Sociodemographic and nutritional determinants of obesity in ... - Nature

International Journal of Obesity (1999) 23, Suppl 1, 1±9 ß 1999 Stockton Press All rights reserved 0307±0565/99 $12.00 http://www.stockton-press.co.uk/ijo

Sociodemographic and nutritional determinants of obesity in Belgium MC Stam-Moraga*1, J Kolanowski2, M Dramaix3, G De Backer4 and MD Kornitzer1 1

Laboratory of Epidemiology and Social Medicine, School of Public Health, UniversiteÂ Libre de Bruxelles, Brussels; 2School of Medicine, UniversiteÂ Catholique de Louvain, Brussels; 3Laboratory of Medical Statistics, School of Public Health, UniversiteÂ Libre de Bruxelles, Brussels and 4Department of Public Health, University of Gent, Belgium

OBJECTIVE: To examine associations between sociodemographic, dietary factors and physical activity and the prevalence of obesity in the Belgian general population. DESIGN: Base-line data from the Belgian Interuniversity Research on Nutrition and Health (BIRNH) study (1979 ± 1984). SUBJECTS: A total of 5837 men and 5243 women aged 25 ± 74 y were included in the analysis. The sample was considered representative of the Belgian population. MEASUREMENTS: Using the body mass index (BMI) as the criterion, obesity was de®ned as a BMI 30 kg=m2. Nutritional data were assessed by a 24 h food record. Physical activity level (PAL) was calculated by dividing total caloric intake by an estimation of basal metabolic rate (BMR, predicted from gender, weight and age). Age-adjusted odds ratios (OR) of the prevalence of obesity were estimated by multilogistic regression analysis. RESULTS: Prevalence of obesity was 12.1% in men and 18.4% in women. In both sexes, prevalence of obesity increased gradually in each 10-year age category (P < 0.0001) and steeply decreased with level of education. Comparing lowest (Q1) to highest quartile (Q4), after adjustment for age, prevalence of obesity decreased with total carbohydrate intake, as well as total sugar intake in both sexes, and increased with total fat intake only in men. Obesity increased with a high fat to sugar ratio (men: ORQ4=Q1 1.56; con®dence interval (CI): 1.25 ± 1.93; women: ORQ4=Q1 1.45; CI: 1.17 ± 1.80). PAL was inversely and very strongly associated with obesity (men: ORQ4=Q1 0.20; CI: 0.15 ± 0.26; women: ORQ4=Q1 0.18; CI: 0.14 ± 0.23). The same associations were observed with the mean BMI. CONCLUSION: This study indicates that prevalence of obesity is particularly high in Belgium. Low level of education and reduced physical activity, increased fat intake and especially elevated fat to sugar ratio appear to be powerful determinants of obesity in this Belgian population. Keywords: obesity; Belgium; dietary factors; sociodemographic factors

Introduction Obesity is a condition af¯icting a growing number of people in most industrialised countries. In a recent press release from the World Health Organisation (WHO), it has been stated that the prevalence of obesity in adults (de®ned as a body mass index (BMI) 30 kg=m2) is 10 ± 25% in most countries of western Europe.1 The experts agree that the principal causes of this worldwide escalating problem are sedentary lifestyles and high-fat, energy-dense diets, both as a consequence of the profound societal and behavioural changes over the last 20 ± 30 years. Although it is certain, from studies focused on the patho-physiological mechanisms of obesity,2 that obesity develops only when energy intake chronically exceeds energy expenditure, epidemiological studies show that obesity is in¯uenced in a complex manner, as well by sociodemographic factors3 ± 19 (such as age, socioeconomic status and marital status), as by

*Correspondence: Margaretha C Stam-Moraga, Laboratoire d'Epidemiologie et de MeÂdecine Sociale, Ecole de SanteÂ Publique, UniversiteÂ Libre de Bruxelles, CP697 808 Route de Lennik B-1070 Brussels.

behavioural and environmental factors,20 ± 38 including food intake and physical activity. In Belgium, the size of the problem of obesity at the general population level has barely been evaluated and neither is there any information available on the sociodemographic distribution of obesity, nor on the lifestyle factors associated with this disease. Using the base-line data from a nutritional survey (Belgian Interuniversity Research on Nutrition and Health (BIRNH)39 study, carried out between 1979 and 1984) the aim of our study is to estimate the prevalence of obesity in a sample, representative of the Belgian general population, and to investigate the associations of obesity with sociodemographic factors, nutritional patterns and physical activity. Even though the data are not very recent, it seems important to publish such information for Belgium. Moreover, determinants of obesity have probably not changed over the last decade.

Materials and methods Study population

The BIRNH study of 1979 ± 1984, aimed to investigate the regional distribution of cardiovascular risk

Obesity in Belgium MC Stam-Moraga et al

2

factors and nutritional habits in Belgium and their relation with cardiovascular mortality. The methodology and results were published elsewhere.39 ± 42 To summarise, in each of 42 Belgian districts 3 ± 5 communes were selected at random (except for the most populated commune which was added ex of®cio) and in each commune, a strati®ed sample with 100 subjects for each 10 y=gender stratum was selected at random from the communal voting lists. In total 11 302 subjects took part in the study (5949 men and 5353 women). As the participation rate was low (36.7%), a 10% non-responder sample, selected at random, was invited to answer questions related to nutritional habits and other lifestyle habits. No major differences were found between participants and non participants. The dietary habits of each participant were assessed using a 24 h food record: the participant was sent a standardised questionnaire several days before the effective day of record. The day following the dietary registration, a well trained dietician veri®ed and quanti®ed the food records; these were then translated into quantities for 156 food items. The conversion program of all food items into nutrients was based on the Dutch Food Composition Tables,43 on the tables from Paul and Southgate44 and on food composition data available in Belgium for local food items (for example, margarine). Before analysis, the macronutrient intakes, initially expressed in grams per day, were translated into percent of total calorie intake using Atwater conversion's factors.45 A diet composition index of total fat with regard to total sugar was calculated by dividing total fat intake by total sugar intake. Physical activity level (PAL) was calculated by dividing the subject's total calorie intake by their basal metabolic rate (BMR); with BMR predicted from gender, age and body-weight (W) according to the FAO=WHO equations.46 This approach seems reasonable provided that: 1) energy balance is maintained and 2) reported calorie intake corresponds to the subject's real intake. If these conditions are present, physical activity may be measured by energy expenditure as well as by energy intake.28 All nutritional variables were categorised by dividing the data into gender-speci®c quartiles. During the medical examination, height and weight were measured on a scale type Seca 710, with the subjects undressed to the waist and shoes off. Obesity was de®ned as a BMI 30 kg=m2. Sociodemographic data were collected by a standardised questionnaire. They were classi®ed as follows: age (25 ± 34 y, 35 ± 44 y, 45 ± 54 y, 55 ± 64 y, 65 ± 74 y), educational level according to the highest grade completed (primary, secondary, technical, university), region of residence (Dutch speaking, French speaking), and marital status (single, married, divorced or separated, widowed).

Statistical methods

All analyses were strati®ed by gender. Univariate relationships of sociodemographic and dietary variables, with BMI in its continuous form, were performed by analysis of variance. In addition, multiple means comparisons were carried out using the KeulsNewman-Student test.47 Bivariate analyses, correcting for age, were performed with analysis of covariance models. For the prevalence of obesity, univariate analyses were performed by Chi-square tests with the calculations of corresponding odds ratios (OR) and 95% con®dence intervals (CI). Age-adjusted OR (CI) of the proportion of obesity were estimated in multilogistic regression models. All analyses were carried out using SPSS 4.0 software.48 The global level of statistical signi®cance was set at a 0.05.

Results The characteristics of the subjects are presented in Table 1. Mean BMI did not differ between the sexes, but when classifying BMI according to Garrow's obesity grades,49 we observed that overweight (BMI 25 and < 30) was more frequently encountered in men (46.5% vs 35.2% in women), while obesity (BMI 30) was more prevalent in women (18.4% vs12.1% in men). Sociodemographic factors and obesity (Table 2 and Table 3)

In univariate analysis, mean BMI, as well as prevalence of obesity, were largely in¯uenced by age. In men, the highest average BMI was observed for the age group 55 ± 64 y, and the proportion of BMI 30 was 2.64 times higher in this age category compared to the group 25 ± 34 y. In women, the positive gradient of BMI with age was even steeper and the prevalence of obesity increased seven-fold between 25 ± 34 y and 65 ± 74 y. In bivariate analysis, no signi®cant relationships were found between region, marital status and BMI (either in its continuous form or expressed as a proportion 30) in men. In women, average BMI and prevalence of obesity were higher in the Dutch speaking subjects. Marital status had an overall effect on BMI, with married women having a signi®cantly higher BMI than single and divorced=separated women, and single women having signi®cantly lower BMI than divorced=separated and widowed women. Even after adjustment for age, a highly signi®cant inverse gradient was found in both sexes between level of education and average BMI, as well as the prevalence of obesity. This negative association was however more pronounced in women, in whom multiple means comparisons between the categories showed


3

Table 1 Characteristics of the study sample (n, means (s.d.) or percentages) Men Continuous variables

n

Age (y) Height (m) Weight (kg) BMI (kg=m2) Total fat intake (%EI) Total carbohydrate intake (%EI) Total sugar intake (%EI)

Mean (s.d.)

5837 5837 5837 5837 5732 5742 5732

Categorical variables

Women

49.2 1.72 77 25.95 41.8 38.7 14.3

n

Region French speaking Dutch speaking Marital status Married Single Divorced=separated Widowed Level of education Primary Secondary Technical University

n

(13.5) (0.07) (12) (3.55) (8.4) (8.0) (6.4)

5243 5243 5243 5243 5165 5165 5165

%

2242 3594

38.4 61.6

4991 404 207 197

86.1 7.0 3.5 3.4

3403 1412 668 313

58.7 24.4 11.5 5.4

Mean (s.d.) 48.3 1.60 66 26.01 42.6 39.3 16.9

n

(13.1) (0.06) (12) (4.59) (8.4) (8.2) (7.0)

%

5241 1980 3261 5219 4260 264 196 499 5220 3537 1135 466 82

37.8 62.2 81.6 5.1 3.7 9.6 67.8 21.7 8.9 1.6

BMI body mass index; %EI % of energy intake.

Table 2 Mean body mass index (BMI, crude and age-adjusted) by age, region, marital status and level of education (mean (s.d.)). Strati®cation by gender BMI (kg=m2) Men Sociodemographic variables Age groups (y) 25 ± 34 35 ± 44 45 ± 54 55 ± 64 65 ± 74 Total 25 ± 74 Region French speaking Dutch speaking Marital status Married Single Divorced=separated Widowed Level of education Primary Secondary Technical University

Crude Mean (s.d.) 24.77 25.77 26.51 26.42 26.08 25.95

(3.20) (3.43) (3.36) (3.69) (3.77) (3.55)

25.89 (3.65) 25.98 (3.48) 25.96 25.76 25.87 26.19

(3.50) (3.88) (3.53) (4.04)

26.23 25.73 25.41 25.03

(3.63) (3.45) (3.30) (3.15)

Women Age-adjusted

P

Mean

P

Crude Mean (s.d.) 23.13 25.01 26.75 27.62 27.86 26.01

***a

NS

26.00 25.86

NS

25.96 25.94 25.96 25.71

***b

26.13 25.84 25.64 25.20

(3.51) (4.24) (4.59) (4.26) (4.64) (4.59)

NS

25.55 (4.60) 26.30 (4.56)

NS

26.00 24.73 24.81 27.30

(4.60) (4.00) (4.62) (4.45)

***

26.90 24.51 23.43 23.05

(4.62) (4.01) (3.47) (3.95)

Age-adjusted P

Mean

P

***

25.49 26.32

***

***d

26.20 24.50 25.36 25.46

***

***e

26.58 25.07 24.32 24.10

***

***c

NS: non-signi®cant; *P < 0.05; **P < 0.01; ***P < 0.0001 Multiple means comparisons: (P < 0.05) a Mean BMI differ signi®cantly two by two, except for groups 45 ± 54 y and 55 ± 64 y; b Mean BMI in primary differs signi®cantly from means in the three other groups; c Mean BMI differ signi®cantly two by two, except for groups 55 ± 64 y and 65 ± 74 y; d Mean BMI in single differs signi®cantly from means in the three other groups, and mean BMI of married differs signi®cantly from mean BMI of divorced=separated; e Mean BMI in primary and secondary differ signi®cantly from means in the three other groups.


4

Table 3 Prevalence of obesity (body mass index, BMI 30 kg=m2 (%)) and its odds ratios (OR, crude and age-adjusted) by age, region, marital status and level of education. Strati®cation by gender BMI 30kg=m2 Men Sociodemographic variables Age groups (y) 25 ± 34 35 ± 44 45 ± 54 55 ± 64 65 ± 74 Total 25 ± 74 Region French speaking Dutch speaking Marital status Married Single Divorced=separated Widowed Level of education Primary Secondary Technical University

Women

Crude %

Age-adjusted

OR (CI)

6.5 10.3 13.4 15.5 14.1 12.1

1.00 1.66 2.23 2.64 2.36

12.3 11.9

1.00 (ref.) 0.97 (0.82 ± 1.14)

11.8 13.6 12.7 15.7

1.00 1.18 1.08 1.39

14.0 10.6 9.3 4.8

1.00 0.73 0.63 0.31

OR (CI)

(ref.) (1.21 ± 2.27)*** (1.64 ± 3.02)*** (1.96 ± 3.56)*** (1.72 ± 3.25)***

Crude %

Age-adjusted

OR (CI)

OR (CI)

4.3 11.1 21.9 26.8 30.8 18.4

1.00 2.80 6.97 8.17 9.96

(ref.) (1.94 ± 4.11)*** (4.93 ± 10.05)*** (5.78 ± 11.77)*** (6.96 ± 14.50)***

1.00 (ref.) 0.99 (0.84 ± 1.16)

16.3 19.7

1.00 (ref.) 1.26 (1.09 ± 1.47)**

1.00 (ref.) 1.33 (1.14 ± 1.56)**

(ref.) (0.86 ± 1.60) (0.74 ± 1.55) (0.92 ± 2.10)

1.00 1.29 1.14 1.07

(ref.) (0.96 ± 1.74) (0.75 ± 1.74) (0.72 ± 1.60)

18.3 9.8 11.7 25.9

1.00 0.49 0.59 1.56

(ref.) (0.32 ± 0.75)*** (0.37 ± 0.94)* (1.25 ± 1.94)***

1.00 0.37 0.67 0.74

(ref.) (0.24 ± 0.56)*** (0.42 ± 1.05) (0.59 ± 0.94)*

(ref.) (0.59 ± 0.89)** (0.47 ± 0.84)** (0.18 ± 0.54)***

1.00 0.82 0.77 0.36

(ref.) (0.67 ± 1.00)* (0.57 ± 1.02) (0.21 ± 0.62)**

23.1 9.7 6.0 4.9

1.00 0.36 0.21 0.17

(ref.) (0.29 ± 0.44)*** (0.14 ± 0.32)*** (0.05 ± 0.46)***

1.00 0.49 0.33 0.28

(ref.) (0.40 ± 0.61)*** (0.22 ± 0.50)*** (0.10 ± 0.78)*

CI 95% con®dence intervals; ref reference group *P < 0.05; **P < 0.01; ***P < 0.0001.

two-by-two signi®cant differences in average BMI, except between women with technical and university degrees. Furthermore, compared to the lowest educated women, the proportion of obese was signi®cantly lower in each higher education group; in men the clearest difference in proportion of obese was found between primary and university levels. Dietary factors and obesity (Table 4 and Table 5)

Nutrient intakes. Total fat intake was signi®cantly positively associated with BMI and prevalence of obesity, but in men only. After adjustment for age, mean BMI remained signi®cantly higher by 0.55 kg=m2 among men in the highest intake quartile compared to those in the lowest quartile. This corresponds to a difference of 2 kg for a 180 cm tall man. For prevalence of obesity, bivariate analysis showed the proportion of the obese to be 1.32 higher in quartile 4 compared to quartile 1 (P < 0.05). Total carbohydrate and total sugar intakes were signi®cantly and inversely related to BMI (in its continuous form, as well as in its categorical form) in both sexes, but the association was more pronounced with total sugar intake. When considering the fat to sugar ratio, we found, in men and in women, that the subjects with a diet that was predominantly composed of fat (fat intake more than four-fold higher than the sugar intake in quartile 4) had signi®cantly higher BMI values than those with a low fat to sugar ratio (fat intake less than two-fold higher than sugar intake). The same association was observed with the

prevalence of obesity. In both cases the association persisted after adjustment for age. Physical activity level (PAL). A very clear inverse (and highly signi®cant) gradient was observed between PAL and BMI for both sexes. Mean BMI differed signi®cantly two by two in all quartiles. Regarding prevalence of obesity, after adjustment for age, the proportion of obese men in the highest PAL quartile was ®ve-fold lower than that in the lowest quartile. In women, the percentage of the obese was lowered by a factor 5.6 between these quartiles.

Discussion Although the base-line data from the BIRNH study are not very recent, the present ®ndings are of national relevance. They give an unique image (because they are based on a national representative general population sample) of the magnitude of the problem of obesity in Belgium. They indicate that a substantial proportion of Belgian men (12.1%) and Belgian women (18.4%) are obese. Furthermore, the percentage of overweight subjects (BMI: 25 ± 29.99) is dramatically high: 46.5% in men and 35.2% in women. In neighbouring countries, like Germany50 and England,3 prevalence of obesity in the late 1980s was similar (17.2% and 13%, respectively, in men and 19.3% and 15%, respectively, in women). In the Netherlands

Obesity in Belgium MC Stam-Moraga et al Table 4 Mean body mass index (BMI, crude and age-adjusted) by total fat, total carbohydrate and total sugar intake, fat to sugar ratio and physical activity level (PAL, mean (s.d.)). Strati®cation by gender BMI (kg=m2) Men Crude Nutrient intake Total fat (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Total carbohydrates (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Total sugar (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Fat : sugar ratio Quartile 1 Quartile 2 Quartile 3 Quartile 4 PAL Quartile 1 Quartile 2 Quartile 3 Quartile 4

Mean (s.d.) 25.80 25.77 25.93 26.32

(3.52) (3.50) (3.45) (3.68)

26.44 26.05 25.77 25.57

(3.73) (3.60) (3.45) (3.33)

26.32 25.97 25.86 25.67

(3.71) (3.54) (3.54) (3.34)

25.68 25.87 25.91 26.37

(3.38) (3.51) (3.60) (3.65)

27.50 26.31 25.40 24.61

(3.83) (3.25) (3.24) (3.14)

Women Age-adjusted

P

Mean

***a

25.79 25.76 25.95 26.34

***b

26.50 26.06 25.75 25.53

***c

26.50 25.97 25.89 25.70

***a

25.70 25.87 25.90 26.36

***e

27.44 26.30 25.41 24.64

P

Crude Mean (s.d.)

***

25.93 26.09 26.02 26.07

(4.45) (4.59) (4.67) (4.63)

***

26.10 26.22 26.02 25.77

(4.63) (4.81) (4.39) (4.49)

***

26.53 26.04 25.92 25.62

(4.85) (4.50) (4.55) (4.38)

***

25.68 25.87 26.11 26.46

(4.35) (4.48) (4.59) (4.87)

***

28.18 26.44 25.33 24.16

(5.08) (4.44) (4.04) (3.67)

Age-adjusted P

Mean

P

NS

25.79 26.02 26.07 26.24

NS

NS

26.31 26.24 25.95 25.63

***

***b

26.53 26.03 25.88 25.68

***

***d

25.68 25.88 26.13 26.50

***

***e

28.01 26.35 25.31 24.44

***

NS: non-signi®cant; *P < 0.05; **P < 0.01; ***P < 0.0001; %EI percent of total energy intake Multiple means comparisons: P < 0.05 Mean BMI in quartile 4 differs signi®cantly from means in the three other quartiles; b Mean BMI in quartile 1 differs signi®cantly from means in the three other quartiles; c Mean BMI in quartile 1 and 2 differ signi®cantly from means in the three other quartiles; d Mean BMI in quartile 4 differs signi®cantly from means in the quartiles 1 and 2, and mean BMI in quartile 3 differs signi®cantly from mean in quartile 1; e Mean BMI in all quartiles differ signi®cantly two by two. a

between 1987 and 1991 the proportion of obese men and women aged 20 ± 59 y, was considerably lower, 7% and 9%, respectively.51 Sociodemographic factors and obesity

The increase of obesity with age differed by gender. In men, the proportion of the obese rises to 45 ± 54 y, then declines. In women, the prevalence of obesity increases continuously to 65 ± 74 y. This discrepancy between gender is in accordance with other crosssectional studies, reporting women to be generally more obese than men, especially after the age of 50 y.3 ± 5 The effect of marital status on BMI and prevalence of obesity also differed by gender. As no effect was observed for men, married women (after adjustment for age) were of a higher weight than single and widowed women. Some former studies did report married people to have a higher BMI than unmarried persons,8 but others did not.6,7 The direction of this association also remains to be clari®ed. Thus Jeffery et al,9 who analysed the nature of social in¯uence on

body weight, observed in both sexes, that persons classi®ed in the lowest and highest quintiles of the BMI distribution, were less likely to be married than those in the middle. In this study, a highly signi®cant inverse gradient (although more pronounced in women than in men) was observed between level of education and obesity (either measured by BMI or by the proportion of obese subjects). Level of education being an alternate measure of socioeconomic status (SES), our results support Sobal and Stunkards10 conclusions in their review of 144 studies examining the relationship between SES and obesity. Although SES measures differed across studies, they concluded consistently and straightforwardly that in the developed societies, higher social class was strongly associated with a lower prevalence of obesity, especially among women. Among men, the relationship was weaker and variable in direction. Although this relation is well established transversally, attempting to explain it is more complex. Stunkard and Sùrensen11 give a good overview of the complexity of the obesity-SES relationship. In

5


6

Table 5 Prevalence of obesity (body mass index, BMI 30 kg=m2 (%)) and its odds ratios (OR, crude and age-adjusted) by total fat, total carbohydrate and total sugar intake, fat to sugar ratio and physical activity level (PAL). Strati®cation by gender BMI 30 kg=m2 Men Crude Nutrient intake Total fat (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Total carbohydrate (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Total sugar (%EI) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Fat : sugar ratio Quartile 1 Quartile 2 Quartile 3 Quartile 4 PAL Quartile 1 Quartile 2 Quartile 3 Quartile 4

(%)

OR (CI)

Women Age-adjusted OR (CI)

Crude %

OR (CI)

Age-adjusted OR (CI)

12.2 10.3 11.0 15.2

1.00 0.83 0.89 1.30

(ref.) (0.65 ± 1.05) (0.71 ± 1.13) (1.04 ± 1.62)*

1.00 0.83 0.90 1.32

(ref.) (0.65 ± 1.05) (0.72 ± 1.14) (1.07 ± 1.64)*

17.7 19.1 18.3 18.5

1.00 1.10 1.04 1.06

(ref.) (0.90 ± 1.35) (0.85 ± 1.28) (0.86 ± 1.30)

1.00 1.14 1.13 1.21

(ref.) (0.94 ± 1.39) (0.92 ± 1.41) (0.97 ± 1.50)

15.8 12.3 11.4 9.1

1.00 0.75 0.68 0.53

(ref.) (0.60 ± 0.93)** (0.55 ± 0.85)*** (0.42 ± 0.67)***

1.00 0.72 0.74 0.59

(ref.) (0.58 ± 0.90)** (0.59 ± 0.91)*** (0.46 ± 0.74)***

19.2 19.3 18.0 17.3

1.00 1.01 0.95 0.88

(ref.) (0.83 ± 1.23) (0.75 ± 1.28) (0.72 ± 1.08)

1.00 0.93 0.82 0.75

(ref.) (0.77 ± 1.13) (0.67 ± 1.00) (0.60 ± 0.93)*

15.6 11.8 11.7 9.6

1.00 0.72 0.72 0.57

(ref.) (0.58 ± 0.90)** (0.57 ± 0.89)** (0.45 ± 0.72)***

1.00 0.74 0.74 0.59

(ref.) (0.60 ± 0.92)** (0.59 ± 0.91)** (0.46 ± 0.74)***

21.2 18.7 17.0 16.9

1.00 0.85 0.76 0.76

(ref.) (0.70 ± 1.04) (0.62 ± 0.93)** (0.62 ± 0.93)**

1.00 0.84 0.73 0.74

(ref.) (0.69 ± 1.02) (0.60 ± 0.88)* (0.60 ± 0.93)*

10.2 11.3 11.9 15.3

1.00 1.12 1.19 1.59

(ref.) (0.88 ± 1.43) (0.94 ± 1.51) (1.26 ± 2.00)***

1.00 1.12 1.18 1.56

(ref.) (0.88 ± 1.41) (0.93 ± 1.48) (1.25 ± 1.93)***

16.1 17.4 19.6 20.7

1.00 1.10 1.27 1.37

(ref.) (0.89 ± 1.36) (1.03 ± 1.56)* (1.11 ± 1.68)**

1.00 1.09 1.32 1.45

(ref.) (0.87 ± 1.34) (1.07 ± 1.64)* (1.17 ± 1.80)**

22.9 12.6 8.0 5.2

1.00 0.49 0.29 0.18

(ref.) (0.40 ± 0.60)*** (0.23 ± 0.37)*** (0.14 ± 0.24)***

1.00 0.50 0.31 0.20

(ref.) (0.41 ± 0.61)*** (0.24 ± 0.38)*** (0.15 ± 0.26)***

33.5 19.9 12.6 7.8

1.00 0.49 0.29 0.17

(ref.) (0.41 ± 0.59)*** (0.23 ± 0.35)*** (0.13 ± 0.21)***

1.00 0.48 0.28 0.18

(ref.) (0.40 ± 0.58)*** (0.23 ± 0.34)*** (0.14 ± 0.23)***

CI 95% con®dence intervals; %EI percent of total energy intake; ref reference group *P < 0.05; **P < 0.01; ***P < 0.0001.

summary, they proposed at least three possibilities, with evidence supporting each of them. Firstly, obesity might in¯uence SES. Indeed longitudinal studies reported, during follow-up, that overweight women married less often, had lower income and completed fewer years of school. Similar but weaker trends were found among the men.12 Secondly, SES might in¯uence obesity. This hypothesis is not only cross-sectionally supported by studies12 ± 14 showing that parental education was signi®cantly lower among the overweight than among the nonoverweight subjects, but also longitudinally by studies reporting that childhood SES predicts the development of obesity in adult life.15,16 The strongest evidence that the SES of the parents in¯uences the BMI of their offspring is given by a Danish adoptive study52 where, after controlling for parental BMI, social class of adoptive parents was inversely associated with BMI of adoptees, while BMI of the adoptive parents and BMI of the adoptees were not correlated. The third possibility, that one or more common factors in¯uence both obesity and SES, considerably complicates the attribution of causality. For example, Stephens et al17 found a positive relation between SES and physical activity. Others18 reported that persons with a high education smoked less and adopted healthier dietary patterns. Finally, in women, a strong association exists between social class and the desire to be thin.19

Dietary factors and obesity

Nutrient intakes. In comparison with the national nutritional recommendations recently published for Belgium,53 the present study population has a rather higher proportional intake of fat. With this respect, Kornitzer and Dramaix54 observed in the same population, that only 8.4% of men and 7% of women were situated under the recommended upper limit of 30% for total fat intake. Furthermore, proportional intake of total carbohydrates is clearly lower than recommended (that is, between 55% and 75% of total energy intake) and intake of total sugar is slightly higher than the recommended upper limit of 10%. The positive associations we found between obesity and total fat intake were more pronounced in men than in women, and stronger with mean BMI than with the prevalence of obesity. After correcting for age, the increase of both obesity indexes was essentially observed between lowest and highest quartiles, and was only signi®cant among men. In the interpretation of the results, it would be erroneous to conclude that total fat intake does not affect the occurrence of obesity. Most cross-sectional studies,20,21,31 ± 33 but not all,34 report the same relationship. Moreover, the in¯uence of fat intake on weight gain has also, but not always,34 been con®rmed prospectively.35,36 The weakness of the association in our study could result


from the method used to assess the dietary intakes. Beaton et al55 indicate, in this review of the sources of error in short-term recalls, that these could attenuate the nutrient-disease relation, if present. While interindividual variations of dietary intake can be reduced by expressing intakes in proportion to energy, a high intraindividual variability is affected to a much lesser degree, leading to a misclassi®cation of the subjects according to their habitual intakes. The physiological arguments for the role of dietary fat as a weight gain promotor are many: 1) dietary fat, due to its well-known palatability and energy density, may promote excessive energy intake56; 2) Dietary fat has only a weak effect on satiety, which can contribute to passive overconsumption of fat57 and 3) In terms of nutrient balance equations (oxidation in response to excessive intake) unlikely to the non-fat nutrients, body fat stores are large and fat intake has no effect on its oxidation.58 ± 61 Therefore, energy balance is essentially equivalent to fat balance, and since fat intake does not stimulate fat oxidation, clearly there is a potential for fat imbalance leading to obesity in subjects consuming a high-fat diet. In this study, total carbohydrate intake and especially total sugar intake was inversely associated with obesity in both sexes. In other surveys, a reverse association was reported especially with the intake of extrinsic sugars (not naturally occurring sugars, such as lactose milk).32,33,36,37 Physiologically, this inverse relation could result from the fact, as mentioned previously, that excess carbohydrate energy, unlike fat energy, produces rapid increases in carbohydrate oxidation and energy expenditure in a way not shown by fat.58,60,61 Epidemiological data report that the apparent protective effect of sugars against obesity may result from a reciprocal linkage in diet between the percent of energy derived from sugars and that from fat.32 Applying this concept, and in agreement with BoltonSmith and Woodward,37 we found that average BMI and prevalence of obesity were substantially higher in the subjects classi®ed in the highest quartile of fat to sugar ratio. Physical activity level (PAL). We evaluated the participants' physical activity by the PAL, dividing their total energy intake by an estimation of their BMR. This approach is inherently reasonable, provided that energy balance is maintained and that reported calorie intakes effectively re¯ects the real calorie amounts ingested by the subjects. With regard to energy balance, assuming that the dynamic phase of obesity is a relatively short period, Baeke et al27 suggested, as a consequence, that in cross-sectional studies most obese people would be in energy balance. Concerning the second assumption, some studies have shown that obese people systematically underreport the quantity of food they eat.24,25 In our study,

as reported by others,20 ± 23 energy intake was inversely associated with the prevalence of obesity, but also to the mean BMI (data not shown). As a possible explanation of this negative correlation, systematic underestimation of energy intake cannot be excluded. However, as suggested by Keen et al,26 underreporting of energy intake occurs especially in clinically obese patients as a behavioural response to obesity. We observed an inverse relationship between energy intake and BMI across the whole range of the index, so this trend was not only con®ned to the manifestly obese subjects. When analysing BMI variations and prevalence of obesity by PAL, we observed a very strong inverse gradient in both sexes, with a ®ve-fold higher proportion of obese subjects in the lowest quartile compared to the highest quartile. These results strongly suggest that body adiposity can be determined by the level of physical activity. In agreement with our ®ndings, Slattery et al 28 showed that among 2346 men, caloric intake (measured by a single 24 h recall) was directly correlated to leisure time and occupational physical activity, but was inversely associated with adiposity (measured by the BMI and the sum of the skinfold thickness). Other authors, like Owens et al29 observed prospectively that women with a high PAL at baseline gained less weight during the three-year followup than those with lower activity levels; the least weight gains being observed among those having increased their PAL during follow-up.

Conclusion Attempting to draw a central strategy for the prevention and treatment of obesity at the Belgian general population level, our results reinforced the idea that special attention should be given to regular practice of a sustained physical activity, more than solely a reduction in energy intake, as obese people already have lower intakes compared to their non-obese congeners. In addition, our ®ndings that low fat, low prevalent obesity, diet is relatively high in sugars, may translate dif®culties of the Belgian population to exchange fat energy for more healthier foods (vegetables, fruits and cereals) instead of foods containing large amounts of sugar. Health messages have also to be positively discriminated towards the less educated social layers. References

1 World Health Organisation. Obesity Epidemic puts millions at risk from related diseases. Press Release WHO=46 (12 June 1997). WHO: Geneva, 1997. 2 Ravussin E, Swinburn BA. Pathophysiology of obesity. Lancet 1992; 340: 404 ± 408. 3 Department of Health. The health of the nation: one year on . . . a report on the progress of the health of the nation. HMSO: London 1993.

7


8

4 Kuskowska-Wolk A, BergstroÈm R. Trends in BMI and prevalence of obesity in Swedish women 1980 ± 89. J Epidemiol Commun Health 1993; 47: 195 ± 199. 5 Cornari-Huntley JC, Harris JB, Everett DF, Albanes D, Micozzi MS, Miles TP, Feldman JJ. An overview of body weight of older persons, including the impact on mortality. The national health and nutrition examination survey I ± epidemiologic follow-up study. J Clin Epidemiol 1991; 44: 743 ± 753. 6 Rissanen A, HelioÈvaara M, Kenkt P, Reunanen A, Aromaa A. Determinants of weight gain and overweight in adult Finns. Eur J Clin Nutr 1991; 45: 419 ± 430. 7 Merkus MP, Mathus-Vliegen LMM, Broekhoft C, Heijnen AMP. Extreme obesity: sociodemographic, familial and behavioural correlates in The Netherlands. J Epidemiol Commun Health 1995; 49: 22 ± 27. 8 Croft JB, Strogarlz DS, James SA, Keenan NL, Ammerman AS, Malarcher AM, Haines PS. Socioeconomic and behavioural correlates of body mass index in black adults: the Pitt county study. Am J Public Health 1992; 82: 821 ± 826. 9 Jeffery FW, Forster JL, Folsom AR, Lueptien RV, Jacobs JR, Blackburn H. The relationship between social status and body mass index in the Minnesota Health Program. Int J Obes 1989; 13: 59 ± 67. 10 Sobal J, Stunkard AJ. Socioeconomic status and obesity: a review of the literature. Psychol Bull 1989; 105: 260 ± 275. 11 Stunkard AJ, Sùrensen TIA. Obesity and socioeconomic status ± a complex relation. N Engl J Med 1993; 329: 1036 ± 1037. 12 Gortmaker SL, Must A, Perrin JM, Sobol AM, Dietz WH. Social and economic consequences of overweight in adolescence and young adulthood. N Engl J Med 1993; 329: 1008 ± 1012. 13 Sonne-Holm S, Sùrensen TIA. Prospective study of attainment of social class of severely obese subjects in relation to parental social class, intelligence and education. BMJ 1986; 292: 586 ± 589. 14 Goldblatt PB, Moore ME, Stinkard AJ. Social factors in obesity. JAMA 1965; 192: 1039 ± 1044. 15 Braddon FE, Rodgers B, Wadsworth ME, Davies JM. Onset of obesity in a 36 year birth cohort study. BMJ 1986; 293: 299 ± 303. 16 Power C, Moynihan C. Social class and changes in weight-forheight between childhood and early adulthood. Int J Obes 1988; 12: 445 ± 543. 17 Stephens T, Jacobs DR, White CC. A descriptive epidemiology of leisure time physical activity. Public Health Rep 1985; 100: 147 ± 158. 18 Jacoben BK, Thelle DS. Risk factors for coronary heart disease and level of education: the Tromso Heart Study. Am J Epidemiol 1988; 127: 923 ± 932. 19 Dornbusch SM, Carlsmith JM, Duncan PD, Gross RT, Martin JA, Ritter PL, Siegel-Garlick B. Sexual maturation, social class and the desire to be thin among adolescent females. J Dev Behav Pediatr 1984; 5: 308 ± 314. 20 Tucker LA, Kano MJ. Dietary fat and body fat: a multivariate study of 205 adult females. Am J Clin Nutr 1992; 56: 616 ± 622. 21 Miller WC, Lindeman AK, Wallace J, Niederpruem M. Diet composition, energy intake and exercise in relation to body fat in men and women. Am J Clin Nutr 1990; 52: 426 ± 430. 22 Romieu I, Willett WC, Stampfer MJ, Colditz GA, Sampson L, Rosner B, Hennekens CH, Speizer FE. Energy intake and other determinants of relative weight. Am J Clin Nutr 1988; 47: 406 ± 412. 23 Keen H, Thomas BJ, Jarret RJ, Fuller JH. Nutrient intake, adiposity and diabetes. BMJ 1979; 1: 655 ± 658. 24 Forbes GB. Diet and exercise in obese subjects: self-report versus controlled measurements. Nutr Rev 1993; 51: 296 ± 300.

25 Prentice AM, Black AE, Coward WA, Davies ML, Goldberg GR, Murgatroyd PR, Ashford J, Sawyer M, Whitehead RG. High levels of energy expenditure in obese women. BMJ 1986; 292: 983 ± 987. 26 Keen H, Thomas BJ, Jarret JR, Fuller JH. Nutrient intake, adiposity and diabetes. BMJ 1979; 1: 655 ± 658. 27 Baeke JAH, Van Staveren WA, Burema J. Food consumption, habitual physical activity and body fatness in young Dutch adults. Am J Clin Nutr 1983; 37: 278 ± 286. 28 Slattery ML, Jacobs DR, Nichaman MZ. An assessment of caloric intake as an indicator of physical activity. Prev Med 1989; 18: 444 ± 451. 29 Owens JF, Matthews KA, Wing RR, Kuller LH. Can physical activity mitigate the effects of aging in middle-aged women? Circulation 1992; 85: 1265 ± 1270. 30 Williamson DF, Madans J, Anda RF, Kleinman JC, Kahn HS, Byers T. Recreational physical activity and ten-year weight change in an US national cohort. Int J Obes 1993; 17: 279 ± 286. 31 Lissner L, Heitmann BL. Dietary fat and obesity: evidence from epidemiology. Eur J Clin Nutr 1995; 49: 79 ± 90. 32 Hill JO, Prentice AM. Sugar and body weight regulation. Am J Clin Nutr 1995; 62 (Suppl 1): S264 ± S274. 33 Bolton-Smith C, Woodward M. The prevalence of overweight and obesity in different fat and sugar consumption groups. Proc Nutr Soc 1993; 52: 383A (Abstr). 34 Kant AK, Graubard BI, Schatztin A, Ballard-Barbasch R. Proprotion of energy intake from fat and subsequent weight change in the NHANES I Epidemiologic Follow-up Study. Am J Clin Nutr 1995; 61: 11 ± 17. 35 Colditz GA, Willett WC, Stampfer MJ, London SJ, Segal MR, Speizer FE. Patterns of weight change and their relation to diet in a cohort of healthy women. Am J Clin Nutr 1990; 51: 1100 ± 1105. 36 Klesges RC, Klesges LM, Haddock CK, Eck LM. A longitudinal analysis of impact of dietary intake and physical activity and weight change in adults. Am J Clin Nutr 1992; 55: 818 ± 822. 37 Bolton-Smith C, Woodward M. Dietary composition and fat to sugar ratios in relation to obesity. Int J Obes 1994; 18: 820 ± 828. 38 Gibson SA. Consumption and sources of sugars in the diets of British school children: are high sugar diets nutritionally inferior? J Hum Nutr Diet 1993; 6: 355 ± 371. 39 Interuniversity Research group-BIRNH. Nutriton and health: an Interuniversity study. Regional differences in dietary habits, coronary risk factors and mortality rates in Belgium. 1. Design and methodology. Acta Cardiol 1984; 39: 285 ± 292. 40 Kornitzer M, Dramaix M. The Belgian Interuniversity Research on Nutrition and Health (BIRNH): general introduction for the BIRNH Study Group. Acta Cardiol 1989; 2: 89 ± 99. 41 Kornitzer M, Bara L. For the BIRNH Study Group. Clinical and anthropometric data. Blood chemistry and nutritional patterns in the Belgian population according to age and sex. Acta Cardiol 1989; 44: 101 ± 144. 42 Kornitzer M, Bara L. (For the BIRNH Study Group). Differences between North and South in coronary risk factors, food habits and mortality in Belgium. Acta Cardiol 1989; 44: 145 ± 155. 43 Voorlichtingsbureau voor de Voeding. Nederlandse voedingsmiddelentabel. Nederland: s'Gravenhage, 1979, druk 32. 44 Paul AA, Southgate DAT. The composition of foods, 4th Edition, London: MRC report, 1980, 297. 45 Sasaki S, Kesteloot H. Value of Food and Agriculture Organization data on food balance sheets as a data source for dietary fat intake in epidemiologic studies. Am J Clin Nutr 1992; 56: 716 ± 723.


46 James WPT, Scho®eld EC. Human energy requirements. A manual for planners and nutritionists. Oxford University Press: New York, 1990. 47 Kleinbaum DG, Kupper LL, Muller KE (eds). Applied Regression Analysis and Other Multivariable Methods (2nd edn). PWS-KENT Publishing Company: Boston, 1988. 48 MarijaÈ J. Statistical package for the social sciences. Norusis=SPSS Inc.: 1990. 49 Garrow JS. Treat obesity seriously. Churchill Livingstone: Edinburgh, 1981, p 3. 50 Hoffmeister H, Mensink GBM, Stozenberg H. National trends in risk factors for cardiovascular disease in Germany. Prev Med 1994; 23: 197 ± 205. 51 Seidell JC, Verschuren WMM, Krombout D. Prevalence and trends of obesity in The Netherlands 1987 ± 1991. Int J Obes 1995; 19: 924 ± 927. 52 Teasdale TW, Sùrensen TIA, Stunkard AJ. Genetic and early environmental components in sociodemographic in¯uences on adult body fatness. BMJ 1990; 300: 1615 ± 1618. 53 Conseil National de la Nutrition. Recommandations nutritionnelles pour la Belgique. MinisteÁre des Affaires Sociales, de la SanteÂ Publique et de l'Environnement: Brussels 1996. 54 Kornitzer M, Dramaix M. The BIRNH Study Group. Prevention of coronary heart disease: how far is the Belgian population from the recommended nutritional goals and ideal serum cholesterol levels? Eur Heart J 1988; 9: 1048 ± 1057.

55 Beaton GH, Milner J, Corey P, McGuire V, Cousins M, Stewart E, de Ramas M, Hewitt D, Grambsch PV, Kassim N, Little JA. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr 1979; 32: 2456 ± 2459. 56 Rolls BJ, Shide DJ. The in¯uence of dietary fat on food intake and body weight. Nutr Rev 1992; 50: 283 ± 290. 57 Blundell JE, Burley VJ, Cotton JR, Lawton L. Dietary fat and the control of energy intake: evaluating the effect of fat on meal size and post-meal satiety. Am J Clin Nutr 1993; 57 (Suppl 5): S772 ± S777. 58 Abbott WG, Howard BV, Christin L, Freymond D, Lillioja S, Boyce VL, Anderson TE, Bogardus C, Ravussin E. Short-term energy balance: relationship with protein, carbohydrate and fat balance. Am J Physiol 1988; 225: E332 ± E337. 59 Flatt JP, Ravussin E, Acheson KJ, JeÂquier E. Effects of dietary fat on postprandial substract oxidation and carbohydrate and fat balance. J Clin Invest 1985; 76: 1019 ± 1024. 60 Schutz Y, Flatt JP, JeÂquier E. Failure of dietary fat intake to promote fat oxidation: a factor favoring the development of obesity .Am J Clin Nutr 1989; 50: 7 ± 14. 61 Flatt JP. The difference in the storage capacities for carbohydrate and for fat, and its implications in the regulation of body weight. Ann NY Acad Sci 1987; 499: 104 ± 123.

9