The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Objective: To examine the association between relative body weight and health status and the potential modifying effects of socioeconomic position and working conditions on this association.
Research Methods and Procedures: The data were derived from three identical cross-sectional surveys conducted in 2000, 2001, and 2002. Respondents to postal surveys were middle-aged employees of the City of Helsinki (7148 women and 1799 men, response rate 67%). BMI was based on self-reported weight and height. Health status was measured by the Short-Form 36 subscales and component summaries.
Results: Body weight was inversely associated with physical health, but in mental health, differences between BMI categories were small and inconsistent. In women, physical health deteriorated monotonically with increasing BMI, whereas in men, poor physical health was found among the obese only. Socioeconomic position did not modify the association between BMI and health. In women, the association between body weight and physical health became stronger with decreasing job control and increasing physical work load, whereas in men, a similar modifying effect was found for high job demands.
Discussion: Body weight was associated with physical health only. Lower levels of relative weight in women than in men may be associated with poor physical health. High body weight combined with adverse working conditions may impose a double burden on physical health.
Overweight and obesity are rapidly increasing all over the world (1, 2, 3). Excess body weight is known to increase the risk for total mortality and that of several chronic diseases, including type 2 diabetes, coronary heart disease, stroke, and some forms of cancer (4, 5). To comprehensively assess the health consequences of this increasingly prevalent public health concern, the use of generic measures of physical, mental, and social functioning and sense of well-being, often referred to as measures of health-related quality of life, has been supported (6, 7, 8). In obesity research, such generic measures of health were first adopted in clinical weight loss studies. More recently, studies examining the association between body weight and various dimensions of health in non-patient populations have emerged (9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19). These studies have found high body weight to be associated with poor physical health, with greater deterioration of health associated with greater degrees of obesity. In contrast, the association between body weight and mental health seems to be modest. These studies have been made in diverse populations, used varying measures of health status, and most have not analyzed men and women separately (9, 11, 12, 14, 16, 18).
Socioeconomic position (SES)1 is inversely associated with body weight (20, 21, 22), and it could also be a potential modifier of the association between body weight and health status. Excess body weight might have a stronger impact on physical health in the lower SES because in these positions, social roles in work and leisure may require better functional ability and more physical strength. In contrast, the impact of obesity on mental health might be expected to be more adverse in the higher SES. Because slimness is likely to be more highly valued in the higher SES, especially women in these positions have stronger normative pressures for maintaining low body weight. Failure to do so may lead to stigmatization and discrimination, resulting in mental and emotional distress (23, 24, 25, 26).
Measures of SES are often based on employment relations and conditions (27, 28). Working conditions also might be expected to modify the association between body weight and health, although research evidence on this is scarce. Physically strenuous work that is more typical to blue-collar occupations could strengthen the effect of excess body weight on physical health. Also, other characteristics of the working environment, such as heat or dust, might be more disturbing for the overweight and obese. Furthermore, work is one potential source of stress that may provide a disproportionately large burden for mental health among those who are already stigmatized and distressed by their weight.
The aim of this study was, firstly, to examine the association between relative body weight and various dimensions of health among a population consisting of middle-aged employees of the City of Helsinki using the Short-Form 36 (SF-36). The SF-36 is one of the most widely used generic measures of health status today. It comprises measures of the ability to perform daily tasks and activities (functioning) and subjective perceptions of how people feel physically and emotionally and how they think and feel about their health (well-being). Secondly, we aimed to examine how SES and various physical and psychosocial working conditions may affect the association between body weight and physical and mental health. SES and working conditions are intrinsically connected, and both could be expected to modify the association between body weight and health in various ways.
Research Methods and Procedures
The data were derived from middle-aged women and men employed by the City of Helsinki (29). The City of Helsinki has nearly 40,000 employees. The main branches of employment include health care and social services, education and culture, general administration, public transport, and technical and construction branches. Three identical cross-sectional surveys for successive age cohorts were conducted in 2000, 2001, and 2002 for research purposes. Each year, employees who reached the age of 40, 45, 50, 55, or 60 that year received a self-administered questionnaire. The response rate was 69% among women and 60% among men. Twenty-three women who reported that they were pregnant were excluded. The final data included 7148 women and 1799 men, reflecting the fact that the employees of the City of Helsinki are predominantly women. Younger respondents and manual workers were slightly underrepresented among the respondents (30).
BMI was calculated from self-reported weight and height using the standard formula: weight (kilograms) divided by height (meters squared). BMI was first divided into the categories of normal weight (BMI < 25 kg/m2), overweight (25 to 30 kg/m2), and obese (>30 kg/m2) (31). Less than 3% of the respondents were underweight (BMI < 18.5 kg/m2), and they were combined to the normal weight category. Furthermore, in relation to the SF-36 physical component summary (PCS) and mental component summary (MCS), BMI was examined in gender-specific deciles and as a continuous variable.
The SF-36 includes 36 items that assess health status on eight empirically distinct domains: physical functioning, role limitations due to physical problems, bodily pain, general health perceptions, mental health, role limitation due to emotional problems, social functioning, and vitality (32). All domains are scored on subscales ranging from 0 to 100, with higher scores indicating better health. In addition, factor analysis has been used to compress these eight subscales into two component summaries describing physical and mental health, where each subscale positively or negatively contributes to both component summaries (33). The PCS and MCS each have a mean of 50 and an SD of 10 in the U.S. general population, and higher scores again indicate better health. Of the eight subscales, the four first mentioned are often referred to as physical subscales, and the four last mentioned as mental subscales, although vitality, and to a lesser degree social functioning and general health, also have notable correlations with the other component summary. The SF-36 has been shown to have good construct validity, high internal consistency, and high test-retest reliability (32, 33, 34, 35).
SES and Working Conditions
SES was measured by occupational class and divided into four hierarchical groups: managers and professionals, semiprofessionals, routine non-manuals, and manual workers. Manual workers and non-manual employees were separated using the socioeconomic classification of Statistics Finland (36), and non-manual employees were further divided into three groups according to the occupational classification of the City of Helsinki.
Working conditions were measured with Karasek's questions on job control and job demands (37) and with an 18-item inventory of potentially harmful characteristics of work and working environment developed at the Finnish Institute of Occupational Health (38). Job control and job demands are commonly used measures for psychosocial working conditions, and they have been shown to be associated with cardiovascular diseases and various other health outcomes (39). In these data, job demands have shown a moderate association with weight gain (40). Our study included 10 questions on job demands (Cronbach's α 0.75) and nine questions on job control (Cronbach's α 0.76). A summary score was calculated for job demands and job control separately and divided into quartiles.
Factor analysis was used to analyze the 18-item inventory of potentially harmful working conditions. The items had four response alternatives ranging from does not exist to exists and bothers a lot. Factor analysis provided a three-factor solution, explaining 43% of the total variance. The factors were interpreted to describe physical work load (six items, Cronbach's α 0.82), hazardous exposures (nine items, Cronbach's α 0.79), and computer work (three items, Cronbach's α 0.80). The items that loaded on the physical work load factor were uncomfortable postures, repetitive trunk rotation, repetitive movements, standing, walking, and heavy physical work (lifting and carrying). Hazardous exposures factor included items such as exposure to dirt and dust, damp and wetness, noise, solvents and other irritating substances, and problems with temperature or lightning. Computer work, using the computer mouse, and sedentary work loaded on the factor labeled as computer work.
All analyses were conducted separately for women and men using SAS version 8.02 for Windows (SAS Institute Inc., Cary, NC). ANOVA was used to assess differences in the eight SF-36 subscales and the two component summaries among normal-weight, overweight, and obese respondents after adjustment for age. Smoking, alcohol use, marital status, and menopausal status (for women) were also considered as potential confounders, but because they did not influence the results, they were omitted from the final analyses. Differences between the BMI categories were tested using the F test, and 95% confidence intervals (CIs) for the group means were calculated. CIs were calculated also for the physical and mental summary means in BMI deciles. The effect of SES and working conditions on the associations between body weight and SF-36 was examined with multiple regression analysis. The PCS and MCS scores were used as dependent variables in the analyses where BMI was used as a continuous variable and SES and working conditions were added as categorical variables. Pair-wise interactions were checked to examine whether the association of BMI with physical and mental health was similar in all categories of SES and working conditions.
The respondents were ∼50 years of age on average, with the mean BMI slightly above the boundary for overweight in both women (25.3 kg/m2) and men (26.4 kg/m2) (Table 1). Women more often than men belonged to the normal-weight category, but in both genders, ∼14% of the respondents were obese. The routine non-manual group was large in women, whereas in men, the groups of both managers/professionals and manual workers were relatively large. Women reported poorer health than men on most SF-36 subscales. However, the gender differences were rather small, except for the subscales describing role limitations due to physical problems and bodily pain.
Table 1. . Basic characteristics of the study sample and SF-36 scores among women (n = 7.148) and men (n = 1.799)
In women, body weight was inversely associated with all SF-36 subscales pertaining to the physical domains of health (Table 2). All these subscales indicated poorer health with increasing body weight. The PCS also showed an inverse association with body weight. In contrast, the associations of body weight with subscales indicating emotional and social domains of health were modest. There were no differences between BMI categories in the subscales describing mental health or role limitations due to emotional problems, and in social functioning and vitality, there was only a slight decrease among the obese as compared with the normal weight. In contrast, the MCS suggested a slightly better health for the obese compared with the normal weight.
Table 2. . Scores for the SF-36 subscales and summary components by BMI categories (mean and 95% CI) among women and men, adjusted for age
In men, there was a worsening of health with increasing body weight in the physical domains of SF-36, but only the obese differed statistically significantly from the other two BMI categories (Table 2). In the mental subscales, no differences were found between the BMI categories. The component summaries reflected these findings: the PCS indicated poorer health especially for the obese, but no differences between the BMI categories were found in the MCS.
To more closely examine the association of body weight with PCS and MCS, BMI was divided into gender-specific deciles. In women (Figure 1), physical health deteriorated monotonically from the lowest to the highest BMI decile (p < 0.001). In men (Figure 2), there were no differences in physical health in the first five BMI deciles, but in the higher BMI categories, physical health became worse. Thus, in men, the association between body weight and physical health was slightly curvilinear (p = 0.002 for the second order term). In mental health, the differences between BMI deciles were small, but in women, mental health was slightly better among those with higher relative weight.
We next examined the contribution of SES and working conditions on the associations between body weight and SF-36 PCS and MCS. SES and working conditions were mainly inversely but modestly associated with BMI (Table 3). SES was positively associated with the PCS but negatively with the MCS. Working conditions showed positive associations with both component summaries.
Table 3. . Correlations of SES and working conditions with BMI and the SF-36 PCS and MCS among women and men, adjusted for age
* Correlations including SES were Spearman correlations, those including working conditions were Pearson correlations analyzing working conditions as unclassified continuous variables.
−0.15 (p < 0.001)
0.15 (p < 0.001)
−0.08 (p < 0.001)
−0.10 (p < 0.001)
0.15 (p < 0.001)
−0.04 (p = 0.10)
−0.04 (p < 0.001
0.15 (p < 0.001)
0.13 (p < 0.001)
−0.04 (p = 0.10)
0.15 (p < 0.001)
0.16 (p < 0.001)
0.01 (p = 0.33)
0.09 (p < 0.001)
0.21 (p < 0.001)
−0.01 (p = 0.98)
0.10 (p < 0.001)
0.21 (p < 0.001)
Physical work load
−0.07 (p < 0.001)
0.30 (p < 0.001)
0.06 (p < 0.001)
−0.03 (p = 0.23)
0.30 (p < 0.001)
0.11 (p < 0.001)
−0.03 (p = 0.02)
0.15 (p < 0.001)
0.12 (p < 0.001)
−0.09 (p < 0.001)
0.21 (p < 0.001)
0.11 (p < 0.001)
0.05 (p < 0.001)
0.04 (p < 0.001)
0.13 (p < 0.001)
0.04 (p = 0.09)
0.07 (p = 0.004)
0.18 (p < 0.001)
Table 4 shows results from multiple regression analyses examining the change in the association between BMI and the SF-36 PCS and MCS when SES and working conditions were controlled for. BMI was analyzed as a continuous variable. To maintain a meaningful interpretation of the regression coefficient, the second order term for BMI was not included in the model despite the slight curvilinearity observed in the association between BMI and the PCS in men. In women, the score of the PCS decreased by 0.45 for 1 unit increase in BMI. Almost a similar decrease was seen in men. SES and working conditions only slightly weakened these associations. In women, the association between body weight and the MCS was slightly positive, but the association was not statistically significant when SES was controlled for. In men, body weight was not associated with the MCS.
Table 4. . The effect of SES and working conditions on the associations of BMI with PCS and MCS; regression coefficient β and the corresponding p value among women and men
Finally, pair-wise interactions of BMI with SES and working conditions were examined in relation to the PCS and MCS. Three interactions were found for the PCS. BMI interacted with job control (p = 0.02) and physical work load (p < 0.001) in women and with job demands (p = 0.003) in men. No interactions were found for the MCS. Table 5 shows the regression coefficients of BMI separately in the categories of working conditions. In women, the association between body weight and physical health strengthened with decreasing job control or increasing physical work load. In men, the association between body weight and physical health was weaker among those with low job demands.
Table 5. . Regression coefficients for BMI in categories of working conditions that showed statistically significant interactions with BMI in relation to SF-36 PCS; regression coefficient, SE, and the corresponding p value
Job control, women
Physical work, women
Job demands, men
Poor working conditions refer to low job control, heavy physical work load, and high job demands.
This study examines the association between relative body weight and health status and the potential modifying effects of SES and working conditions on this association. The analyses of middle-aged employees of the City of Helsinki provide several important findings.
First, our study provides renewed empirical support for a clear inverse association between body weight and physical health in non-patient populations. All SF-36 physical subscales and the PCS uniformly showed that body weight is associated with poor health. In women, we found a step-wise deterioration of health with increasing body weight, whereas in men, poor physical health was found among the obese only.
In contrast, for the mental domains of health, differences between BMI categories were small and inconsistent. Also, previous studies have found no or only small differences in mental health by body weight, although mental health problems could be expected because of impaired physical health (41) and the stigma associated with excess body weight (42). All previous studies using the SF-36 show more consistent association of body weight with physical than with mental health (10, 11, 12, 15, 17). In our study, social functioning and vitality were slightly poorer among the obese than among the normal weight, but for the MCS, there was a slight improvement of health with increasing BMI in women. The reason for this discrepancy may be that the MCS is factored as a complement to the PCS; therefore, mental health problems that are induced by poor physical health are artificially detached from the MCS. Previous studies have found more mental health problems among the obese with comorbidities (12, 18).
Second, our study suggests a potential gender difference in the association between body weight and physical health. For all physical subscales, overweight women reported poorer health than the normal weight, but in men, poor physical health was more clearly associated with obesity. A gender difference was also apparent when BMI was divided into deciles; in women, physical health deteriorated steadily with increasing body weight also among the normal weight, although the difference between women in the upper and lower ends of the normal-weight range was only moderate. In men, physical health deteriorated only among those above the BMI of 27. Two previous studies have reported a gender difference in the association between body weight and the physical functioning subscale of the SF-36. In the United Kingdom, Stafford et al. (43) found a steady deterioration of physical functioning in women but a threshold effect of 27 kg/m2, i.e., similar to ours, in men. In the U.S., Yan et al. (19) found that the physical functioning subscale of the SF-36 was associated with obesity in both genders but with overweight in women only. Furthermore, among the Spanish elderly, López-García et al. (17) found the best health for all SF-36 subscales among overweight men, whereas in women, for most subscales, the normal weight reported best health. Together, these findings suggest that lower levels of relative weight in women than in men may be associated with poor physical health.
Third, the effect of SES on the association between BMI and health was negligible, and no interactions between SES and BMI were found. We assumed that BMI might have a stronger association with physical health in the lower occupational classes but with mental health in the higher occupational classes. The latter assumption not being supported is not surprising, given that overall the differences in mental health by body weight were modest. In our middle-aged respondents, body image is likely to have less influence on self-esteem and peer pressures than in younger age groups. Furthermore, as the prevalence of overweight and obesity increase in the population, previously deviant weight may be seen as more normal, weakening the stigma attached to excess weight (17).
Fourth, working conditions had some effects on the association between body weight and physical health. In women, the association between excess body weight and physical health became stronger with decreasing job control and increasing physical work load, although the effect was not large. In men, a similar but less consistent modifying effect was found for high job demands. High job demands, heavy physical work, and low job control were all themselves associated with poor physical health. Thus, excess body weight affected physical health most in those categories of working conditions that were already associated with poor physical health, strengthening the effect of adverse working conditions on physical health.
Our study has some limitations. Because the data were derived from cross-sectional surveys, strict causal judgments should be avoided. The sample consisted of middle-aged employees, the variation in body weight may be truncated from both ends of the BMI range. On the one hand, the number of respondents with low BMI was rather small because of the age of the respondents. On the other hand, the seriously obese also may be underrepresented because such respondents may have been selected out of the workforce and, thus, from our study sample. Similarly, those with poor health may have been selected out of the data due to the healthy worker effect (44). If there is selection of the data due to both excess body weight and poor health, the associations observed are likely to be underestimated. Another limitation of the study is that body weight was based on self-reports. People are known to underestimate their weight, which may lead to misclassification for overweight and obesity (45, 46). The underestimation of weight may be more pronounced among the heavier people, women, older people, and higher socioeconomic groups. However, such underestimation has to be relatively large to alter the main findings, and in population-level studies, self-reported weight is usually considered an acceptable indicator of body weight.
In conclusion, body weight was consistently associated with physical but not with mental health. In women, physical health deteriorated gradually with increasing body weight, whereas in men, impaired physical health was clearly associated with obesity only. SES had a negligible effect on the association between body weight and health, thus reinforcing the view that the association is valid for different population subgroups. Instead, we found some evidence on the influence of working conditions on the association between body weight and physical health. Excess body weight together with adverse physical or psychosocial working conditions may impose a double burden on physical health.
The Helsinki Health Study was supported by the Academy of Finland (Grants 48118, 53245, and 1210191) and by the Finnish Work Environment Fund (Grant 99090). Additional grants were provided by the Academy of Finland (Grant 204894 to M.L. and Grants 48600 and 210752 to P.M.).