International Journal of Obesity (2006) 30, 191200 & 2006 Nature Publishing Group All rights reserved 0307-0565/06

$30.00 www.nature.com/ijo

ORIGINAL ARTICLE
A life-course approach in explaining social inequity in obesity among young adult men and women
m1 M Novak1, C Ahlgren2 and A Hammarstro
1

University, Umea , Sweden and 2Occupational Family Medicine, Department of Public Health and Clinical Medicine, Umea University, Umea , Sweden Medicine, Department of Public Health and Clinical Medicine, Umea

Objective: To examine the cumulative influence of adverse behavioural, social, and psychosocial circumstances from adolescence to young adulthood in explaining social differences in overweight and obesity at age 30 years and if explanations differ by gender. Design: A 14-year longitudinal study with 96.4% response rate. Subject: Data from 547 men and 497 women from a town in north Sweden who were baseline examined at age 16 years and prospectively followed up to age 30 years. Measurements: Overweight and obesity were ascertained at ages 16 and 30 years. Occupation and education were used to measure socioeconomic status. The explanatory measurements were: age at menarche, smoking, physical activity, alcohol consumption, TV viewing, home and school environment, social support, social network, and work environment. Results: No gender or social difference in overweight was observed at age 16 years. At age 30 years, significantly more men than women (odds ratio (OR) 2.81, 95% confidence interval (CI) 2.143.68) were overweight or obese. Educational level was associated with overweight at age 30 years, but not occupational class. Both men (OR 1.55, 95% CI 1.102.19) and women (OR 1.78, 95% CI 1.162.73) with low education (p11 years) were at risk of overweight. The factors that explained the educational gradient in overweight among men were low parental support in education during adolescence, and physical inactivity, alcohol consumption, and nonparticipation in any association during young adulthood. The educational gradient in overweight in women was explained mostly by adolescence factors, which include early age at menarche, physical inactivity, parental divorce, not being popular in school, and low school control. Restricted financial resource during young adulthood was an additional explanatory factor for women. All these factors were significantly more common among men and women with low education than with high education. Conclusion: Social inequities in overweight reflect the cumulative influence of multiple adverse circumstances experienced from adolescence to young adulthood. Underlying pathways to social inequity in overweight differ between men and women. Policy implications to reduce social inequity in overweight include reduction of social differences in health behaviours and social circumstances that take place at different life stages, particularly psychosocial circumstances during adolescence. International Journal of Obesity (2006) 30, 191200. doi:10.1038/sj.ijo.0803104; published online 20 September 2005 Keywords: gender; inequity; behavioural factors; social factors; psychosocial factors; longitudinal study

Introduction
The prevalence of obesity has increased dramatically worldwide, particularly in developed countries, and today obesity is regarded as a major contributor to the global burden of disease and disability.1 Numerous health risks are known

Correspondence: M Novak, Family Medicine, Department of Public Health University, Umea SE-901 85, Sweden. and Clinical Medicine, Umea E-mail: masuma.khatun@fammed.umu.se Received 19 July 2004; revised 17 February 2005; accepted 7 April 2005; published online 20 September 2005

to be associated with obesity, including coronary heart disease, hypertension, certain cancers, noninsulin-dependent diabetes mellitus, musculoskeletal disorder, and infertility.1,2 According to some experts, the current generation of children are likely to have shorter life expectancies than their parents because of obesity.3 High body fatness is a strong predictor of long-term sick leave4 and early work disability pension.5 Besides the physical health consequences of obesity, obese people, particularly obese women, suffer from social stigmatization, prejudice, and discrimination, which may even influence their social class.6,7 Thus, both direct and indirect costs attributed to obesity are huge

Explanations to social inequity in obesity M Novak et al

192
for a society. In Sweden, direct costs for obesity-related complications are estimated to be SEK 3 billion per year, and indirect costs (such as for sick leave, early pensions, etc.) are presumed to be as high as for direct costs.8 Therefore, it is essential for a society to identify risk groups as well as underlying factors that contribute to the development of obesity. Across all industrial countries, considerable evidence suggests that people with low socioeconomic status (SES) are at greater risk of becoming overweight and obese than people with high SES.913 Social differences in obesity are reported to be widening in Sweden, particularly among women.14 However, the mechanisms by which social inequities in health are generated and maintained in particular contexts are not fully understood. In understanding the mechanisms, the importance of life-course perspective has increasingly been recognized.15,16 In a lifecourse perspective, cumulative differential exposure is one of the important mechanisms, also commonly known as the social causation hypothesis. According to the social causation hypothesis, social differences in health occur due to accumulation of multiple adverse circumstances experienced by the lower SES groups throughout the life course, as compared to the more favourable conditions experienced by the higher SES groups.15 With regard to obesity, it has been suggested that psychosocial stress induced by multiple negative life circumstances may lead to unhealthy behaviour, resulting in increased risk of obesity. A study of middle-aged women reported that psychosocial stress (job strain, poor social support) and recent unhealthy behaviour together with a reproductive history of early menarche and high parity explained a large proportion of the association between low social class and obesity.17 Other studies investigating the role of recent health behaviour suggest that health behaviour account for only some of the class variance in body mass index (BMI, kg/m2).18,19 However, these studies1719 focus on recent life circumstances only, use limited indicators, and are cross-sectional in design, which makes it difficult to determine the direction of causality. Research on the causes of obesity emphasizes the importance of long-term influence of early life circumstances,16 particularly psychosocial circumstances.20 Parental social class,10 health-related behaviour,21 and social acceptance22 during childhood were reported to be associated with increased body fat in longitudinal studies. In a longitudinal study, teachers assessment of parental neglect during childhood was found to be a risk factor for obesity in young adulthood.20 However, the influence of early adverse circumstances in explaining social gradients in adult obesity has rarely been studied longitudinally. In addition, research on cumulative influences of adverse circumstances during lifetime (from early to recent years) on social inequities in obesity is sparse, mainly because few studies are designed longitudinally. This longitudinal study aims to examine explanations for social inequities in obesity (if any) among 30-year-old men and women. We focused on the cumulative
International Journal of Obesity

influence of behavioural, social, and psychosocial circumstances from adolescence to young adulthood in explaining social differences in overweight and obesity at age 30 years and if the explanations differ by gender.

Population and methods

The data and the subjects The analysis is based on longitudinal data from a 14-year follow-up study. The study was carried out in an industrial town in the north of Sweden. The baseline survey was conducted in 1981, when all the 16-year-old pupils in the last year of compulsory school (grade 9) completed a comprehensive self-administrated questionnaire in the classroom (577 boys and 506 girls). In the Swedish school system, all the pupils follow the same core curriculum until grade 9. The pupils were followed up again at ages 18, 21 and 30 years using the same survey questionnaire. Extensive work was carried out to reduce the nonresponse rate to a minimum. The total response rate was 96.4% throughout the 14-year follow-up. The present study data are based on the questionnaire surveys at age 16 and 30 years, school records from age 16 years, and interviews with the form teachers at age 16 years. A total of 65 teachers were interviewed with structural questionnaires. In Sweden, the form teachers usually have regular contact with the pupils, they teach at least one major subject, and coordinate the class work. They also have the main responsibility for contact with parents. The form teachers had known the pupils for at least 6 months.

Outcome variable The BMI was calculated from measured weight and height data recorded in school health records at baseline, and from self-reported data at age 30 years. The BMI at 30 years was categorized according to the criteria recommended by WHO.23 The categories for men and women are: nonoverweight o25 kg/m2, preobesity (or overweight) 2530 kg/m2, and obesity 430 kg/m2. Owing to the low prevalence of obesity (men 7.4% and women 3.8%), the outcome variable was dichotomized by combining preobesity and obesity together. Nonoverweight was used as a reference category. The BMI at 16 years was categorized according to the international reference values recommended by Cole et al.24 as follows: overweight 23.9028.88 kg/m2 for boys and 24.3729.43 kg/m2 for girls, and obesity 428.88 kg/m2 for boys and 429.43 kg/m2 for girls.

Measures of SES SES was assessed using the Swedish socioeconomic classification (SEI) of occupational categories.25 Manual workers were grouped into blue-collar workers and nonmanual into whitecollar workers. At age 16 years, the classification was based on parents occupation, primarily the fathers occupation,

Explanations to social inequity in obesity M Novak et al

193
but in single-mother households, the mothers occupation was used. At age 30 years, the subjects own occupation and educational level was used. Educational level was categorized as low (p11 years of schooling) or high (X12 years). assessed based on a question reading can you raise a sum of US$1860 in a week by any means? with the alternative responses yes or no. Experience of physical violence and or threats in the last year was coded as no versus yes. The perceived availability of social network and social support was assessed. The social network index represented the number of persons with whom links were recognized (range 424) and the social support index reflected the degree of emotional and material support received when needed (range 623). Social activity was assessed by the participation or nonparticipation in any association. Which kind of association was not specified in the questionnaire, but associations could, for example, be related to sport-, cultural-, or political activities, etc.

Maturation and health-related behaviour, age 16 years The girls age at menarche was assessed in years. Factors used as measures of health behaviour were smoking, alcohol consumption, physical activity, and TV viewing. Smoking was dichotomized as never or stopped smoking versus currently smoking. Alcohol consumption was estimated in centilitres of pure alcohol per year. We assessed the frequency of physical activity using a score from 0 to 2, with 0 indicating regular, 1 infrequent, and 2 no activity. The subjects were asked to report how many TV programmes they watched. Alternative responses were from less than one programme/week to several programmes/day (scales 15).

Home-related factors, age 16 years Parental divorce was defined as if the participants were living with one parent. Parental support in education was assessed based on teachers assessments of parental support in their childrens studies with the response range 15 (very much to very little). An index of contact with parents was constructed based on participants assessment of two separate questions related to contact with father and mother. The alternative responses were on a five-grade scale ranging from very good to very bad (15). The index range was 29.

Work-related factors, age 30 years Psychosocial job characteristics were assessed using two 6-item scales adopted from Karasek and Theorell26 on job demand (referring to quantity of work, intellectual requirements, and time constraints of the job) and on job control (referring to the possibilities of making decisions, being creative, and using and developing ones own abilities), both ranging from 6 to 24. The scales were dichotomized according to the median split (9 and 8). All factors were constructed from questionnaires that had previously been validated.2628

School-related factors, age 16 years A composite index of school control was constructed based on three questions: (1) ability to decide in school; (2) ability to pursue own interests; and (3) if anything they learnt at school could be useful in future. The alternative responses to each question ranged from 1 to 5 (very much to very little) and the composite index range was 315. An index of popularity in school was constructed based on teachers assessment of two questions related to popularity among students and among teachers. The alternative responses were on a six-grade scale ranging from most to least popular (16). The index range was 212.

Health-related behaviour, age 30 years Similar behavioural factors and coding as at age 16 years were used for age 30 years, except for frequency of watching TV, which was ascertained as number of hours per week.

Social factors, age 30 years Marital status was defined as married or cohabiting versus single or never married. The number of children at age 30 years ranged from 0 to 4. Restricted financial resources were

Statistical analysis Statistical analyses were performed in SPSS version 11.5. (SPSS Inc., Chicago, IL, USA). All the indices were calculated based on factor analysis (varimax), and items with factor loading X0.50 were included in the indices. All the scales were dichotomized in order to enhance the accuracy of the multivariate analysis.29 We used three approaches for the data analysis. First, we assessed the distribution of the selected factors (for this study) between men and women (Table 1) as well as between the SES groups (Table 3). The distributions of the factors were expressed in proportion, mean, and median. Differences in the distributions between groups were tested using w2 tests for discrete variables, t-tests for continuous variables with normal distribution, and nonparametric test (Mannwhitney U-test) for continuous variables with skewed distribution. A P-value o0.05 was considered significant. Secondly, we estimated whether there were gender and SES differences in overweight at ages 16 and 30 years (Table 2). Binary logistic regressions were used to assess the associations, which were expressed in odds ratios (OR) with 95% confidence intervals (CI). In the last step of analyses, a series of logistic regressions were conducted to examine the influence of factors from adolescence to young adulthood on SES differences in overweight at age 30 years. We fitted separate univariable logistic models to estimate the effect of each explanatory
International Journal of Obesity

Explanations to social inequity in obesity M Novak et al

194
Table 1 Prevalence of overweighta and obesity at ages 16 and 30 years and the distribution of factors thought to be associated with overweight or obesity at age 30 years by gender Men (n 547) Variables from age 16 years Overweighta Obesity Physical activity: never Currently smoking Alcohol consumption, median (P25P75) TV viewing (several programmes/day) Parents social class (blue-collar) Divorced parents Low parental support in education Bad contact with parents Not popular in school Low school control Variables from age 30 years Overweighta Obesity Physical activity: never Currently smoking Alcohol consumption, median (P25P75) TV viewing Single Number of children, median (P25P75) Years of education, mean (7s.d.) Restricted financial resources Nonparticipation in any association Poor social network Poor social support Violence High job demand Low job control Women (n 497) Men OR (95% CI) 8.3 0.7 11.6 18.5 1.2 (0.012.5) 39.9 50.3 22.6 22.6 20.1 53.7 35.7 6.3 0.8 7.2* 34.3*** 1.4 (0.09.7) 17.0*** 45.0 21.0 21.0 32.6*** 40.3*** 34.6 Age 16 years Parents social class White-collar Blue-collar Age 30 years Own social class White-collar Blue-collar Own education High education Low education Gender differences Age 16 years Age 30 years Women OR (95% CI) Table 2 Social class and gender difference in overweighta at age 16 and 30 years, according to bivariate logistic regression Class differences Overweighta

1.00 1.35 (0.751.44)

1.00 1.36 (0.712.61)

1.00 1.29 (0.901.83) 1.00 1.55 (1.102.19)

1.00 1.49 (0.972.23) 1.00 1.78 (1.162.73)

1.03 (0.671.66) 2.81 (2.143.68)

1.00 1.00

39.7 7.4 28.4 25.2 17.7 (9.133.8) 44.9 33.3 0.0 (0.01.0) 11.9 (71.6) 14.5 60.9 32.6 44.0 9.7 48.0 36.1

20.3*** 3.8*** 27.2 34.2*** 5.6 (2.613.2)*** 32.0 *** 21.2*** 1.0 (0.02.0)*** 12.3 (71.7)** 29.8*** 62.3 44.1*** 60.5*** 8.0 48.9 40.0

a Includes both overweight and obesity; high education X12 years; low education p11 years; OR odds ratio; CI 95% confidence interval.

subsequent multivariable model (Table 5). A stepwise approach was used to construct this model, following a temporal sequence: first adolescence factors were included and then adulthood factors. The method allowed for the timing of particular events and circumstances, providing an indication of the relative and cumulative effects of particular explanations.30 The data analysis approach applied here has been used in previous studies.30,31 Collinearity between the independent variables was explored using the technique developed by Wax32 and found results unaffected. All the analyses were performed separately for men and women.

Values are percentages of subjects, unless stated otherwise. n number of participants. aObesity is not included; P25 25th percentile, P75 75th percentile; s.d. standard deviation; *, **, *** P-value o0.05, o0.01, o0.001. P-values represent the differences between men and women.

Results
The distribution of factors theoretically thought to be associated with overweight at age 30 years are shown for men and women at age 30 years in Table 1. According to Table 1, about half of the adverse factors were equally distributed between men and women. Among the rest, more risk factors were found to be common among men than in women. More men had been physically inactive, watched more TV, and were reported as being less popular at school at age 16 years. At age 30 years, a higher proportion of men were overweight and obese, consumed more alcohol, more often watched TV, were single, and had fewer years of education than women. On the other hand, at age 16 years, more women were smoking and experienced bad contact with their parents. At age 30 years, more women had children, reported restricted financial resources, poor social support, and poor social network than men.

factor on SES differences in overweight, using the differences between the unadjusted and adjusted OR. The OR of overweight among low SES group compared with high SES group at age 30 years was the base model from which the contributions of individual factors were estimated. Since obesity during early years is known to be associated with later obesity as well as with later SES,6,7 BMI at age 16 years is controlled for in the base model. First, we determined whether the adjustment for an individual explanatory factor in the base model resulted in a reduction of the OR (Table 4). The contribution of factors was measured by the percentage reduction in the OR ((unadjusted ORadjusted OR) 100/ (unadjusted OR1)).17 Factors that resulted in a reduction of the OR by 10% or more were then selected for inclusion in a
International Journal of Obesity

Explanations to social inequity in obesity M Novak et al

195
The differences in overweight between SES groups and gender at ages 16 and 30 years are shown in Table 2. No SES or gender differences in overweight were observed at age 16 years. At age 30 years, no SES difference was observed for men or women when an occupational measure of SES was used. Using educational measures, men and women with low education had a significantly higher risk of overweight. For instance, the OR for overweight in women is 1.78 (95% CI 1.162.73) among those who attended only p11 years of schooling compared to women with longer education. Gender difference was also evident at age 30 years, where men had almost three times higher risk of being overweight than women (OR 2.81, 95% CI 2.143.68). Since social difference in obesity was found only when the educational measurement was used, only results of the educational measurements are presented in the following sections. The distribution of factors thought to be associated with overweight at age 30 years is presented in Table 3 by educational level and gender. Almost every adverse factor from early years onward was significantly more prevalent among men and women who attended only p11 years of schooling. For instance, 12% women with low education had experienced violence as compared to 5.5% women with high education. The contribution of factors from adolescence to early adulthood in explaining educational differences in overweight at age 30 years is presented in Table 4 by gender. The ORs for educational differences are given with adjustments separately for each factor, in order to identify factors that had substantial influence on educational differences in overweight, that is, factors that reduced the OR for educational differences by 10% or more. A total of 11 factors were identified, among which four factors for men and six factors for women substantially accounted for the educational gradient in overweight (Table 4). For instance, among men, an estimated 29% of the increased risk in overweight of the lower educational group can be attributed to low parental support in education during adolescence. Among women, an estimated 26% of the increased risk of overweight of the lower educational group can be attributed to low school control during adolescence. The 11 factors with substantial effect on the overweight differentials were selected from Table 4 for multivariate analysis, as shown in Table 5. Controlling for all the selected factors simultaneously, the ORs for educational differences in overweight reduced from 1.51 to 1.16 in men and from 1.70 to 1.22 in women. This means that the selected factors play a significant role in explaining educational differences in overweight.

Table 3

Distribution of factors thought to be associated with overweight or obesity at age 30 years by educational level and gender Men (n 547) Educational level at 30 years High (X12 years) Low (p11 years) Women (n 497) Educational level at 30 years High (X12 years) Low (p11 years)

Variables from age 16 years Physical activity: never Currently smoking Alcohol consumption, median (P25P75) TV viewing (several programmes/day) Age at menarche, mean (7s.d.) Parents social class (blue-collar) Divorced parents Low parental support in education Bad contact with parents Not popular in school Low school control Variables from age 30 years Physical activity: never Currently smoking Alcohol consumption, median (P25P75) TV viewing Single Number of children, median (P25P75) Restricted financial resources Nonparticipation in any association Poor social network Poor social support Violence High job demand Low job control

6.0 10.2 0.2 (0.06.2) 38.3 F 37.5 19.2 8.7 14.6 40.7 24.0

18.2*** 27.4*** 3.2** (0.015.8) 41.4 F 64.3*** 25.5* 38.4*** 25.6*** 67.8*** 48.0***

3.1 20.8 0.5 (0.05.3) 12.9 12.8 (71.2) 40.4 15.8 13.0 27.5 29.8 28.7

11.9*** 52.4*** 3.6*** (0.015.5) 22.1** 12.7 (71.2) 51.9** 29.2*** 31.8*** 40.9** 54.1*** 40.9***

16.4 18.1 16.1 (8.529.9) 41.5 30.4 0 (01) 7.6 50.4 28.8 46.7 9.0 54.6 24.8

41.2*** 32.9*** 19.3* (10.436.6) 48.4 36.8 0 (02) 22.4*** 72.1*** 35.7 41.1 10.7 41.2*** 48.7***

23.3 26.0 6.0 (2.513.2) 28.8 20.5 1 (02) 20.5 55.5 39.9 64.8 5.5 50.8 32.7

31.5 46.0*** 5.1 (2.611.8) 36.9 21.7 2 (02)*** 42.2*** 71.8*** 50.3* 54.6* 12.0* 44.6 50.7***

Values are percentages of subjects, unless stated otherwise. n number of participants; P25 25th percentile; P75 75th percentile; s.d. standard deviation; *, **, *** P-value o0.05, o0.01, o0.001. P-values represent the differences between the high and low educational groups within gender.

International Journal of Obesity

Explanations to social inequity in obesity M Novak et al

196
Table 4 OR for educational differences in overweighta at age 30 years unadjusted for explanatory variables (top) in a separate univariable logistic regression analysis Men (n 547) OR Educational level at age 30 years (low versus high) Age 16 years Physical activity: never Currently smoking Alcohol consumption TV viewing Age at menarche Parents social class Divorced parents Low parental support in education Bad contact with parents Not popular in school Low school control Age 30 years Physical activity: never Currently smoking Alcohol consumption TV viewing Single Number of children Restricted financial resources Nonparticipation in any association Poor social network Poor social support Violence High job demand Low job control
b

Explanatory variables

Women (n 497) % reduction F OR 1.70 95% CI 1.082.67 % reduction F

95% CI 1.052.17

1.51

1.50 1.57 1.50 1.51 F 1.56 1.49 1.36 1.50 1.62 1.55

1.022.20 1.092.32 1.042.17 1.052.18 F 1.072.28 1.042.15 0.912.04 1.042.16 1.112.38 1.062.26

2 0 2 0 F 0 4 29 2 0 0

1.58 1.68 1.82 1.73 1.63 1.69 1.55 1.72 1.69 1.57 1.52

0.9892.52 1.042.57 1.142.89 1.092.76 1.022.58 1.072.66 0.982.46 1.062.76 1.072.62 0.982.51 0.962.42

17 3 0 0 10 1 21 0 1 19 26

1.34 1.50 1.42 1.49 1.51 1.51 1.66 1.36 1.51 1.48 1.49 1.52 1.50

0.911.95 1.052.17 0.9952.06 1.032.15 1.052.15 1.052.17 1.142.42 0.922.00 1.052.18 1.022.13 1.032.15 1.052.19 1.032.19

33 2 18 4 0 0 0 29 0 6 4 0 2

1.72 1.74 1.80 1.70 1.69 1.72 1.56 1.74 1.75 1.77 1.73 1.70 1.68

1.062.73 1.082.82 1.132.83 1.072.70 1.072.67 1.092.81 0.982.49 1.092.79 1.112.77 1.112.82 1.092.72 1.082.67 1.052.55

0 0 0 0 1 0 20 0 0 0 0 0 3

Each estimate indicates the OR for educational differences in overweight resulting from an adjustment for the specified variable in a bivariate model. aIncludes both overweight and obesity. bControlled for BMI at age 16 years. n number of participants; OR odds ratio; CI 95% confidence interval. % reduction indicates percentage reduction of OR for educational differences in overweight resulting from an adjustment for the specified variable. The reduction was estimated as: ((unadjusted ORadjusted OR) 100/(unadjusted OR1)).

Discussion
In this Swedish cohort, SES inequity in overweight was observed at age 30 years, but not at age 16 years. Both men and women with low educational level were more likely to be overweight. The educational gradient in overweight was wider among women than among men. The results revealed that cumulative exposures to adverse behavioural and psychosocial factors from both adolescence and recent years played a substantial role in the development of obesity among the low SES groups. The analysis also demonstrated that the explanatory factors to obesity were different for men and women, which is consistent with the study by Ball et al.19 The strength of the present study includes the availability of longitudinal data, collected prospectively throughout different life stages of the same populations. The life-course approach, therefore, could be used in explaining social gradients in overweight. The longitudinal data employed here are also quite unique for the extremely high response rate during a 14-year follow-up (96.4%). In addition, the questionnaire was sent to all the pupils of a certain age group
International Journal of Obesity

in the municipality. Therefore, this cohort could be representative of young people of middle-range industrial towns in Sweden during the 1980s and 1990s. The major limitation of this study includes the use of self-reported measures of BMI at age 30 years. The misclassification associated with self-reports may cause bias in the prevalence of overweight and differentials between population groups, for example, social classes.33 In a Swedish study, underestimation of BMI was found among all SES groups, but the highest underestimation was found among men with high SES and women with low SES.33 Therefore, it is suggested that using selfreported BMI, the class differences in overweight and obesity may be overestimated among men, but underestimated among women.33 If so, it would imply that the class differences observed in our study might be smaller among men and larger among women than was reported. The present analyses demonstrated that the occupational position was not associated with overweight for either men or women. A review by Sobal and Stunkard34 revealed that in many countries, low occupational position is strongly associated with high obesity among women but less consistent among men. Differences in physical demands in

Explanations to social inequity in obesity M Novak et al

197
Table 5 Cumulative effect of the variables from adolescence to adulthood that contributed to a reduction in educational differences in overweighta among men and women at age 30 years in a multivariable logistic regression model Overweighta at age 30 years OR (95% CI) Men Education (base modelb) High (X12 years) Low (p11 years) Adjusted for parental support in education, age 16 years and physical activity, age 30 years and alcohol consumption, age 30 years and participation in any association, age 30 years Women Education (base modelb) High (X12 years) Low (p11 years) Adjusted for age at menarche, age 16 years and physical activity, age 16 years and parental divorce, age 16 years and popularity in school, age 16 years and school control, age 16 years and restricted financial resources, age 30 years

1.00 1.51 (1.052.17) 1.36 1.24 1.20 1.16 (0.912.04) (0.811.88) (0.791.83) (0.751.79)

1.00 1.70 (1.082.67) 1.63 1.53 1.45 1.40 1.33 1.22 (1.022.58) (0.962.46) (0.902.34) (0.862.29) (0.812.19) (0.712.11)

a Includes both overweight and obesity. b The base model is controlled for BMI at age 16 years. OR odds ratio; CI 95% confidence interval.

mens and womens occupation have been put forward as one possible explanation for these inconsistent findings.35 The authors suggest that low-status jobs for women are more sedentary, while mens jobs are more physically demanding, which might be protective against obesity. In the Swedish context, low-status jobs for women are not necessarily characterized as sedentary. For instance, nursing assistant or cleaner are typical womens jobs and also physically demanding.36 On the other hand, some typical mens jobs such as drivers may be more sedentary.36 Low job control was suggested as another reason for the association found between low occupational position and obesity among women. Low job control is associated with low-status jobs, and is more common among women than in men. Low job control might make it difficult for women to manage time effectively or to adopt a healthy lifestyle, leading to obesity. In our study, low job control among women was as common as among men. This could be a possible explanation for the lack of association found between occupational position and overweight for women. On the other hand, education may be related to obesity primarily through generating knowledge about eating behaviour, physical activity, and the negative health effect of obesity.37 Therefore, one might expect an equivalent effect of education for both men and women. In our study, the association found between education and obesity for men and women is consistent with other studies.35,38,39 However, the categorization of low

(p11 years) and high (X12 years) educational groups in our study requires further discussion. In other countries, p11 years of schooling may not be considered low education. In Sweden, 9 years of schooling (elementary) is compulsory for all, and more than 85% of the study subjects went to 2 more years of upper secondary schooling. Therefore, 11 years of schooling was taken as a cutoff point. Like SES differences in overweight, gender difference in overweight was observed in this study at age 30 years (but not at age 16 years), where higher proportions of men than women were overweight and obese. Our finding partly contradicts the general trend that is observed in the WHO MONICA study on 48 different European countries, where women were more likely than men to be obese, whereas men were more likely to be overweight (age 3564 years).1 However, the prevalence rate of overweight and obesity found among men and women in our study is quite comparable to the national statistics for Sweden. Although comparison was problematic since our prevalence rate applies to a specific age group, that is, 30 years, while a wider age range is being used in the national statistics. The closest age range available for comparison was between 25 and 34 years from survey year 1998. There an overweightand obesity prevalence for men was given as 43 and 6%, respectively, and for women the corresponding rates were 25 and 5%.40 These findings are close to our prevalence rate with 39.7 and 7.4% among men and 20.3 and 3.8% among women (with survey year 1995). An interesting observation in our study was that men had larger social networks and better social support than women. This finding is also quite contrary to earlier literatures. Our speculation is that in the study town there is only one Technical University, where most men are studying. Therefore, most likely women might have moved out of the town for higher education. As consequence, they have less social support and smaller social networks in a new town compared to the men who did not move from their hometown.

Explanations for educational differences in overweight As shown in a previous study,17 early menarche was found to be one of the explanatory factors for a higher rate of obesity among women with low SES in this study. Although early menarche was not more common among girls in lower than in higher SES groups in the present study, it is plausible that early menarche in combination with other adverse behavioural factors among the lower SES group (for instance, physical inactivity and high alcohol consumption) may make them more prompt to develop obesity. Although the mechanism of the association between early menarche and adult obesity is uncertain, it has been suggested that it might be both biological and sociocultural. Various endocrine factors influence the accumulation of body fat,41 and also that those girls who mature early are fatter at the time maturation begins than other girls of the same age.42 Owing to the differences in body appearance between early and late
International Journal of Obesity

Explanations to social inequity in obesity M Novak et al

198
maturing girls, early mature girls may develop a negative body image. A negative body image is found to be one of the main causes of depressive symptoms among adolescent girls,43 which may lead to comfort-eating behaviour or withdrawal from group activities such as school sports or physical activities, which in turn predisposes them to obesity. In previous studies, early maturing girls were reported to have negative body image, lower self-esteem,44 and to be physically less active than late maturing girls of the same age.45 Previous studies have shown a protective effect of physical activity on fatness.46,47 In our study, recent physical activity was found to be important for men, while physical activity during adolescence was important for women. In addition to physical activity, several other factors from adolescence were important for women in the development of obesity, while several recent factors were important for men. A similar trend was observed in an earlier study carried out on this cohort, where adolescence factors were important for women and recent factors for men in the development of class-based musculoskeletal disorders.48 In the present study, recent nonparticipation rate in association activities was found to be significantly higher among men with low education than men with high education, and this factor was also found to contribute to high obesity among lower educated men at age 30 years. The differences in participation rate could be due to economy (e.g. sports club activities) and or differences in preferences for social activities between the SES groups. The present study showed that an unfavourable environment at home and in school during adolescence, such as parental divorce, not being popular in school and low school control was associated with adult obesity in women. The relationship between unfavourable life circumstances during adolescence and the later development of obesity is complex, but several potential mediators appear plausible. Adolescence is an important period of life for an individual course of development. An unfavourable environment during this period might negatively affect their self-worth and also cause stress. Negative self-worth and stress due to unfavourable circumstances may lead to unhealthy behaviours, resulting in an increased risk of obesity. In a Swedish study of school children (912 years), parental divorce was found to be associated with somatic ill health and stress among girls.49 Boys, on the other hand, were reported to react differently to the same situation, increasing their alcohol consumption.50 Evidence from an earlier study conducted on our study cohort suggests that both among boys and girls the lack of school control at age 16 years was significantly associated with increased alcohol consumption at the age of 18 years and the association was stronger among girls than among boys.51 Among men, but not among women, low parental support in education during adolescence explained a large part of the class variance in later obesity. Lissau and Sorensen,20 in their longitudinal study conducted in Copenhagen, showed that low parental
International Journal of Obesity

support during childhood (910 years) was a strong predictor of later obesity (1920 years). In our study, we have assessed parental support in relation to the participants education only, whereas later studies have assessed parental support in general. Nevertheless, it is interesting that this factor had an influence on later obesity in both studies. In the study by Lissau and Sorensen,20 men and women were analysed together, so it is not known whether parental support had a similar effect for both men and women. In our study, it is not clearly understood why parental support in education during adolescence is less important for girls. However, in relation to parental support in education during adolescence, one may discuss it from the perspective of gender construction in society. Girls are expected to do well and take responsibilities both for themselves and for others from an early age.52 Therefore, guidance or support from parents in schoolwork might be more beneficial for boys in school performance as well as attaining higher education. Low education is known as a strong predictor of obesity. Lack of parental support in education during adolescence may be associated with later obesity through educational attainment, whereby boys with low parental support in education may do poorly in school, and consequently attain low education as adults. In many respects, Sweden is known to be one of the most equitable countries in the world, yet women in general still receive lower salary than men for the same occupational position.53 As shown in this study, more women reported having restricted financial resources than men, particularly women with low education, and this was part of the explanation for obesity among them. One potential mediator might be psychological stress associated with restricted economy. It has been suggested that psychological stress may influence the metabolism and distribution of body fat.54 Another possibility is that a restricted economy is associated with obesity primarily through less healthy diets, which may be related to both lack of knowledge and lack of financial resources. Nutritious foods are, in general, more expensive than foods with high fat and sugar content. Therefore, it might be difficult to keep up with the healthy diets for women with limited financial resources, resulting in an increase risk of obesity among them. All the selected risk factors that were associated with obesity among men and women (described above) were also significantly more common among lower SES groups. The factors could also explain why lower SES groups had higher overweight and obesity. These results support the social causation hypothesis that social differences in health occur due to multiple adverse circumstances experienced by the lower SES groups throughout the life course as compared to the higher social groups.15

Conclusions Social inequities in overweight reflect the cumulative influence of multiple adverse circumstances experienced

Explanations to social inequity in obesity M Novak et al

199
from adolescence to young adulthood. The underlying pathways to social inequity in overweight differ between men and women. Policy implications to reduce social inequity in overweight include reduction of social differences in health behaviours and psychosocial circumstances that take place at different life stages, particularly psychosocial circumstances during adolescence.
17 Wamala SP, Wolk A, Orth-Gomer K. Determinants of obesity in relation to socioeconomic status among middle-aged Swedish women. Prev Med 1997; 26: 734744. 18 Jeffery RW, French SA, Forster JL, Spry VM. Socioeconomic status differences in health behaviors related to obesity: the Healthy Worker Project. Int J Obes Relat Metab Disord 1991; 15: 689696. 19 Ball K, Mishra GD, Crawford D. Social factors and obesity: an investigation of the role of health behaviours. Int J Obes Relat Metab Disord 2003; 27: 394403. 20 Lissau I, Sorensen TI. Parental neglect during childhood and increased risk of obesity in young adulthood. Lancet 1994; 343: 324327. 21 Parsons TJ, Power C, Logan S, Summerbell CD. Childhood predictors of adult obesity: a systematic review. Int J Obes Relat Metab Disord 1999; 23 (Suppl 8): S1S107. 22 French SA, Perry CL, Leon GR, Fulkerson JA. Self-esteem and change in body mass index over 3 years in a cohort of adolescents. Obes Res 1996; 4: 2733. 23 WHO. Obesity: preventing and managing the global epidemic. Report of a WHO Consultation, Geneva, 35 June, WHO/NUT/98.1. Geneva: WHO, 1998. 24 Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 2000; 320: 12401243. 25 Statistics Sweden. Socioekonomisk indelning (SEI). Meddelande i gor 1982:4 (Swedish Socio-Economic Classification). samordningsfra Stockholm: Statistics Sweden, 1983. 26 Karasek RA, Theorell R. Healthy Work: Stress Productivity and Reconstruction of Working Life. New York: Basic Books, 1990. 27 Henderson S, Duncan-Jones P, Byrne DG, Scott R. Measuring social relationships. The Interview Schedule for Social Interaction. Psychol Med 1980; 10: 723734. berg R. Welfare in Transition. A Survey of Living 28 Erikson R, A Conditions in Sweden 19681981. Oxford: Clarendon Press, 1987. 29 Harrell Jr FE, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression modelling strategies for improved prognostic prediction. Stat Med 1984; 3: 143152. 30 Power C, Manor O. Explaining social class differences in psychological health among young adults: a longitudinal perspective. Soc Psychiatry Psychiatr Epidemiol 1992; 27: 284291. 31 Power C, Stansfeld SA, Matthews S, Manor O, Hope S. Childhood and adulthood risk factors for socio-economic differentials in psychological distress: evidence from the 1958 British birth cohort. Soc Sci Med 2002; 55: 19892004. 32 Wax Y. Collinearity diagnosis for a relative risk regression analysis: an application to assessment of dietcancer relationship in epidemiological studies. Stat Med 1992; 11: 12731287. 33 Bostrom G, Diderichsen F. Socioeconomic differentials in misclassification of height, weight and body mass index based on questionnaire data. Int J Epidemiol 1997; 26: 860866. 34 Sobal J, Stunkard AJ. Socioeconomic status and obesity: a review of the literature. Psychol Bull 1989; 105: 260275. 35 Wardle J, Waller J, Jarvis MJ. Sex differences in the association of socioeconomic status with obesity. Am J Public Health 2002; 92: 12991304. , Messing K, Thorbjo rnsson C. Womens Health at Work. 36 Kilbom A Helsingborg: National Institute for Working Life, 1998. 37 Sobal J. Obesity and socioeconomic status: a framework for examining relationships between physical and social variables. Med Anthropol 1991; 13: 231247. 38 van Lenthe FJ, Droomers M, Schrijvers CT, Mackenbach JP. Sociodemographic variables and 6 year change in body mass index: longitudinal results from the GLOBE study. Int J Obes Relat Metab Disord 2000; 24: 10771084. 39 Power C, Manor O, Matthews S. Child to adult socioeconomic conditions and obesity in a national cohort. Int J Obes Relat Metab Disord 2003; 27: 10811086. 40 Statistics, Sweden. Women and Men in Sweden: Facts and Figures. Stockholm: Statistics Sweden, 2002.

Acknowledgements
This study was financed by the Swedish Council for Working Life and Social Research and the National Institute of Public Health, Sweden.

References
1 WHO. Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000; 894: ixii, 1253. 2 James WP. What are the health risks? The medical consequences of obesity and its health risks. Exp Clin Endocrinol Diabetes 1998; 106 (Suppl 2): 16. 3 Jain A. Fighting obesity (Editorial). BMJ 2004; 328: 13271328. 4 Moreau M, Valente F, Mak R, Pelfrene E, de Smet P, De Backer G et al. Obesity, body fat distribution and incidence of sick leave in the Belgian workforce: the Belstress study. Int J Obes Relat Metab Disord 2004; 28: 574582. 5 Rissanen A, Heliovaara M, Knekt P, Reunanen A, Aromaa A, Maatela J. Risk of disability and mortality due to overweight in a Finnish population. BMJ 1990; 301: 835837. 6 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: 10081012. 7 Sargent JD, Blanchflower DG. Obesity and stature in adolescence and earnings in young adulthood. Analysis of a British birth cohort. Arch Pediatr Adolesc Med 1994; 148: 681687. r medicinsk utva rdering. Fetma 8 Asp N-G. Statens beredning fo tga versikt (Obesity problem och a rder: en systematisk litteraturo Problems and Measures: A Systematic Literature Review). Stockholm: SBU, 2002 (in Swedish). 9 Martikainen PT, Marmot MG. Socioeconomic differences in weight gain and determinants and consequences of coronary risk factors. Am J Clin Nutr 1999; 69: 719726. 10 Power C, Moynihan C. Social class and changes in weight-forheight between childhood and early adulthood. Int J Obes Relat Metab Disord 1988; 12: 445453. 11 Rahkonen O, Lundberg O, Lahelma E, Huuhka M. Body mass and social class: a comparison of Finland and Sweden in the 1990s. J Public Health Policy 1998; 19: 88105. 12 Hardy R, Wadsworth M, Kuh D. The influence of childhood weight and socioeconomic status on change in adult body mass index in a British national birth cohort. Int J Obes Relat Metab Disord 2000; 24: 725734. 13 Power C, Parsons T. Nutritional and other influences in childhood as predictors of adult obesity. Proc Nutr Soc 2000; 59: 267272. 14 Persson G. Health in Sweden: The National Public health Report 2001. Oslo: Scandinavian University Press, 2001. 15 Kuh DL, Ben-Shlomo Y. A Life Course Approach to Chronic Disease Epidemiology: Tracing the Origins of Ill Health from Early to Adult Life. Oxford: Oxford University Press, 1997. 16 Kuh D, Hardy R. A Life Course Approach to Womens Health. Oxford: Oxford University Press, 2002.

International Journal of Obesity

Explanations to social inequity in obesity M Novak et al

200
41 van Lenthe FJ, Kemper CG, van Mechelen W. Rapid maturation in adolescence results in greater obesity in adulthood: the Amsterdam Growth and Health Study. Am J Clin Nutr 1996; 64: 1824. 42 van Lenthe FJ, Kemper HC, van Mechelen W, Post GB, Twisk JW, Welten DC et al. Biological maturation and the distribution of subcutaneous fat from adolescence into adulthood: the Amsterdam Growth and Health Study. Int J Obes Relat Metab Disord 1996; 20: 121129. 43 Judith M. Body image, perceived pubertal timing, and adolescent mental health. J Adolesc Health 1999; 25: 155165. 44 Brooks-Gunn J. Antecedents and consequences of variations in girls maturational timing. J Adolesc Health Care 1988; 9: 365373. 45 Post GB, Kemper HC. Nutrient intake and biological maturation during adolescence. The Amsterdam growth and health longitudinal study. Eur J Clin Nutr 1993; 47: 400408. 46 DiPietro L. Physical activity in the prevention of obesity: current evidence and research issues. Med Sci Sports Exerc 1999; 31: S542S546. 47 Sternfeld B, Wang H, Quesenberry Jr CP, Abrams B, Everson-Rose SA, Greendale GA et al. Physical activity and changes in weight and waist circumference in midlife women: findings from the Study of Womens Health Across the Nation. Am J Epidemiol 2004; 160: 912922. 48 Khatun M, Ahlgren C, Hammarstrom A. The influence of factors identified in adolescence and early adulthood on social class inequities of musculoskeletal disorders at age 30: a prospective population-based cohort study. Int J Epidemiol 2004; 33: 13531360. Gillander-Gadin K, Hammarstrom A. School-related health a cross-sectional study among young boys and girls. Int J Health Serv 2000; 30: 797820. Berg Kelly K, Ehrver M, Erneholm T, Gundevall C, Wennerberg I, Wettergren L. Self-reported health status and use of medical care by 3500 adolescents in western Sweden. I. Acta Paediatr Scand 1991; 80: 837843. Hammarstrom A, Janlert U, Theorell T. Youth unemployment and ill health: results from a 2-year follow-up study. Soc Sci Med 1988; 26: 10251033. Jones A. Becoming a girl: post-structuralist suggestions for educational research. Gender Educ 1993; 5: 157166. f T, Pokarzhevskaya G. Da rfo r tja nar ma n mer a n kvinnor. Ljunglo neskillnadernas utvekling bland SACO-fo rbundens medleLo mmar mellan 1996 och 2001 (Reasons that men earn more than women. The development of salary differences among SACOunion members between 1996 and 2001). Sveriges Akademikers Centralorganisation (Swedish Academic Central Organization, SACO) 2003. Epel ES, McEwen B, Seeman T, Matthews K, Castellazzo G, Brownell KD et al. Stress and body shape: stress-induced cortisol secretion is consistently greater among women with central fat. Psychosom Med 2000; 62: 623632.

49

50

51

52 53

54

International Journal of Obesity