIs, were more likely to be obese, and were more likely to want to reduce their weight (Table 3). The proportion of obese subjects and of subjects with desire for weight reduction was largest among severe underreporters (EI:BMR < 1.14). Compared with subjects reporting normal EI:BMRs, overreporters of EI were younger, had lower BMIs, were more likely to be lean, and were more likely to want to increase their weight. Few male overreporters wanted to reduce their weight, but as many as 22% of the female overreporters expressed a desire for weight reduction. The proportion of underreporters of EI increased with BMI and was highest in subjects who wanted to reduce their weight. This occurred in all categories of BMI (Figure 1). Even though the proportion of underreporters was high among obese subjects who wanted to reduce their weight, only 5% of the subjects were obese and as many as 651 (52%) of the 1249 underreporters had a BMI < 25. The proportion of overreporters was highest among subjects who wanted to increase their weight (Figure 2). Note that only 2 subjects were included in the category of subjects with BMIs of 2529 who had a desire to increase their weight. Most of the

25 20 50 5 15 59 21 5 2 60 38 64 22 14

EI, energy intake; BMR, basal metabolic rate. x SD. 3 Signicantly different from men, P< 0.001 (t test). 4,5 Signicant difference between sexes in all categories (chi-square test): 4 P< 0.001, 5 P<0.05.

TABLE 2 Energy intake, body weight status, and lifestyle variables related to age1
Age (y) 1619 y (n = 89) 2029 y (n = 269) 13.7 5.13 7.8 0.83 1.77 0.73 78 12 181 7 23.8 3.03 5.8 2.03 2.9 0.83 3039 y (n = 330) 12.3 4.0 7.6 0.6 1.63 0.5 80 12 181 7 24.6 3.2 5.3 1.7 3.1 0.8 Men 4049 y (n = 296) 11.3 3.3 7.6 0.5 1.51 0.5 80 11 179 6 24.9 2.9 5.2 1.8 3.2 0.7 5059 y (n = 221) 10.5 3.4 7.6 0.6 1.39 0.5 81 12 178 7 25.4 3.1 4.7 1.8 3.2 0.8 6069 y (n = 149) 9.7 3.0 6.9 0.5 1.41 0.4 80 11 178 7 25.1 2.8 4.2 1.9 3.4 0.8 7079 y (n = 107) 9.5 2.9 6.6 0.6 1.45 0.5 78 12 176 7 25.0 3.1 4.1 1.9 3.5 0.9 1619 y (n = 96) 2029 y (n = 369) 3039 y (n = 321) Women 4049 y (n = 289) 8.1 2.4 5.8 0.4 1.40 0.4 66 11 166 6 23.9 3.8 4.8 1.7 3.4 0.6 5059 y (n = 202) 7.7 2.3 5.8 0.3 1.33 0.4 66 10 166 6 24.0 3.1 4.7 1.7 3.4 0.8 6069 y (n = 166) 7.5 2.3 5.5 0.5 1.38 0.5 69 14 164 6 25.5 4.9 4.4 1.7 3.7 0.7 7079 y (n = 116) 7.6 2.3 5.4 0.4 1.44 0.5 66 11 163 6 24.7 4.1 4.0 1.9 3.9 0.9

Energy (MJ/d)2,3 15.4 5.63 BMR (MJ/d) 7.7 0.8 EI:BMR 2.01 0.73 Weight (kg) 72 113 Height (cm) 179 8 22.3 2.53 BMI (kg/m2) Activity score 6.7 1.93 Dietary consciousness 2.9 1.13 Percentage of subjects (%) EI:BMR6 < 1.14 8 1.14-1.34 8 1.35-2.39 62 2.4 23 BMI (kg/m2) 6 < 20 15 20-24 73 25-29 10 >30 2 Desire for weight change6 Increase 16 No change 81 Reduction 3 Smoking habits6 Smokers 23
1 2

9.8 3.43 9.1 3.53 8.8 3.0 6.0 0.53 6.0 0.63 5.7 0.4 1.66 0.64 1.55 0.74 1.55 0.54 60 83 63 104 64 12 167 6 167 7 168 63 21.4 2.63 22.4 3.23 22.8 3.83 6.8 1.43 5.2 1.93 4.7 1.8 3.2 0.85 3.2 0.73 3.4 0.7

UNDER- AND OVERREPORTING OF ENERGY INTAKE

14 11 62 13 8 62 27 3 9 73 18 38

14 19 58 9 2 57 37 4 3 75 22 47

19 22 55 4 3 51 41 5 2 75 23 45

30 19 48 2 1 47 43 9 2 81 17 37

31 21 46 2 3 43 49 5 3 71 26 32

24 19 53 4 4 50 40 6 4 86 10 22

17 16 58 9 33 60 5 1 2 62 35 32

24 17 52 7 21 61 16 2 4 59 37 40

16 22 57 5 20 62 12 6 2 64 34 44

26 22 50 2 9 61 23 7 2 54 44 41

33 26 39 2 6 62 26 5 1 60 39 31

34 19 44 3 4 52 34 10 58 42 30

27 20 50 3 10 47 34 9 2 70 28 13

BMR, basal metabolic rate; EI, energy intake. SD. x 3 Signicantly different from ages 4079 y, P <0.05 (one-way ANOVA with Bonferroni correction). 4 Signicantly different from ages 40 69 y, P <0.05 (one-way ANOVA with Bonferroni correction). 5 Signicantly different from ages 60 79 y, P<0.05 (one-way ANOVA with Bonferroni correction). 6 Signicant difference between age groups within sexes in all categories (chi-square test): P < 0.001, except for desire for weight change in women (P<0.05).

269

Downloaded from www.ajcn.org by guest on May 18, 2011

270

JOHANSSON ET AL

TABLE 3 Body weight status and lifestyle variables related to groups with different ratios of energy intake to basal metabolic rate (EI:BMR; <1.14 to 2.4) Men <1.14 (n = 286) Age (y) Energy (MJ/d) Body weight (kg) Body height (cm) BMI (kg/m2) Physical activity score Dietary consciousness Percentage of subjects (%) BMI (kg/m2)3 <20 2024 2529 3034 >35 Desire for weight change3 Increase No change Reduction Smoking habits4 Nonsmokers Light smokers, <110 cigarettes/d Heavy smokers, 11 cigarettes/d 1 x SD.
2 3

1.141.34 (n = 259) 45.6 15.12 9.4 1.02 81.7 11.62 180 7 25.3 2.92 4.9 1.82 3.1 0.8

1.352.39 (n = 808) 41.2 16.0 12.8 2.2 77.5 10.9 180 7 24.0 2.8 5.3 2.0 3.1 0.8

2.4 (n = 108) 32.4 13.92 21.9 5.22 74.2 9.72 179 6 23.1 2.52 5.9 1.92 3.0 1.1

<1.14 (n = 386)

Women 1.141.34 1.352.39 (n = 318) (n = 785) 40.3 16.6 9.6 1.6 62.2 9.2 166 6 22.5 3.1 5.0 1.8 3.4 0.7

2.4 (n = 70) 34.4 16.22 16.6 4.52 57.6 8.12 166 6 21.0 2.72 5.2 2.1 3.1 0.92

48.0 15.71,2 7.3 1.42 83.9 13.52 179 7 26.1 3.52 4.7 1.92 3.3 0.8

44.6 17.22 42.9 15.9 5.5 1.12 7.2 0.72 69.8 13.62 66.5 10.52 166 6 166 6 25.2 4.72 24.0 3.62 4.6 1.82 4.8 1.92 3.4 0.8 3.4 0.7

0.3 38 50 11 1 1 64 35 67 17 16

1 43 52 4 1 2 72 26 64 17 19

6 62 30 3 0.1 6 80 14 60 22 17

9 70 20 1 12 87 1 49 32 19

7 47 34 9 3 0.3 45 54 62 23 15

8 61 25 5 2 0.6 57 43 68 18 14

19 66 13 2 1 3 68 29 64 22 14

44 47 9 12 66 22 50 30 20

Downloaded from www.ajcn.org by guest on May 18, 2011

Signicantly different from EI: BMR category 1.352.39, P < 0.05 (one-way ANOVA with Bonferroni correction). Signicant difference between EI: BMR categories within sexes in all categories, P < 0.001 (chi-square test). 4 Signicant difference between EI: BMR categories within sexes in all categories, P < 0.05 for men (chi-square test).

178 overreporters (84%) had a BMI <25. Although a desire for weight change as well as BMI were good predictors for under- and overreporting of EI, only a minor part of the total number of underand overreporters could be identied with information about BMI and attitudes about change of body weight. Smoking was more common among overreporters than among subjects with an EI:BMR within the normal range. The proportion of underreporters was lower (39% compared with 43%, P = 0.02) and the proportion of overreporters was higher (8% compared with 5%, P <0.001) among smokers than nonsmokers. Smokers had a lower BMI and a lower physical activity score than nonsmokers. Despite this, male smokers reported a higher EI (12.3 compared with 11.5 MJ/d) and a higher EI:BMR (1.7 compared with 1.5) than nonsmokers. Dietary habits Daily intake of most foods was lower among underreporters of EI and higher among overreporters than among subjects with an EI:BMR within the normal range. After the daily intake of foods per 10 MJ EI was computed, both male and female underreporters had a lower intake of foods rich in fat and sugar, such as cakes, potato chips, edible fats, chocolate and sweets, and sugar-containing soft drinks, and a higher intake of potatoes, meat, sh, and nonalcoholcontaining beverages than subjects with an EI:BMR in the normal range (Table 4). Underreporters also had a lower percentage of energy derived from fat and sugar in their diet, whereas the intake of ber and vitamin C per 10 MJ was higher than that in subjects with an EI:BMR in the normal range. Male as well as female overreporters of EI had a higher intake of potato chips and whole milk, and more energy derived from sugar, than subjects with a normal EI:BMR.

Multiple regression analyses EI and EI:BMR were positively correlated with desire to increase weight, physical activity score, education, blue-collar work, and residence in rural areas and were negatively correlated with desire to reduce weight, after adjustment for age and sex (Table 5; step 1). Furthermore, EI was negatively correlated with degree of dietary consciousness and BMI, and EI:BMR was positively correlated with smoking. When data from men and women were analyzed separately, smoking was positively correlated with both EI and EI:BMR in men, but not in women. Desire for weight change and physical activity score remained signicantly correlated with both EI and EI:BMR when adjusted for sociodemographic and lifestyle variables (Table 5; step 2). Together the independent variables explained 29% and 13% of the variation in EI and EI:BMR, respectively. BMI was only included in the model for EI. When BMI was included in the model for EI:BMR, all the independent variables remained signicantly correlated with EI:BMR, except for smoking habits, and the explained variation increased from 13% to 16%.

DISCUSSION Level of EI Evaluation studies have shown that the questionnaire used in our dietary survey can be used for assessing EI among individuals and groups. The questionnaire gave a higher EI than did a 48-h recall in adult men and women (9.2 compared with 8.6 MJ/d) (18). When EI according to the questionnaire was compared with that from 14-d weighed food records from elderly

UNDER- AND OVERREPORTING OF ENERGY INTAKE women, the difference was not statistically significant (8.7 compared with 7.8 MJ/d) (17), although this may have been due to the relatively low number of participants. The correlation coefficient between EI reported with the food-frequency questionnaire and the 48-h recall was 0.47, and was 0.61 between the questionnaire and 14-d weighed records. In 3 reproducibility studies the correlation coefficients for EI were 0.54, 0.67, and 0.77, respectively, when the present questionnaire was used (1618). The reported EI among men aged 1629 y in our study was within the range of the reference values for seated work with regular leisure activity; men aged 3079 y had values within the range of reference values for seated work with low leisure activity (24). The reported EI among women in our survey was below the Nordic reference values. Measurements with DLW suggested a daily energy expenditure of 1.41.7 BMR for seated work and 2.02.4 BMR for individuals with strenuous physical activity (5). The EI:BMR in our study was 1.58 among men and 1.48 among women, which indicates a sedentary lifestyle. A review of 37 dietary surveys showed a mean EI:BMR of 1.50 for men and 1.37 for women (1). The mean EI:BMR differed between methods and was 1.31, 1.47, and 1.60 when the intake was assessed by 24-h recall, dietary records, and dietary history, respectively. Thus, the EI:BMR reported in the present study was higher than that typically reported by 24-h recall and dietary records and lower than that by dietary history. Most dietary surveys underestimate habitual EI when it is evaluated against the calculated EI:BMR (1).

271

FIGURE 2. Percentage of overreporters of energy intake (EI) [ie, those with a ratio of EI to basal metabolic rate (EI:BMR) 2.4] by different desires concerning change in body weight in 3 categories of BMI. Values in brackets are n values for the numbers of subjects desiring each weight change.

Downloaded from www.ajcn.org by guest on May 18, 2011

Level of under- and overreporting of EI The average proportion of underreporters of EI in our study was of the same magnitude (4050%) as shown in national surveys in

TABLE 4 Intake of foods and nutrients in subgroups with different ratios of energy intake to basal metabolic rate (EI:BMR; <1.14 to 2.4)1 < 1.14 (n = 286) Food intake (g/10 MJ) Bread and cereals 262 80 Cakes 25 302 Fruit and berries 187 163 Vegetables 120 93 Potatoes, total 177 1052 Potato chips 2 52 Meat and meat products 121 542 Fish and sh products 79 642 Milk, total 546 348 Whole milk (3.8% fat) 81 2072 Skimmed milk (0.1% fat) 189 3232 Edible fats, total 25 182 Chocolate and sweets 6 122 Soft drinks, 121 2492 sugar containing Soft drinks, 110 328 non-sugar containing Coffee 704 5242 Beer, wine, liquor 146 212 Nutrient intake Fat (% of energy) 29 62 Sugar (% of energy) 7 52 Alcohol (% of energy) 23 Dietary ber (g/10 MJ) 24 72 Vitamin C (mg/10 MJ) 96 552 1 x SD.
2

1.14-1.34 (n = 259)

Men 1.35-2.39 (n = 808) 252 65 30 27 182 138 108 72 130 68 36 110 42 68 48 487 255 83 173 94 195 37 21 9 13 189 254 66 195 426 340 108 149 32 5 95 22 22 6 86 42

2.4 (n = 108) 217 632 36 39 169 137 84 542 105 752 6 112 101 35 49 462 444 216 125 1912 38 1202 45 212 16 232 315 3392 30 842 286 2612 132 279 34 62 12 72 23 19 52 75 382

< 1.14 (n = 386) 260 722 24 222 284 229 190 1372 160 1072 3 62 119 552 80 632 460 3132 53 163 181 284 26 202 8 142 105 2172 187 4392 763 6942 65 105 28 62 7 52 12 28 82 135 782

Women 1.141.34 1.352.39 (n = 318) ( n = 785) 238 59 29 232 249 159 168 102 147 732 3 62 120 472 74 452 504 328 43 114 182 306 31 212 11 16 119 1742 162 3692 577 4652 73 126 30 62 8 52 12 25 72 118 58 231 57 36 29 259 176 154 100 124 67 36 105 40 66 41 512 283 60 153 143 254 35 22 11 14 171 242 111 300 455 413 63 149 31 6 95 12 24 6 113 58

2.4 (n = 70) 207 712 42 38 261 232 124 732 110 66 8 142 87 372 60 522 445 242 113 1672 59 592 34 20 16 22 276 390 129 270 290 2892 64 111 32 6 12 82 11 22 62 107 63

258 65 30 29 184 137 116 71 156 882 34 116 48 70 50 481 291 55 1342 121 240 32 212 7 10 139 1722 81 235 575 4112 128 164 30 52 8 42 23 23 62 90 42

Signicantly different from EI:BMR category 1.352.39, P <0.05 (Mann-Whitney U test).

272

JOHANSSON ET AL

TABLE 5 Energy intake and energy intake to basal metabolic rate (EI:BMR) related to weight status, lifestyle, and sociodemographic variables1 Energy intake (kJ/d) R2 -coefcient P Step 12 Desire to reduce weight versus others Desire to increase weight versus others Physical activity score, 110 Education, <13 versus 13 years Rural districts versus cities Blue versus white collar workers Smokers versus nonsmokers Dietary consciousness, 15 BMI, 1458 Step 23 Desire to reduce weight versus others Desire to increase weight versus others Physical activity score, 110 Age, 1679 y Sex, men versus women Education, <13 versus 13 years Rural versus other districts Smokers versus nonsmokers Dietary consciousness, 15 BMI, 1458
1 2

EI:BMR R2 -coefcient P <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.056 0.016 <0.001 9.6 7.0 5.9 5.2 6.0 5.7 5.3 5.3 0.268 0.444 0.027 0.090 0.111 0.087 0.053 -0.024 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.010 0.060

26.5 25.8 25.8 25.8 26.2 25.6 25.2 25.4 25.6

1106 1937 186 628 853 653 252 201 78

918 1332 199 54 3137 450 549 224 146 8

<0.001 <0.001 <0.001 <0.001 <0.001 0.004 0.003 0.095 0.089 0.680

0.239 0.344 0.029 0.005 0.052 0.067 0.060 0.043 0.009

<0.001 <0.001 <0.001 <0.001 0.007 0.005 0.030 0.033 0.498

Downloaded from www.ajcn.org by guest on May 18, 2011

Adjusted R2,

coefcient, and P values are for independent variables in models for multiple regression.

Each independent variable was entered separately together with age and sex forced into a regression model. Sex and age together explained 25% and 5% of the variation in energy intake and EI:BMR, respectively. 3 Independent variables were included in 1 block in a regression model: desire for weight change, physical activity score, smoking habits, level of education, residence, degree of dietary consciousness, age and sex. BMI was included in the model for energy intake, but not for EI:BMR. R2 values for energy intake and EI:BMR are 29.2 and 12.6, respectively.

the United Kingdom (25), United States (26), and Finland (27). In our study, 6% of the subjects reported EIs above the range suggested as the maximum for a sustainable lifestyle. Because few data are published about overreporting from national surveys, we did not know what proportion of overreporting we could expect to find. In DLW studies, an EI:BMR 2.4 was measured only in soldiers, athletes, and explorers with strenuous, long-endurance physical activity (5). The possibility of finding more than a few subjects with such heavy activity in our random sample was very low. When differences in the proportion of subjects having an EI:BMR below or above cutoff values used in our study are compared, it is of great importance to keep in mind that these values are blanket definitions that do not take into account the true energy expenditure of each individual. The FAO/WHO/UNU suggested that the average daily energy requirements for men are 1.55, 1.78, and 2.10 BMR (in MJ/d) for light, moderate, and heavy occupational work, respectively, whereas the requirements for women are 1.56, 1.64, and 1.82 BMR (2). The cutoff values used by us assume a sedentary lifestyle among all subjects. DLW measurements have shown that assuming a mean energy expenditure for an inactive population can identify only a proportion of underreporters (28). Furthermore, DLW measurements have shown that underreporting as well as overreporting of EI may occur across all levels of energy expenditure (29). Only those EIs that are outside the statistically probable limits of normal variation (EI:BMR < 1.14) can be identified as biased. Adequate identification of false reporting across the full range of

EIs requires knowledge of an individuals physical activity. We had no way to assign individual- or group-adjusted cutoffs for the EI:BMRs in our study. Under- and overreporting of EI related to sex and age Women were more likely to underreport EI than were men in our study as well as in other national studies (2527) when the same cutoff value for both sexes was used. This was also true when DLW was used, showing that fewer men than women underreported EI (16% compared with 35%), whereas the percentage of overreporters was much larger among men than among women (15% compared with 2%) (29). When adjusted for body size, men had an 11% greater energy expenditure (5) and a higher mean EI:BMR (1, 5) than women. The mean EI:BMR measured with DLW was 1.62 for women and 1.78 for men (21). If sex-specific EI:BMR values were taken as true population energy expenditures and used in place of 1.55 to derive cutoff values, the difference in proportion of underreporters between sexes in our study would decrease. In our study fewer underreporters and more overreporters were found in the youngest than in the middle-aged groups. Even though absolute EI:BMR values in our study were lower than those reported in studies using the DLW method (5), they peaked in the youngest age groups and declined with age, as expected. However, in our study fewer of the 7079-y-olds than the 5059-y-olds were underreporters. This may have been related to the relatively high proportion of physically active respondents in the oldest age group. Men and women aged 7079 y in our study

UNDER- AND OVERREPORTING OF ENERGY INTAKE had EI:BMRs similar to those reported in 14-d weighed records for Norwegian elderly women (17), as well as those reported by dietary history among Swedish 70-y-old men and women (30). Although our physical activity score is crude, it was positively associated with EI and EI:BMR. It was higher in the youngest age groups in both sexes and higher in men than in women. Male and female underreporters had lower physical activity scores and male overreporters had higher scores than subjects with normal EI:BMRs. Thus, some of the variation in EI:BMR, as well as some of the differences in percentages of under- and overreporting between subgroups, may be explained by real differences in energy expenditure because of differences in physical activity. Because body weight was used for estimating BMR, the degree of under- and overreporting may also be biased because of self-reported weight. On average, self-reported weight and height were given with small errors when evaluated against measurements in large samples of the US (31) and Finnish (32) populations. However, errors in self-reported weight were directly related to BMI. Underweight American men and women overreported their weight by an average of 2.3 and 0.4 kg, respectively, whereas overweight men and women underreported their weight by 1.4 and 3.4 kg, respectively (31). Lifestyle and social class Male smokers had a higher EI and EI:BMR and smokers of both sexes had a lower physical activity than nonsmokers. Despite this, male and female smokers had a lower BMI than nonsmokers. A higher EI and a lower BMI in smokers than in nonsmokers were also found in the Multiple Risk Factor Intervention Trial (33). Because smoking increases energy metabolism (9), the difference in EI:BMR between smokers and nonsmokers observed in our study may be explained by a higher energy expenditure among smokers. Contrary to our present findings, underreporting of EI was more common in British smokers (34) and US female smokers than in nonsmokers (6). Furthermore, underreporting of EI was more common in subjects with a short education (< 12 y) than in those with a long education ( 12 y) in 2 US surveys (6, 26) and in British manual social classes (34). In our survey, short education, blue-collar work, and residence in rural areas were positively associated with EI and EI:BMR. Body weight status DLW measurements have shown that habitual energy expenditure is substantially raised in obese individuals (8), who tend to underreport EI (1, 35, 36). Underreporting was also more common in obese than in lean subjects in our study. The physical activity score did not differ significantly between groups of subjects with different BMIs or desire to change weight except that obese women had a lower physical activity score than lean women. Men probably use exercise for weight control more than do women, whereas women may preferentially use dietary restriction (37). In our study, BMI was negatively correlated with EI, but this correlation was no longer significant when we adjusted for desire to change weight. In an American survey, adjustment for dieting among the participants also reduced the inverse correlation between EI and BMI (6). Our findings indicated that individual attitudes about their own body weight strongly influenced reported intake, even if they were asked to report the usual food intake during the previous year. This is of great importance

273

because > 30% of all the participants wanted to change their body weight. Among lean women, 12% wanted to increase their weight and as many as 22% wanted to reduce their weight. Even if attitude about weight was strongly correlated with EI:BMR, the other attitude variable included in our analysisdegree of dietary consciousnesswas not correlated with EI:BMR. Although desire for weight change as well as BMI were good predictors of under- and overreporting of EI in the present study, only a small part of the total number of under- and overreporters could be identified with information about BMI and attitudes about change in body weight. Conclusions Our study showed that a large fraction of participants underreported EI and that a small fraction reported EI that was higher than likely. Underreporting was more common among women than among men and in the middle-aged than in the youngest age groups when related to cutoff values for a sedentary lifestyle. However, some of the variation in EI:BMR, as well as in differences in the percentage of under- and overreporters between subgroups, may be explained by real differences in energy expenditure due to differences in physical activity and smoking habits. Under- and overreporting of EI were strongly associated with BMI and attitudes about body weight. Desire for weight reduction was associated with underreporting both among normalweight and overweight subjects. Although reported desire for weight change was a good predictor of under- and overreporting of EI, the largest number of under- and overreporters in our survey was found among normal-weight subjects and subjects who did not want to change their body weight. Our findings indicate that it is important to detect the degree and distribution of misreported EI because this may be important in the evaluation of the relation between diet and health.
The study was carried out in cooperation with the Norwegian Food Control Authority. The contributions of Christina Bergsten, Bodil Blaker, Elin B Lken, and Gunnar mlid were greatly appreciated.

Downloaded from www.ajcn.org by guest on May 18, 2011

REFERENCES
1. Black AE, Goldberg GR, Jebb SA, Livingstone MB, Cole TJ, Prentice AM. Critical evaluation of energy intake data using fundamental principles of energy physiology. 2. Evaluating the results of published surveys. Eur J Clin Nutr 1991;45:58399. 2. WHO. Energy and protein requirements. Report of a Joint FAO/WHO/UNU Expert Consultation. World Health Organ Tech Rep Ser 1985:724. 3. Bahr R. Excess post exercise oxygen consumptionmagnitude, mechanisms and practical implications. PhD thesis. University of Oslo, 1991. 4. International Dietary Energy Consultant Group. The doubly labelled water method for measuring energy expenditure: technical recommendations for use in humans. Vienna: International Atomic Energy Agency, 1990. 5. Black AE, Coward WA, Cole TJ, Prentice AM. Human energy expenditure in afuent societies: an analysis of 574 doubly-labelled water measurements. Eur J Clin Nutr 1996;50:7292. 6. Ballard-Barbash R, Graubard I, Krebs-Smith SM, Schatzkin A, Thompson FE. Contribution of dieting to the inverse association between energy intake and body mass index. Eur J Clin Nutr 1996;50:98106. 7. Livingstone MBE. Assessment of food intakes: are we measuring what people eat? Br J Biomed Sci 1995;52:5867. 8. Prentice AM, Black AE, Coward WA, Cole TJ. Energy expenditure in overweight and obese adults in affluent societies: an analysis of

274

JOHANSSON ET AL
319 doubly-labelled water measurements. Eur J Clin Nutr 1996;50:937. Moffatt RJ, Owens SG. Cessation from cigarette smoking: changes in body weight, body composition, resting metabolism and energy consumption. Metabolism 1991;40:46570. Tverdal A. Height, weight and body mass index for men and women aged 4042 years. Tidsskr Nor Laegeforen 1996;116:21526. Seidell JC. Obesity in Europe: scaling an epidemic. Int J Obes Relat Metab Dis 1995;19(suppl 3):14. Galuska DA, Serdula M, Pamuk E, Siegel PZ, Byers T. Trends in overweight among US adults from 1987 to 1993: a multistate telephone survey. Am J Public Health 1996;86:172935. Johansson L, Solvoll K, Bjrneboe G-EA, Drevon CA. Dietary habits among Norwegian men and women. Scand J Nutr 1997;41:6370. Statistics Norway. Statistical yearbook 1995. Official statistics of Norway. Oslo: Grndahl og Dreyers Forlag A/S, 1995. Blaker B, Aarsland M. Household measures and weights of foods. Oslo: National Association for Nutrition and Health, 1989 (in Norwegian). Johansson L, Solvoll K, Opdahl S, Bjrneboe G-EA, Drevon CA. Response rates with different distribution methods and reward, and reproducibility of a quantitative food frequency questionnaire. Eur J Clin Nutr 1997;51:34653. Nes M, Frost Andersen L, Solvoll K, et al. Accuracy of a quantitative food frequency questionnaire applied in elderly Norwegian women. Eur J Clin Nutr 1992;42:80921. Solvoll K, Lund-Larsen K, Syland E, Sandstad B, Drevon CA. A quantitative food frequency questionnaire evaluated in a group of dermatologic outpatients. Scand J Nutr 1993;37:1505. Frost Andersen L, Solvoll K, Drevon CA. Very long-chain n 3 fatty acids as biomarkers for intake of fish and n 3 fatty acid concentrates. Am J Clin Nutr 1996;64:30511. Commission of The European Communities. Reports of the Scientific Committee for Food: nutrient and energy intakes for the European Community. Luxembourg: Office for Official Publications of the European Communities, 1992. Goldberg GR, Black AE, Jegg SA, et al. Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. Derivation of cut-off limits to identify underrecording. Eur J Clin Nutr 1991;45:56981. Statistics Norway. Standard classication of municipalities. Ofcial Statistics of Norway. Oslo: Akademica, 1994. SPSS Inc. SPSS for Windows, release 6.1. Chicago: SPSS Inc, 1994. Sandstm B, Lyhne N, Pedersen JI, Aro A, Thorsdottir I, Becker W. Nordic nutrition recommendations 1996. Scand J Nutr 1996;40:1615. Gregory J, Foster K, Tyler H, Wiseman M. The dietary and nutritional survey of British adults. London: Her Majestys Stationery Office, 1990. Klesges RC, Eck LH, Ray JW. Who underreports dietary intake in a dietary recall? Evidence from the second National Health and Nutrition Examination Survey. J Consult Clin Psychol 1995;63:43844. Hirvonen T, Mnnist S, Roos E, Pietinen P. Increasing prevalence of underreporting does not necessarily distort dietary surveys. Eur J Clin Nutr 1997;51:297301. Black AE, Johansson G, Welch A, Bingham SA. Energy intake: basal metabolic rate (EI:BMR) for the evaluation of reported energy intake of individuals. Proc Nutr Soc 1997;56:122A. Black AE. Under-reporting of energy intake at all levels of energy expenditure: evidence from doubly labelled water studies. Proc Nutr Soc 1997;56:121A. Rothenberg E, Bosaeus I, Steen B. Evaluation of energy intake estimated by a diet history in three free-living 70 year old populations in Gothenburg, Sweden. Eur J Clin Nutr 1997;51:606. Rowland ML. Self-reported weight and height. Am J Clin Nutr 1990;52:112533. Jalkanen L, Tuomilehto J, Tanskanen A, Puska P. Accuracy of selfreported body weight compared to measured body weight. A population survey. Scand J Soc Med 1987;15:1918. Stamler J, Rains-Clearman D, Lenz-Litzow K, Tillotson JL, Grandits GA. Relation of smoking at baseline and during trial years 16 to food and nutrient intakes and weight in the special intervention and usual care groups in the Multiple Risk Factor Intervention Trial. Am J Clin Nutr 1997;65(suppl):374S402S. Pryer JA, Vrijheid M, Nichols R, Kiggins M, Elliot P. Who are the low energy reporters in the dietary and nutritional survey of British adults? Int J Epidemiol 1997;26:14654. Lichtman SW, Pisarska K, Berman ER, et al. Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. N Engl J Med 1992;27:18938. Heitmann BL, Lissner L. Dietary underreporting by obese individualsis it specific or non-specific? Br Med J 1995;311:9869. Drewnowski A, Yee DK. Men and body image: are males satisfied with their body weight. Psychosom Med 1987;49:62634.

9.

25.

10. 11. 12.

26.

27.

13.

28.

14. 15.

29.

30.

16.

31. 32.

Downloaded from www.ajcn.org by guest on May 18, 2011

17.

18.

33.

19.

20.

34.

21.

35.

36. 37.

22. 23. 24.