Factors associated with overweight and central body fat in the city of Rio de Janeiro

Factors associated with overweight and central body fat in the city of Rio de Janeiro

Public Health (2001) 115, 236±242 ß R.I.P.H.H. 2001 www.nature.com/ph Factors associated with overweight and central body fat in the city of Rio de J...

81KB Sizes 1 Downloads 60 Views

Public Health (2001) 115, 236±242 ß R.I.P.H.H. 2001 www.nature.com/ph

Factors associated with overweight and central body fat in the city of Rio de Janeiro: results of a two-stage random sampling survey VM Ramos de Marins1, RMR Varnier Almeida2*, RA Pereira3 and MBA Barros4 1

Nutrition Department, Fluminense Federal University, NiteroÂi, RJ, Brazil; 2Program of Biomedical Engineering=COPPE= Federal University of Rio de Janeiro, City of Rio de Janeiro, RJ, Brazil; 3Nutrition Institute, Federal University of Rio de Janeiro, City of Rio de Janeiro, RJ, Brazil; and 4Medical School, University of Campinas UNICAMP, Campinas SP, Brazil The purpose of the survey was to investigate the association of overweight (body mass index, BMI) and central body fat distribution (waist=hip girth ratio WHR) with socio-economic, demographic, lifestyle and dietary variables in the adult population of Rio de Janeiro City, Brazil, 1995 ± 1996. A two-stage random sample population-based survey was performed, with 1455 males and 1906 females above 20 y old resident in Rio de Janeiro City, Brazil. Data were obtained by direct interview and physical examination of the subjects. The intake of selected nutrients (fat, saturated fat, cholesterol) and energy was obtained from a semi-quantitative food frequency questionnaire (FFQ). Variables with at least a marginal univariate association with the dependent variables (BMI and WHR) were selected as predictors in two logistic regression models, and variables statistically signi®cant (P < 0.05) were retained in them. Overweight prevalence was 44.9%, and 39.2% of the subjects had excessive central body fat distribution (elevated WHR). The proportions of subjects with an excessive intake of fat, saturated fat and cholesterol were respectively 31%, 42% and 47%. For the BMI model, the variables retained were age (OR ˆ 1.5, 95% CI ˆ 1.3 ± 1.7), schooling (OR ˆ 1.7, 1.4 ± 1.9) and smoking (OR ˆ 0.8; 0.7 ± 0.9); and for WHR, age (OR ˆ 1.8, 1.5 ± 2.1), schooling (OR ˆ 2.2, 1.9 ± 2.6), occupation (OR ˆ 1.8, 1.1 ± 2.0) and gender (OR ˆ 3.9, 3.2 ± 4.7). Obesity and excessive central body fat are highly prevalent health problems in the studied population. As suggested by the identi®ed risk factors, they should be urgently addressed through health nutrition education and physical activity programs; particularly those directed to the middle aged and female groups. Public Health (2001) 115, 236±242. Keywords: overweight; body mass index; waist-to-hip ratio; food intake

Introduction In the last decades, public health efforts in industrialised countries have been mostly concerned with actions such as the prevention of chronic diseases, for instance, through programs for weight reduction and the promotion of better eating habits.1 However, results for overweight reduction in these countries have been discouraging. For instance, in the United States, the National Health and Nutritional Examination Survey (NHANES III), 1988 ± 19912 showed that 33% of adults could then be classi®ed as overweight. More recent studies indicate that this proportion is still increasing, approaching 50% of the population.3,4 This increase has been partially explained by the consequences of modern civilisation, such as an increase in the availability of energy-dense foods, decreased levels of ®bre consumption and sedentary lifestyles.3

*Correspondence: RMR Varnier Almeida, PEB=COPPE=Federal University of Rio de Janeiro, Caixa Postal 68510, Cidade UniversitaÂria Rio de Janeiro, RJ, Brazil, 21945-970. Accepted 29 January 2001

In developing nations, although severe undernutrition problems persist in many areas, the problem of excessive food intake has also been arising.5 In Brazil, a comparison of two national surveys in the last two decades, the National Study on Family Expenditures (ENDEF)6 and the National Survey on Health and Nutrition (PNSN)7 reveals that overweight has been increasing in the country, and that it has a complex relationship with socio-economic factors. For instance, in women, the detected relationship between income and overweight appeared non-linear, increasing at the lower income strata and showing a decrease on the highest income quartile.8 It is well known that overweight is not just an aesthetic problem, but a health risk strongly associated with mortality and important chronic diseases, particularly hypertension, alteration of serum lipids and plasma lipoproteins, type-2 diabetes mellitus and other insulin metabolic alterations, ostheoarthritis, some cancers and gall bladder diseases.4 These concerns prompted the present investigation on overweight prevalence and associated factors in the City of Rio de Janeiro, Brazil. More speci®cally, this study investigates overweight and central body fat prevalence and their associated socio-economic, demographic, life style

Factors associated with overweight and central body fat VM Ramos de Marins et al

and food dietary variables in the adult (above 20-y-old) population of Rio de Janeiro City, Rio de Janeiro State, Brazil, 1995 ± 1996.

Materials Rio de Janeiro is a city with approximately 5 000 000 inhabitants, located in the southern region of the country. The study used data from a random sample populationbased survey. Data were obtained by direct interview and physical examination of subjects selected by a two-stage sampling process. In the ®rst stage, 60 census sectors were randomly selected, with probability proportional to their population size, and, in the second, 34 residences were sampled from inside each of the primary sampling units, a total of 2040 residences. Data quality was veri®ed by means of telephone interviews that randomly re-checked answers on 5% of the questionnaires.

237

Demographic variables These were gender and age, calculated from birth to interview date, in years. Age was divided into below and above 45-y-old. Socio-economic variables For residence area, subjects were classi®ed either as slum dwellers or non-slum dwellers, according to their residence location. Income per capita was calculated as the total income of a family, divided by the number of persons in the family. This variable was dichotomised using as cut off point its sample median (approximately US$ 200). Schooling was measured as number of years of schooling. It was dichotomised as `lower schooling' (no more than eight years of schooling) and `higher schooling' (more than eight years). Lifestyle variables

Anthropometric measurements and nutritional assessment Anthropometric measurement procedures were standardised during a training period consisting of eight sessions. Data collection teams were formed by an anthropometrist and an interviewer. Interviews took place inside the subjects' residences. Weight measurements were made with digital scales with 100 g variation and 150 kg maximum capacity. Height was measured in centimetres with a metric tape ®xed to a wall with no baseboard, and subjects stood in their erect position with bare feet and arms along the body, according to the standardised procedures proposed by Lohman.9 Two height measurements were performed, with the mean height being recorded unless a difference of 1 cm or more between them was identi®ed. In this case, both measurements were repeated. Waist and hip girth were measured with subjects wearing light clothing. Waist girth was measured at the minimum circumference between the iliac crest and the rib cage. Hip girth was measured at the maximum width over the greater trochanters.9

Smoking was dichotomised into `smokers' and `non-smokers'. Type and quantity of smoking were not considered, and ex-smokers were grouped into the `non-smokers' category. Alcohol consumption was classi®ed as `lower consumption' (abstinence or consumption of less that 14 glasses per week) and `higher consumption' (otherwise). Physical activity Physical activity took into account both leisure time and work-related physical activities during the preceding month. Two physical activity variables were de®ned. The ®rst was physical activity other than locomotion to work, (walking, jogging, sports) dichotomised as `yes' or `no'. The second (occupation) tried to classify subjects in terms of their job-related energy expenditure. Occupations were classi®ed either as `low energy expenditure' (eg teachers, bank clerks, cashiers, housewives and other white-collar jobs) or as `high energy expenditure' (eg janitors, repair men, car mechanics). Dietary variables

Overweight and central fat risk variables The body mass index (BMI), de®ned as weight (kg) divided by the height square (m2) was used to characterise overweight,10 which was de®ned by the cut off point of 25 kg=m2. Central body fat distribution was evaluated by the waist-to-hip girth ratio (WHR), with cut off points for central body fat excess of 0.95 (males) and 0.80 (females).11

A semi-quantitative food frequency questionnaire (FFQ) was used for the assessment of food intake. This questionnaire had been previously validated by four non-consecutive 24-h food consumption recalls in a sample of 88 adult employees of a city university. The FFQ presented an acceptable correlation with the mean nutrient intake obtained from the four 24-h recall data. Information on food ingestion included eighty food items suggested by ENDEF, that are frequently used in Public Health

Factors associated with overweight and central body fat VM Ramos de Marins et al

238

the Brazilian diet. The daily food ingestion was converted into energy intake, fat consumption, saturated fat consumption and cholesterol consumption by a speci®cally developed computer application, taking into account the amount and frequency of ingested items and food composition according to the US Department of Agriculture database.12 The dietary variables were dichotomised as low=high energy intake (lowest through 2500 Kcal=day or above); low=high total fat consumption (up to 30% of total diet calories or more than 30%); low=high saturated fat consumption (up to 10% of total diet energy or more than 10%)

and low=high cholesterol consumption (up to 300 mg daily or more than 300 mg).13 Data analysis All variables were categorised into `0' and `1', with `0' corresponding to the reference category. At ®rst, the association between BMI, WHR and the independent variables was established through the estimation of their odds ratio (OR), 95% con®dence intervals (95% CI) and P-

Table 1 Sample distribution of variables used, Rio de Janeiro City, 1995 ± 6

Public Health

Variables

Total

Gender Female Male Body mass index > 25 18.5±25 Waist=hip girth ratio Risk No risk Age  45 y < 45 y Schooling Lower ( 8 y) Higher ( >8 y) Residence area Slum Non-slum Income per capita (US$) < 200.00  200.00 Smoking Smoker Non-smoker Alcohol consumption  14 weekly doses < 14 weekly doses Activity other than locomotion to work No Yes Occupation Low energy expenditure High energy expenditure Energy intake (Kcal) > 2500  2500 Fat consumption > 30% of total diet  30% of total diet Saturated fat consumption > 10% of diet total energy  10% of diet total energy Cholesterol consumption (mg daily) > 300  300

3361 3252 3176 3361 3324 2276 3169 3335 2498 2992 2728 3025 3025 3025 3025

Group size

%

1906 1455

56.7 43.3

1459 1793

44.9 55.1

1245 1931

39.2 60.8

1441 1920

42.9 57.1

1378 1946

41.5 58.5

896 1880

32.3 67.7

1620 1549

51.1 48.9

903 2432

27.1 72.9

217 2281

8.7 91.3

2340 652

78.2 21.8

2462 266

90.2 9.8

1669 1356

55.2 44.8

939 2083

31.1 68.9

1278 1740

42.3 57.7

1419 1606

46.9 53.1

Mean  s.d.

25  4.8

44.1  16.2

2822 1147 85.6  44.8

348.8  271.1

Factors associated with overweight and central body fat VM Ramos de Marins et al

values in a univariate analysis. Following that, multiple logistic regression models were used to assess the in¯uence of the independent variables on overweight and central body fat.14 Variables included in the models were those that presented at least a univariate marginal association with each dependent variable (P-value < 0.10). These variables were then retained in a model as long as their 95% con®dence intervals on the multivariate model did not include the value `1'. Interaction was investigated for the variables pertaining to physical activity (occupation and physical activity) and schooling, and to income and schooling. The SPSS software version 8 was used for all data manipulation and model estimation.

239

Results From the 2040 residences at ®rst selected, 1668 agreed to participate in the study, implying a non-response rate of 18.2% (323 refusals to participate; 49 closed or otherwise inaccessible). The original sample consisted of 3997 adults above 20-y-old, and anthropometric measurements could be obtained for 3469 of them. Subjects with BMI below 18.5 were further excluded, resulting in a sample of 3361 subjects (43.3% males, 56.7% females). Table 1 shows the sample distribution of the studied variables. The average age of the subjects was 44 y. Overweight had a prevalence of 44.9%, and high waist=hip girth

Table 2 Univariate analysis (odds ratio (OR), 95% con®dence intervals and P-values), body mass index (BMI), waist=hip girth ratio (WHR) and main independent variables BMI Variables Gender Female Male Age  45 y < 45 y Schooling Lower ( 8 y) Higher ( >8 y) Residence area Slum Non-slum Income per capita (US$) < 200.00  200.00 Smoking Smoker Non-smoker Alcohol consumption  14 weekly doses < 14 weekly doses Activity other than locomotion to work No Yes Occupation Low energy expenditure High energy expenditure Energy intake (Kcal) > 2500  2500 Fat Consumption > 30% of diet energy  30% of diet energy Saturated fat consumption > 10% of diet energy  10% of diet energy Cholesterol consumption (mg daily) > 300  300

WHR

Total

% in risk

OR (95% CI)

P-values

Total

% in risk

OR (95% CI)

P-values

1838 1414

44.9 44.8

1.0 (0.8±1.1)

0.98

1793 1383

52.0 22.6

3.7 (3.2±4.3)

0.000

1392 1857

50.9 40.3

1.5 (1.3±1.8)

0.000

1355 1817

46.7 33.5

1.7 (1.5±2.0)

0.000

1341 1874

52.4 39.5

1.7 (1.5±1.9)

0.000

1304 1837

50.2 31.2

2.2 (1.9±2.7)

0.000

869 1830

44.0 46.7

0.9 (0.8±1.0)

0.19

842 1788

38.5 40.5

0.9 (0.8±1.0)

0.32

1561 1507

45.3 44.4

1.0 (0.8±1.1)

0.60

1538 1454

41.0 37.6

1.1 (1.0±1.3)

0.006

870 2356

39.4 46.8

0.7 (0.6±0.9)

0.000

850 2301

37.8 39.5

0.9 (0.8±1.0)

0.36

210 2218

44.8 45.8

0.9 (0.7±1.2)

0.77

204 2158

36.8 40.7

0.8 (0.6±1.1)

0.28

2257 635

46.1 40.8

1.2 (1.0±1.5)

0.18

2223 615

41.9 29.3

1.7 (1.4±2.1)

0.000

2387 263

45.4 43.3

1.0 (0.8±1.4)

0.52

2329 258

40.7 27.1

1.8 (1.4±2.5)

0.000

1623 1303

44.5 45.0

1.0 (0.8±1.1)

0.82

1589 1280

38.8 39.0

1.0 (0.8±1.1)

0.90

907 2016

45.0 44.4

1.0 (0.9±1.2)

0.66

890 1976

40.1 38.3

1.0 (0.9±1.3)

0.35

1236 1683

45.0 44.4

1.0 (0.9±1.2)

0.75

1205 1657

38.8 38.9

1.0 (0.8±1.2)

0.99

1379 1547

45.0 44.5

1.0 (0.8±1.2)

0.76

1348 1521

38.4 39.3

1.0 (1.0±1.3)

0.65

Public Health

Factors associated with overweight and central body fat VM Ramos de Marins et al

240

ratio 39.2%. A large proportion of the subjects had a high intake of energy (55%), total fat (31%), saturated fat (42%) and cholesterol (47%). Table 2 shows the univariate association between the independent variables, BMI and WHR. Variables considered as presenting a statistically signi®cant association with BMI were: age, schooling and smoking. For the waist=hip girth ratio model, the variables selected were: gender, age, schooling, income, activity other than locomotion to work and occupation. Finally, Table 3 presents the results of the logistic regression models for BMI and waist=hip girth ratio, with the adjusted OR and their respective 95% con®dence intervals. The variables retained for the WHR model were age (1.8, 1.5 ± 2.1), schooling (2.2, 1.9 ± 2.6), gender (3.9, 3.2 ± 4.7) and occupation (1.5 1.1 ± 2.0). For the BMI model, these were age (1.5, 1.3 ± 1.7), schooling (1.7, 1.4 ± 1.9) and smoking (0.8, 0.7 ± 0.9). Both models had an overall correct classi®cation rate close to 70%. No interaction terms could be incorporated into the models.

Discussion The demographic and ecomomic transition that many developing countries are undergoing contributes to changes in lifestyle and diet such as a shift towards energy-dense diet, high in saturated fat and re®ned carbohydrates, and a low consumption of complex carbohydrates ®bres.15 This could be observed in the high proportion of subjects with elevated fat and cholesterol intake observed in the study. Weight and height are the most commonly used anthropometric measures in epidemiological studies. They are easily performed, usually precise, and well accepted by communities. Although not precisely representing body fat consumption, the BMI index is also widely used, due to its ease of measurement and the abundance of data referring to body mass and height. The present study, as well as others,16,17 has concluded that WHR is a complementary

Table 3 WHR and BMI risk predictors, logistic models, bvalues, adjusted odds ratios and 95% con®dence intervals. Model w2 P-values < 0.0001 WHR

b

OR

95% CI

Retained variables Age Schooling Gender Occupation

0.57 0.79 1.4 0.39

1.8 2.2 3.9 1.5

[1.5±2.1] [1.9±2.6] [3.2±4.7] [1.1±2.0]

BMI

b

OR

95% CI

0.40 0.51 70.27

1.5 1.7 0.8

[1.3±1.7] [1.4±1.9] [0.7 ± 0.9]

Retained variables Age Schooling Smoking Public Health

index to the more used BMI in obesity evaluation. Waisthip ratio is an easy to measure indicator, that does not demand scales or more sophisticated fat measurement instruments, and therefore should be recommended for obesity and overweight monitoring in Third World populations. Body fat location is possibly more important than the total amount of fat present in the body. Central, abdominal or `android' obesity consists of fat of easy mobilisation, being considered a serious risk factor for cardiovascular diseases, diabetes mellitus, hypertension and some forms of cancer.4 Although it is known that body fat distribution is basically genetically determined, the role of other factors, such as age, gender, lifestyle factors and dietary composition on fat deposition are not yet well understood.16 ± 19 Overweight prevalences obtained in the present study were 44.8 for males and 44.9 for females. That indicates a worrying trend of prevalence increase in the country. These proportions for Brazil Southern region were respectively 18 ± 28% in 1974 ± 19756 and 32 ± 42% in 1989.7 This trend has also been observed in other countries of Latin America5 and indicates that prevalence is quickly approaching the very high levels presently observed in the United States.3 Overweight prevalence usually is higher among women and increases with age in both men and women.19 ± 23 In the present study, age was an important predictor for both outcomes, even controlling for schooling (BMI) and schooling, occupation and gender (WHR). Agerelated WHR increase, particularly in women, can be attributed to estrogen and physical activity reduction that accompanies age increase and menopause.24 Epidemiological investigation has established a close link between overweight prevalence and socio-economic status, and, at least in developed nations, overweight is more frequently observed in the lower socio-economic strata.25 This phenomenon has been attributed to the nutritional style associated with poverty, traditionally based on cheaper and easier to obtain food such as simple carbohydrates.26 Schooling was an important predictor for both analysed outcomes, as has been previously detected, especially for BMI measures.5,27,28 Martorell et al,5 in a survey of results from ten Latin-American countries, found a protective effect of education in studies relative to all of them, except in the two poorer ones (Haiti and Guatemala), which probably had differentiated socio-economic and cultural characteristics.5 Energy consumption variables did not present an association with the analysed outcomes, when controlling for occupation, demographic and socio-economic variables. This is a dif®cult to interpret phenomenon, but a frequent ®nding in transverse studies, and the possibility of inverse causality cannot be excluded. Other studies have reported low consumption levels in overweight subjects, either due to the under-reporting proclivities in this sub-group or to their self-controlled consumption reduction (dieting).29,30

Factors associated with overweight and central body fat VM Ramos de Marins et al

Besides, energy expenditure level is a dif®cult to control confounding factor in this kind of study. Other studies have also detected an inverse relationship between smoking and body weight.28,31 ± 34 For instance, in a WHO collaborative study comprising 48 populations, in not one of them did smoking occur in higher BMI.32 Leisure time physical activity (activity other than locomotion to work) could not be retained in the WHR and BMI models. Possibly, this re¯ects the need for more detailed and precise information collection, including the duration and intensity of the activities. In contrast to other studies in the area, occupation was also used as an indicator of physical activity levels. This is an easy form of energy expenditure measuring, being simple to recall, well de®ned, and truly re¯ecting, most of the time, the real energy expenditure level of subjects. Occupation was the energy expenditure variable selected in the WHR models, re¯ecting the fact that occupations with low energy expenditure usually imply large amounts of time being spent in a seated position, with subsequent abdominal fat deposition. This variable, however, was not included in the BMI models, as can be accounted for by the fact that BMI is an indicator of total body fat that does not take into consideration fat location. Cross-sectional studies are widely used as a means of identifying population sub-groups more vulnerable to a speci®c health problem. The adverse effects of elevated waist-to-hip ratio and total body fat are well known, but strategies for reducing overweight and upper body obesity are not well established. Studies such as the present one, therefore, can be very useful in the de®nition and design of public health strategies for weight reduction. References 1 National Research Council (US). Committee on Diet and Health. Diet and health: implications for reducing chronic diseases risk. Academy Press: Washington, DC, 1989. 2 Flegal KM, Harlan WR, Landis JR. Secular trends in body mass index and skinfold thickness with socioeconomic factors in young adult men. Am J Clin Nutr 1988; 48: 544 ± 551. 3 Hill JO, Peters JC. Environmental contributions to the obesity epidemic. Science 1998; 280: 1371 ± 1374. 4 Must A et al. The disease burden associated with overweight and obesity. JAMA 1999; 282: 1523 ± 1529. 5 Martorell R et al. Obesity in Latin American women and children. J Nutr 1998; 128: 1464 ± 1473. 6 ENDEF (Estudo Nacional de Despesa Familiar). Tabela de composicËaÄo de alimentos. Secretaria de Planejamento de PresideÃncia da RepuÂblica. FundacËaÄo Instituto Brasileiro de Geogra®a e EstatõÂstica: Rio de Janeiro, Brasil, 1985. 7 Instituto Nacional de AlimentacËaÄo e NutricËaÄo. Pesquisa Nacional de SauÂde e NutricËaÄo. Ministry of Health: Brasilia, Brazil, 1991. 8 Sichieri R et al. High temporal, geographic and income variation in body mass index among adults in Brazil. Am J Public Health 1994, 84: 793 ± 798.

241

9 Lohman TG, Roche F, Martorell R. Anthropometric Standardization Reference Manual. Human Kinetics Books: Illinois, USA, 1988. 10 WHO Expert Committee. Physical status: the use and interpretation of anthropometry. Technical Report Series no 854. WHO: Geneva, 1995. 11 Keenan NL, Strogatz DS, James SAS, Ammerman AS, Rice BL. Distribution and correlates of waist-to-hip ratio in black adults: the Pitt County study. Am J Epidemiol 1992; 135: 678 ± 684. 12 United States Department of Agriculture. Food consumption, prices and expenditures. USA: Washington, DC, 1997. 13 WHO. Diet, nutrition and prevention of chronic disease. Report of a WHO study group. Technical Report Series 797. WHO: Geneva, 1990. 14 Hosmer DW, Lemeshow S. Applied Logistic Regression Wiley: New York, 1989. 15 Caballero B, Rubstein S. Environmental factors affecting nutritional status in urban areas of developing countries. Arch Latinoam Nutr 1997; 47(2 Suppl 1): 3 ± 8. 16 Kaye AS et al. Psychosocial correlates of body fat distribution in black and white young adults. Int J Obes Relat Metab Disord 1993; 17: 271 ± 277. 17 Laws A, Terry RB, Barret-Connor E. Behavioral covariates of waist-hip ratio in Rancho Bernardo. Am J Public Health 1990; 80: 1358 ± 1362. 18 Wing RR et al. Change in waist-hip ratio with weight loss and its association with change in cardiovascular risk factors. Am J Nutr 1992; 55: 1086 ± 1092. 19 Lissner L et al. Secular increases in waist-hip ratio among Swedish women. Int J Obes Relat Metab Disord 1998; 22: 1116 ± 1120. 20 Gray DS. Diagnosis and Prevalence of Obesity. Med Clin North Am 1989; 73: 1 ± 13. 21 Shimokata H et al. Studies in the distribution of body fat. Effect of age, sex and obesity. J Gerontol 1998; 44: M67 ± 73. 22 Hubbard VS. Introduction. Obes Res 1995; 3(Suppl): S75 ± S76. 23 Al-Nuaim AA, Bamgboye EA, Al-Rubeaan KA, Al-Mazrou Y. Overweight obesity in Saudi Arabian adult population, whole of socio-demographic variables. J Comm Health 1997; 22: 211 ± 221. 24 Taylor RW, Keil D, Gold EJ, Williams SM, Goulding A. Body mass index, waist girth, and waist-to-hip ratio as indexes of total regional adiposity in women: evaluation using receiver operating characteristic curves. Am J Nutr 1998; 67: 44 ± 49. 25 Van Itallie TB. Health implications of overweight and obesity in adults in the United States. Ann Intern Med 1985; 103: 983 ± 988. 26 Rolland-Cachera MF, Bellisle F, Tichet J, Chantrel AM, Bataille MG, Vol S, Pequignot G. Relationship between adiposity and food intake: an example of psuedo-contradictory results obtained in case-control versus between population studies. Int J Epidemiol 1990; 19: 571 ± 577. 27 Flegal KM, Carrol MD, Kuczmarski RJ, Johnson CL. Overweight and obesity in the United States: prevalence and trends, 1960 ± 1994. Int J Obes Relat Metab Disord 1998; 22: 39 ± 47. 28 Paeratakul S, Popkin BM, Keyou G, Adair LS, Stevens L. Changes in diet and physical activity affect the body mass index of Chinese. Int J Obes Relat Metab Disord 1998; 22: 424 ± 431. Public Health

Factors associated with overweight and central body fat VM Ramos de Marins et al

242

 , Vrijheid M, Nichols R, Kiggins M, Elliot P. Who 29 Pryer JA are the low energy reporters in the Dietary and Nutritional Survey of British Adults? Int J Epidemiol 1997; 26: 146 ± 154. 30 Willett W. Future directions in the development of foodfrequency questionnaires. Am J Clin Nutr 1994; 59: 171S ± 174S. 31 Flegal KM, Troiano RP, Pamuk ER, Kuczmarski RJ, Campbell SM. The in¯uence of smoking cessation on the prevalence of overweight in the United States. New Engl J Med 1995; 333: 1165 ± 1170.

Public Health

32 Mollarius A, Seidell, Kuulasmaa K, Dobson AJ, Sans S. Smoking and relative body weight: an international perspective from the WHO MONICA Project. J Epidemiol Community Health 1997; 51: 252 ± 260. 33 Coackley EH, Rimm EB, Colditz G, Kawachi I, Willett W. Predictors of weight change in men: results from the Health Professionals Follow-up Study. Int J Obes Relat Metab Disord 1998; 2: 89 ± 96. 34 Sundquist J, Johansson SE. The in¯uence of socioeconomic status, ethnicity and lifestyle on body mass index in a longitudinal study. Intl J Epidemiol 1998; 27: 57 ± 63.