Associations of job strain and working overtime with adverse health behaviors and obesity: Evidence from the Whitehall II Study, Helsinki Health Study, and the Japanese Civil Servants Study

Social Science & Medicine 66 (2008) 1681e1698 www.elsevier.com/locate/socscimed

Associations of job strain and working overtime with adverse health behaviors and obesity: Evidence from the Whitehall II Study, Helsinki Health Study, and the Japanese Civil Servants Study Tea Lallukka a,*, Eero Lahelma a, Ossi Rahkonen a, Eva Roos a,b, Elina Laaksonen a, Pekka Martikainen c, Jenny Head d, Eric Brunner d, Annhild Mosdol d, Michael Marmot d, Michikazu Sekine e, Ali Nasermoaddeli e, Sadanobu Kagamimori e a

Department of Public Health, P.O. Box 41 (Mannerheimintie 172), 00014 University of Helsinki, Finland b Folkha¨lsan Research Center, Helsinki, Finland c Population Research Unit, Department of Sociology, University of Helsinki, Helsinki, Finland d Department of Epidemiology & Public Health, University College London, London, UK e Department of Welfare Promotion and Epidemiology, University of Toyama, Toyama, Japan Available online 8 February 2008

Abstract Adverse health behaviors and obesity are key determinants of major chronic diseases. Evidence on work-related determinants of these behavioral risk factors is inconclusive, and comparative studies are especially lacking. We aimed to examine the associations between job strain, working overtime, adverse health behaviors, and obesity among 45e60-year-old white-collar employees of the Whitehall II Study from London (n ¼ 3397), Helsinki Health Study (n ¼ 6070), and the Japanese Civil Servants Study (n ¼ 2213). Comparable data from all three cohorts were pooled, and logistic regression analysis was used, stratified by cohort and sex. Models were adjusted for age, occupational class, and marital status. Outcomes were unhealthy food habits, physical inactivity, heavy drinking, smoking, and obesity. In London, men reporting passive work were more likely to be physically inactive. A similar association was repeated among women in Helsinki. Additionally, high job strain was associated with physical inactivity among men in London and women in Helsinki. In London, women reporting passive work were less likely to be heavy drinkers and smokers. In Japan, men working overtime reported less smoking, whereas those with high job strain were more likely to smoke. Among men in Helsinki the association between working overtime and non-smoking was also suggested, but it reached statistical significance in the ageadjusted model only. Obesity was associated with working overtime among women in London. In conclusion, job strain and

* Corresponding author. Tel.: þ358 919127566; fax: þ358 919127570. E-mail address: [email protected] (T. Lallukka). 0277-9536/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2007.12.027

1682

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

working overtime had some, albeit mostly weak and inconsistent, associations with adverse health behaviors and obesity in these middle-aged white-collar employee cohorts from Britain, Finland, and Japan. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: Psychosocial; Unhealthy food habits; Physical inactivity; Heavy drinking; Current smoking; International comparisons; UK; Finland; Japan; Work

Introduction Psychosocial job strain (Karasek, 1979) has been shown to be a major, independent cardiovascular risk factor (Belkic, Landsbergis, Schnall, & Baker, 2004). Health behaviors are also important determinants of these largely preventable chronic diseases (Ezzati, Henley, Thun, & Lopez, 2005; Hu et al., 2005). Working overtime is also related to health behaviors and health (Caruso, Hitchcock, Dick, Russo, & Schmit, 2004). However, associations of these working conditions with adverse health behaviors have not been thoroughly studied. Nevertheless, it has been suggested, that psychosocial factors including working conditions might shape health behaviors (Stansfeld & Marmot, 2002), and may partly explain inequalities in metabolic syndrome and cardiovascular disease by affecting employees’ health behaviors (Chandola, Brunner, & Marmot, 2006; Marmot & Theorell, 1988). This highlights the importance of deepening our understanding of how working conditions might relate to adverse health behaviors and subsequent obesity. International comparisons provide a wider perspective to study these associations, while also seeking to identify similarities or dissimilarities across national contexts thereby showing more valid results. A pathway between psychosocial working conditions and adverse health behaviors has been suggested as a response to environmental challenges such as high job strain and overtime work that may culminate in behavioral modification (Bhui, 2002). Accordingly, it can be hypothesized that employees compensate for high psychosocial job strain and working overtime with adverse behaviors. First, job strain may increase the consumption of fatty and sweet foods (Oliver, Wardle, & Gibson, 2000), while intake of fruits, vegetables, fish and meat may be reduced (Oliver & Wardle, 1999), at least among susceptible individuals (Oliver et al., 2000; Wardle, Steptoe, Oliver, & Lipsey, 2000). High job strain and overtime work are also potential barriers to physical activity (Schneider & Becker, 2005). Additionally, chronic job strain could increase heavy drinking, prevent smokers from quitting, or induce quitters to relapse. Since smoking is assumed to ease stress,

smokers are likely to smoke more under psychosocially strenuous work (Parrott, 1999). However, as smoking is initiated usually already in young adulthood (Paavola, Vartiainen, & Haukkala, 2004), the relationship between psychosocial working conditions and current smoking is likely to rather reflect smoking intensity or maintenance of the harmful habit (Green & Johnson, 1990; Johansson, Johnson, & Hall, 1991). While strenuous work is hypothesized to increase drinking, heavy drinking may in turn affect both perceptions about psychosocial working conditions and the actual work (Cargiulo, 2007; Zins et al., 1999). Additionally, suggested behavioral modifications might predispose employees to subsequent obesity. These hypotheses need, however, further clarification and examination. As a large number of employed people are continuously exposed to psychosocially strenuous working conditions for a substantial part of their active time, it is important to examine whether this has detrimental effects on health behaviors and obesity across different countries and among both genders. In line with these suppositions about assumed theoretical pathways, earlier research has shown that job strain and its dimensions, i.e., job demands and job control have some, albeit limited and inconclusive, associations with health behaviors and body mass index (Hellerstedt & Jeffery, 1997; Kouvonen, Kivima¨ki, Va¨a¨na¨nen, et al., 2007). Most previous studies have focused only on smoking (Green & Johnson, 1990; Kouvonen, Kivima¨ki, Virtanen, Pentti, & Vahtera, 2005) or heavy drinking (Bobak et al., 2005; Kouvonen, Kivima¨ki, Cox, Poikolainen, et al., 2005; van Loon et al., 2004). Physical activity has also received some attention (Kouvonen, Kivima¨ki, Elovainio, et al., 2005), but studies about food habits are especially scarce (Kawakami et al., 2006). Working overtime may, in turn, influence employees’ health by causing strain or changing health behaviors (Caruso et al., 2004). However, we lack evidence linking working arrangements with several health behaviors. Moreover, these previous results are based on non-comparative studies, with usually a limited number of working conditions and health behaviors. Instead, previous international comparisons have focused on different outcomes,

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

such as self-rated health, while examining working conditions and health behaviors as determinants (Singh-Manoux et al., 2006). Thus, we sought to widen and corroborate the current understanding of the associations between psychosocial working conditions, adverse health behaviors and obesity by pooling comparable employee cohorts from Britain, Finland and Japan. More specifically, by conducting this international comparison, we aimed to identify both common associations across countries and cohort-specific variations. When undertaking the present study design, some clarifications about the cohorts and justification for the comparisons need to be pointed out. In general, it should be borne in mind that the included countries have different welfare regimes (Dahl et al., 2006). First, despite many similarities as Western European countries, Finland and Britain, nevertheless, differ with regard to labor markets, welfare provision, income inequalities, as well as social and family structures (Esping-Andersen, 1990; Smeeding & Gottschalk, 1999), which in turn may contribute to health behaviors. For instance, in Finland the government and the labor unions intervene with the labor market conditions which may possibly lead to smaller differences in working conditions. This may affect the associations of working conditions with health behaviors as well. Secondly, concerning Japan, more secure possibilities for a lifelong job as well as better pay and promotion prospects with increasing age may also affect health behaviors and thus distinguish Japanese employees from their Western European counterparts. While social, cultural, and worksite contexts are likely to add to the complexity of the associations, as has been speculated previously (Hellerstedt & Jeffery, 1997), the socio-cultural diversity also provides fruitful contrasts for comparisons (Markus & Kitayama, 2003). Furthermore, clear counterpoints between Japanese distant cultural features and Western European countries (Kitayama, Snibbe, Markus, & Suzuki, 2004; Markus & Kitayama, 2003) enable us to produce a more comprehensive picture on the associations between psychosocial working conditions and behavioral risk factors. Since Japan is the only non-Western, highly industrialized and educationally advantaged nation, its usefulness in comparative studies has been previously acknowledged as well (Inaba et al., 2005). Thus, by comparing British and Finnish employees with their Japanese counterparts, our key aim is to be able to assess the significance of job strain and working overtime to adverse behaviors and obesity, which might mediate the previously well-established relationship between job

1683

characteristics and (cardiovascular) health (Belkic et al., 2004). Methods Data We analyzed pooled cross-sectional data of employees from three countries: Britain, Finland and Japan. The data collection in Finland and Japan largely followed the protocol of the British Whitehall II Study (London). The data from these countries were collected separately, and harmonized for pooling and comparability purposes. In general, all the items were practically identical in the three cohorts except that information about physical activity was more exhaustive in Britain than in the two other countries. Thus, the levels of physical activity are not directly comparable, but the associations between working conditions and physical inactivity are unlikely to be strongly affected. In each country, the employees were contacted at their work-place by sending a postal questionnaire. They were given key information about the study, including voluntary participation and independence of the studies from the employers. In the Whitehall II Study, people invited at baseline were informed about the goal of ‘‘identifying the characteristics of work and personal environment which may adversely or beneficially affect people’s health.’’ They were invited to a medical screening and asked to complete a questionnaire. Main goal is to examine what underlies the social gradient in death and disease among both women and men. Similarly, at phase 5, participants were also invited to a medical check-up and asked to bring a completed questionnaire with them. Correspondingly, in the Helsinki Health Study the potential participants were informed about the general goal of examining ‘‘how work and other living environment and health behaviors/lifestyle together with biological factors contribute to health and well-being’’. The goals of the Japanese Civil Servants Study presented in the cover letter were, in brief, as follows: ‘‘The purpose of this questionnaire survey is to evaluate work stress and health among employees and the data will be used for the purposes of health management and disease prevention.’’ The cover letters in any of the countries did not focus on health behaviors. For the present comparative study, we included employees of similar ages and positioning in whitecollar jobs. Further information is available at the Whitehall II Study and the Helsinki Health Study websites (http://www.ucl.ac.uk/whitehallII, http://www. kttl.helsinki.fi/hhs). Concerning the Japanese Civil

1684

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

Servants Study, further information is available in a previous report (Sekine, Chandola, Martikainen, Marmot, & Kagamimori, 2006). Britain, Whitehall II Study, London The Whitehall II Study is a cohort of 10,308 men and women working in 20 civil service departments in London at the time of recruitment (1985e1988) (Marmot & Brunner, 2005). The initial response rate was 73%. After the clinical examination, postal questionnaires were administered at 2.5 yearly intervals and three further clinical examinations were also conducted. We used questionnaire data from phase 5 collected in 1997e1999 (n ¼ 7830). Employees aged 45e60 years, who were still working in civil service were included (n ¼ 3397), in order to get data comparable to the other two cohorts in terms of age and working status. Consequently, participants over 60 years old (26%), as well as those retired, not in civil service or out of work were excluded (55%). The majority of the participants were men (72%), and the cohort does not include manual workers. Finland, Helsinki Health Study Postal questionnaire data for the Helsinki Health Study were collected in 2000e2002 among middleaged employees of the City of Helsinki (n ¼ 8960, response rate 67%) (Lahelma, Martikainen, Rahkonen, Roos, & Saastamoinen, 2005). Nearly 80% of the participants were women, corresponding to staff of the City of Helsinki. For comparisons, white-collar employees aged 45e60 years were included (n ¼ 6070). Japan, Civil Servants Study The Japanese data were derived from a postal survey (n ¼ 2213, response rate 88%) among civil servants of a local government on the west coast of Japan in 1998e1999 (Kagamimori, Sekine, Nasermoaddeli, & Hamanisi, 2002). The Whitehall II Study questionnaire items were translated into Japanese for the survey, and back-translated to English. The back-translated items reported in this study (Appendix 1) correspond exactly to the original ones. All employees were invited to the study. For the present study, white-collar employees aged 45e60 years were included as in Helsinki. Similarly to London, the majority of the participants were men (71%). Measures Outcomes Outcome variables were comprised of unhealthy food habits, physical inactivity, heavy drinking, current

smoking, and obesity. First, healthy food habits were defined as consumption of fruit and vegetables at least twice a day, and wholegrain bread and low-fat milk as typical choices in place of white bread and whole milk. Those with no such healthy food habits were classified as having unhealthy food habits in this study. Proxy measures using only a few indicators of healthy food habits have been proved to sufficiently reflect socio-demographic and socio-economic differences in food habits, and thus serve as an applicable method to examine adherence to general dietary guidelines (Dynesen, Haraldsdottir, Holm, & Astrup, 2003). The Japanese data did not include information on food habits. Physical inactivity was measured by reported amount and intensity of leisure time physical activity. In London, physical activity was calculated from responses to questions about mild, moderate, and vigorous leisure time physical activities. The physically inactive category comprised of employees who reported 1 h or less of such physical activities per week. Broadly similar criteria within these data have also been previously applied (Brunner, Shipley, Blane, Davey Smith, & Marmot, 1999; Ferrie, Martikainen, Shipley, & Marmot, 2005). For Helsinki and Japan, responses to questions about mild, moderate, and vigorous physical activities were also summed up, and the sum was dichotomized to indicate physical inactivity. The lowest quintile of the distribution of the sum was used as cutoff point reflecting a similar proportion of physically inactive employees as in London. Heavy drinking was measured as the consumption of cut-off point more than 280 g of pure alcohol per week for men and 140 g per week for women. Correspondingly, gender-specific cut-off points have been used previously (Kouvonen, Kivima¨ki, Cox, Poikolainen, et al., 2005; Romelsjo¨ et al., 1992). The consumption was estimated based on reported units of different alcohol types consumed during the previous week (London and Japan) or average weekly consumption (Helsinki). The alcoholic content of each reported drink was multiplied by the number of units consumed and the total weekly consumption was then examined. Alcoholic beverages in Japan are somewhat different from those in Britain and Finland with drinks such as sake and shochu being typical choices. However, as we examined the consumption of absolute alcohol in grams, the type of the beverage does not distort the findings. Smoking was dichotomized as current smokers and non-smokers, including ex-smokers. In London and Helsinki, the items about smoking included cigarettes, cigars or pipes, whereas in Japan the questionnaire only mentioned cigarettes. This is unlikely to affect

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

1685

Japanese adults (Yang, Shiwaku, Nabika, Masuda, & Kobayashi, 2007).

the results, since smoking cigars or pipes is very rare in Japan. The number of women smoking in Japan was very low (Table 1). Consequently, the results need to be interpreted with caution as the 95% confidence intervals are wide. Obesity was defined as body mass index (BMI) of at least 30. BMI was computed from self-reported height and weight in Helsinki and from measured height and weight in London and Japan (weight/height  height) weight/height2. Self-reported measures may somewhat under-report obesity (Visscher, Viet, Kroesbergen, & Seidell, 2006). In the Japanese data, only about 1% of participants had a BMI over 30. Therefore, those with a BMI of 25 or more were classified as obese according to the definition of the Japan Society for the Study of Obesity. This lower cut-off point has been applied in previous Japanese studies as well (Tsutsumi, Kayaba, Hirokawa, Ishikawa, & The Jichi Medical School Cohort Study Group, 2006). Moreover, recent findings testing the criteria suggested by the Regional Office for the Western Pacific Region of WHO indicate that 25 is a more suitable cut-off point for obesity for

Independent variables Self-reported psychosocial working conditions were modeled using Karasek’s job strain (Karasek, 1979), and working overtime. We included four similar items on job demands and eight on job control in all the countries. These items are listed and compared in Appendix 1. The median in the distribution of the sum score was used as a cut-off point for high job demands and high job control, separately by cohort and gender. The interaction of these dimensions was then examined. According to Karasek’s model (Karasek, Baker, Marxer, Ahlbom, & Theorell, 1981), a low-strain category combines low job demands with high job control, whereas the category of high-strain job is the opposite situation with high job demands and low job control to meet these demands. Additionally, a combination of low job demands and low job control is considered to be a passive work environment, while high job demands

Table 1 Basic characteristics of participants London (%) Women

Helsinki (%) Men

Women

Japanese province (%) Men

Women

Men

Age (mean, years) 45e49 50e54 55e60

52.4 32 38 29

51.7 36 44 20

51.6 28 27 45

52.4 24 24 52

50.9 42 34 24

51.3 41 33 26

Socio-economic position Routine non-manual/clerical/support Professional/semi-professional/executive Managers/administrative

31 45 24

6 44 50

49 43 8

12 61 26

35 64 2

28 54 18

Married or cohabiting

67

85

67

81

87

95

Job strain Low strain Active Passive High strain

32 27 30 11

29 27 30 14

28 23 28 21

25 26 26 23

27 28 33 13

34 25 25 16

Working overtime

43

60

14

21

55

49

Unhealthy food habits

12

13

10

21

e

e

Physical inactivity

21

23

20

21

18

22

Heavy drinking

13

13

7

8

4

23

Current smoking

13

17

21

25

4

46

Obesity (BMI  30; in Japan  25)

19

13

15

15

21

27

967

2430

4984

1086

650

1563

N (total)

Prevalence (%) of adverse health behaviors, obesity and working conditions in the Whitehall II Study (London), Helsinki Health Study, and the Japanese Civil Servants Study (data derived from a local government on the west coast of Japan).

1686

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

coupled with high control is assumed to be an active situation with good learning opportunities. The lowstrain category was used as a reference group in all the analyses. The high-strain job situation exposing to work stress is suggested to have the most detrimental health effects by increasing the risk of cardiovascular diseases (Belkic et al., 2004) and affecting both mental (Tsutsumi, Kayaba, Theorell, & Siegrist, 2001) and physical health (Sekine et al., 2006). Thus, when focusing on adverse health behaviors and obesity, it is possible to shed light on their potential mediating role. Accordingly, working conditions are likely to first influence health behaviors and subsequently generate increased risk for chronic diseases (Siegrist & Ro¨del, 2006; Stansfeld & Marmot, 2002). Working overtime was dichotomized in all countries based on reporting working 40 h or less, or 41 h or more per week. In Japan, weekly working hours were computed from reported length of a normal working day multiplied by five. It is possible that this may slightly underestimate the length of an average week for some participants. In Britain and Japan, the participants were asked to report time spent at the main job including work brought home, while in Finland time spent working in paid employment was more generally enquired. This was assumed to include all working time. Covariates We used age in all the models as a covariate. Occupational class was used as an indicator for socioeconomic position. It was categorized into three comparable, hierarchical groups, judged by current occupation. The lowest group consisted mostly of routine, non-manual workers in Helsinki and clerical and support staff in London and Japan. Executives, professionals and semi-professionals were included in the intermediate group, whereas managers in Helsinki and administrative employees in London and Japan formed the highest category. Marital status was dichotomized as married or cohabiting and single, divorced, or widowed. Missing values In Helsinki, the number of missing values was generally low, 2e3%, whereas in London, it was around 10%. In Japan, the proportion missing was equal or higher to that of London. Missing items in Japan were partly due to data collection in four separate postal questionnaires with somewhat varying responses, i.e., every employee did not return all the four questionnaires. With regard to all the outcomes, those with missing values were excluded from the analyses.

Concerning psychosocial working conditions, for those missing one item out of four items for job demands or two items out of eight for job control at the most, missing values were included in the analyses and replaced by mode to avoid unnecessary loss of data in the multivariate logistic regression models. This method was applied to 1e3% of participants only suggesting that the selected imputation method does not bias the findings. Missing values for working overtime were excluded from the analyses. We also conducted control analyses including missing values for job demands, job control, and working overtime as separate categories in these variables. The results were unaffected (data not shown). Thus, we preferred to examine only those with no missing data about these working conditions. Statistical methods We first computed distributions (%) of outcomes and independent variables in London, Helsinki and Japan (Table 1). Logistic regression models were used to examine the associations between the outcomes and independent variables (Tables 2e6). Model 1 shows the age-adjusted individual effects of the examined working conditions, whereas in Model 2 also occupational class and marital status are adjusted. In Model 3, all the working conditions and covariates are simultaneously adjusted for. All the analyses were carried out using an SAS statistical program, version 8.2. Results Descriptive statistics The mean age in London and Helsinki was 52 years, and in Japan 51 years in both genders (Table 1). In each country, more men than women were managers or administrators. More men than women were married in all the countries. High job strain was more prevalent in Helsinki than London and Japan, whereas working overtime was more common in London and Japan than Helsinki. The prevalence of the outcome variables differed substantially by country and gender. Unhealthy food habits were found among a tenth of women and men in London and women in Helsinki, while 21% of men in Helsinki had unhealthy food habits. Heavy drinking was found among 13% of women and men in London, whereas only among 7% of women and 8% of men in Helsinki. In Japan, a fourth of men, but very few women reported heavy drinking. Likewise, the smoking prevalence differed greatly by country and gender. In

Table 2 Associations between working conditions and unhealthy food habitsa: logistic regression analyses among employees of the Whitehall II Study (London), and Helsinki Health Study Men

Model 1 OR London Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) Helsinki Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week)

Model 2 CI 95%

OR

Model 3 CI 95%

OR

Model 1 CI 95%

OR

n ¼ 777

Model 2 CI 95%

Model 3

OR

CI 95%

OR

CI 95%

n ¼ 2130

1.00 0.78 0.96 0.80

(ref) 0.43 0.55 0.36

1.40 1.67 1.75

1.00 0.90 0.73 0.70

(ref) 0.49 0.40 0.31

1.67 1.34 1.56

1.00 0.90 0.73 0.70

(ref) 0.48 0.40 0.31

1.68 1.34 1.56

1.00 1.08 1.67 1.09

(ref) 0.75 1.20 0.70

1.55 2.33 1.69

1.00 1.20 1.29 0.94

(ref) 0.83 0.91 0.61

1.74 1.84 1.47

1.00 1.26 1.25 0.96

(ref) 0.87 0.88 0.61

1.83 1.78 1.49

1.00 0.83

(ref) 0.52

1.30

1.00 1.00

(ref) 0.61

1.64

1.00 1.00

(ref) 0.61

1.64

1.00 0.65

(ref) 0.51

0.84

1.00 0.78

(ref) 0.60

1.02

1.00 0.78

(ref) 0.59

1.04

n ¼ 4724

n ¼ 1025

1.00 1.01 1.22 1.22

(ref) 0.76 0.94 0.92

1.34 1.59 1.62

1.00 1.03 1.17 1.18

(ref) 0.77 0.89 0.89

1.37 1.54 1.57

1.00 1.02 1.17 1.18

(ref) 0.76 0.89 0.89

1.38 1.54 1.57

1.00 1.30 1.08 0.78

(ref) 0.86 0.71 0.49

1.96 1.65 1.24

1.00 1.29 1.06 0.78

(ref) 0.85 0.69 0.49

1.95 1.63 1.24

1.00 1.22 1.08 0.76

(ref) 0.80 0.70 0.48

1.86 1.66 1.21

1.00 0.99

(ref) 0.75

1.31

1.00 1.01

(ref) 0.76

1.35

1.00 1.03

(ref) 0.78

1.38

1.00 1.32

(ref) 0.92

1.88

1.00 1.33

(ref) 0.92

1.90

1.00 1.30

(ref) 0.89

1.89

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

Women

a

No item of three healthy items, i.e., not eating fresh vegetables or fruit at least twice a day, not choosing wholegrain bread, and not drinking low-fat milk; Model 1: Age-adjusted; Model 2: Age, occupational class, and marital status adjusted for; Model 3: All working conditions and covariates mutually adjusted for.

1687

1688 Table 3 Associations between working conditions and physical inactivitya: logistic regression analyses among employees of the Whitehall II Study (London), Helsinki Health Study, and the Japanese Civil Servants Study Women

Men

Model 1 OR

Working overtime No Yes (over 40 h per week) Helsinki Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) Japanese province Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) a

CI 95%

OR

Model 3 CI 95%

OR

Model 1 CI 95%

OR

n ¼ 782

Model 2 CI 95%

Model 3

OR

CI 95%

OR

CI 95%

n ¼ 2135

1.00 0.88 1.49 0.77

(ref) 0.47 0.85 0.32

1.65 2.61 1.85

1.00 1.29 0.83 0.57

(ref) 0.66 0.45 0.23

2.54 1.54 1.40

1.00 1.33 0.83 0.57

(ref) 0.67 0.45 0.23

2.64 1.53 1.42

1.00 1.16 1.82 1.97

(ref) 0.83 1.34 1.36

1.63 2.48 2.83

1.00 1.19 1.65 1.88

(ref) 0.85 1.19 1.30

1.68 2.28 2.72

1.00 1.20 1.64 1.88

(ref) 0.85 1.18 1.30

1.69 2.27 2.72

1.00 0.59

(ref) 0.36

0.95

1.00 0.83

(ref) 0.49

1.41

1.00 0.81

(ref) 0.47

1.38

1.00 0.84

(ref) 0.67

1.06

1.00 0.93

(ref) 0.73

1.19

1.00 0.97

(ref) 0.75

1.24

n ¼ 4089

n ¼ 857

1.00 0.99 1.29 1.25

(ref) 0.79 1.05 1.00

1.24 1.58 1.56

1.00 0.96 1.30 1.25

(ref) 0.77 1.05 1.00

1.21 1.61 1.57

1.00 0.96 1.31 1.25

(ref) 0.76 1.05 1.00

1.20 1.62 1.57

1.00 1.12 1.22 1.48

(ref) 0.69 0.75 0.92

1.82 1.97 2.39

1.00 1.14 1.18 1.47

(ref) 0.70 0.72 0.91

1.85 1.92 2.38

1.00 1.11 1.18 1.44

(ref) 0.67 0.73 0.89

1.81 1.93 2.35

1.00 1.01

(ref) 0.81

1.26

1.00 1.00

(ref) 0.80

1.25

1.00 1.05

(ref) 0.83

1.31

1.00 1.11

(ref) 0.74

1.66

1.00 1.16

(ref) 0.77

1.75

1.00 1.14

(ref) 0.75

1.74

n ¼ 302

n ¼ 785

1.00 1.84 1.05 1.19

(ref) 0.81 0.46 0.40

4.19 2.42 3.54

1.00 1.82 0.88 1.09

(ref) 0.78 0.36 0.36

4.24 2.16 3.28

1.00 1.83 0.87 1.09

(ref) 0.78 0.35 0.36

4.28 2.16 3.29

1.00 1.08 1.43 1.25

(ref) 0.68 0.91 0.74

1.72 2.24 2.09

1.00 1.07 1.34 1.17

(ref) 0.67 0.85 0.69

1.72 2.12 1.96

1.00 1.08 1.34 1.17

(ref) 0.67 0.85 0.69

1.74 2.12 1.98

1.00 1.04

(ref) 0.57

1.91

1.00 1.05

(ref) 0.55

2.00

1.00 0.95

(ref) 0.49

1.84

1.00 0.94

(ref) 0.67

1.33

1.00 0.96

(ref) 0.68

1.36

1.00 0.98

(ref) 0.68

1.41

Physical inactivity (lowest active quintile); Model 1: Age-adjusted; Model 2: Age, occupational class, and marital status adjusted for; Model 3: All working conditions and covariates mutually adjusted for.

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

London Job strain Low strain Active Passive High strain

Model 2

Table 4 Associations between working conditions and heavy drinkinga: logistic regression analyses among employees of the Whitehall II Study (London), Helsinki Health Study, and the Japanese Civil Servants Study Women

Men

Model 1 OR

Working overtime No Yes (over 40 h per week) Helsinki Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) Japanese province Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week)

CI 95%

OR

Model 3 CI 95%

OR

Model 1 CI 95%

OR

n ¼ 766

Model 2 CI 95%

Model 3

OR

CI 95%

OR

CI 95%

n ¼ 2115

1.00 0.98 0.25 0.67

(ref) 0.61 0.13 0.34

1.56 0.49 1.33

1.00 0.73 0.38 0.86

(ref) 0.45 0.19 0.42

1.20 0.78 1.74

1.00 0.74 0.38 0.86

(ref) 0.45 0.19 0.42

1.22 0.78 1.75

1.00 0.93 0.87 1.00

(ref) 0.66 0.62 0.67

1.30 1.21 1.49

1.00 0.92 0.85 0.98

(ref) 0.65 0.60 0.65

1.30 1.22 1.47

1.00 0.90 0.87 0.97

(ref) 0.63 0.61 0.64

1.27 1.25 1.46

1.00 1.41

(ref) 0.94

2.11

1.00 0.95

(ref) 0.61

1.48

1.00 0.95

(ref) 0.61

1.50

1.00 1.15

(ref) 0.88

1.50

1.00 1.15

(ref) 0.87

1.53

1.00 1.15

(ref) 0.86

1.53

n ¼ 4698

n ¼ 1022

1.00 0.89 0.77 0.77

(ref) 0.66 0.57 0.55

1.21 1.04 1.07

1.00 0.77 0.93 0.85

(ref) 0.56 0.68 0.61

1.05 1.28 1.19

1.00 0.77 0.93 0.85

(ref) 0.56 0.68 0.61

1.06 1.28 1.19

1.00 0.74 0.94 0.58

(ref) 0.39 0.52 0.29

1.39 1.71 1.16

1.00 0.68 1.04 0.60

(ref) 0.36 0.57 0.29

1.28 1.90 1.21

1.00 0.60 1.08 0.56

(ref) 0.31 0.59 0.27

1.16 1.98 1.15

1.00 1.08

(ref) 0.79

1.48

1.00 0.94

(ref) 0.68

1.30

1.00 0.98

(ref) 0.70

1.35

1.00 1.53

(ref) 0.91

2.57

1.00 1.38

(ref) 0.82

2.34

1.00 1.64

(ref) 0.94

2.86

n ¼ 298

n ¼ 778

1.00 1.07 0.52 0.45

(ref) 0.27 0.11 0.05

4.20 2.42 4.22

1.00 1.01 0.35 0.37

(ref) 0.25 0.06 0.04

4.15 1.89 3.57

1.00 1.04 0.33 0.37

(ref) 0.25 0.06 0.04

4.29 1.84 3.58

1.00 0.90 0.79 0.69

(ref) 0.58 0.51 0.41

1.38 1.23 1.18

1.00 0.92 0.80 0.72

(ref) 0.60 0.51 0.42

1.41 1.26 1.23

1.00 0.96 0.79 0.74

(ref) 0.62 0.50 0.43

1.49 1.24 1.28

1.00 0.92

(ref) 0.30

2.83

1.00 0.91

(ref) 0.27

3.01

1.00 0.77

(ref) 0.23

2.64

1.00 0.84

(ref) 0.60

1.19

1.00 0.84

(ref) 0.60

1.18

1.00 0.83

(ref) 0.59

1.19

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

London Job strain Low strain Active Passive High strain

Model 2

a

Alcohol consumption over 280 g per week for men and over 140 g per week for women; Model: 1 Age-adjusted; Model: 2 Age, occupational class, and marital status adjusted for; Model: 3 All working conditions and covariates mutually adjusted for. 1689

1690 Table 5 Associations between working conditions and smokinga: logistic regression analyses among public sector employees of the Whitehall II Study (London), Helsinki Health Study, and the Japanese Civil Servants Study Women

Men

Model 1 OR

Working overtime No Yes (over 40 h per week) Helsinki Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) Japanese province Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) a

Current smoker.

OR

Model 3 CI 95%

OR

Model 1 CI 95%

OR

n ¼ 780

Model 2 CI 95%

Model 3

OR

CI 95%

OR

CI 95%

n ¼ 2130

1.00 0.77 0.69 0.86

(ref) 0.45 0.40 0.43

1.32 1.19 1.73

1.00 0.97 0.53 0.74

(ref) 0.55 0.29 0.36

1.70 0.96 1.50

1.00 0.92 0.53 0.72

(ref) 0.52 0.29 0.35

1.64 0.96 1.46

1.00 0.97 1.01 1.09

(ref) 0.71 0.76 0.76

1.31 1.36 1.56

1.00 1.09 0.76 0.93

(ref) 0.78 0.56 0.65

1.48 1.04 1.35

1.00 1.07 0.77 0.93

(ref) 0.78 0.56 0.64

1.46 1.06 1.34

1.00 1.08

(ref) 0.71

1.65

1.00 1.31

(ref) 0.83

2.07

1.00 1.29

(ref) 0.81

2.04

1.00 0.95

(ref) 0.76

1.20

1.00 1.16

(ref) 0.91

1.48

1.00 1.10

(ref) 0.86

1.41

n ¼ 4685

n ¼ 1023

1.00 0.88 1.19 1.21

(ref) 0.72 0.98 0.99

1.08 1.44 1.48

1.00 0.99 0.96 1.04

(ref) 0.80 0.78 0.85

1.22 1.17 1.28

1.00 0.98 0.96 1.04

(ref) 0.79 0.78 0.85

1.22 1.17 1.29

1.00 0.95 0.97 0.83

(ref) 0.64 0.65 0.54

1.41 1.45 1.26

1.00 1.03 0.85 0.81

(ref) 0.68 0.56 0.53

1.54 1.28 1.24

1.00 1.09 0.84 0.83

(ref) 0.72 0.55 0.54

1.65 1.26 1.27

1.00 0.95

(ref) 0.77

1.16

1.00 1.04

(ref) 0.84

1.28

1.00 1.03

(ref) 0.84

1.28

1.00 0.68

(ref) 0.47

0.99

1.00 0.77

(ref) 0.52

1.13

1.00 0.73

(ref) 0.49

1.09

n ¼ 301

n ¼ 790

1.00 0.19 0.16 1.99

(ref) 0.02 0.02 0.45

1.79 1.51 8.76

1.00 0.19 0.18 2.10

(ref) 0.02 0.02 0.46

1.83 1.86 9.62

1.00 0.18 0.20 2.08

(ref) 0.02 0.02 0.45

1.77 2.05 9.57

1.00 1.14 1.09 1.45

(ref) 0.79 0.75 0.95

1.66 1.58 2.21

1.00 1.13 1.10 1.45

(ref) 0.78 0.75 0.94

1.64 1.61 2.23

1.00 1.25 1.06 1.56

(ref) 0.86 0.72 1.01

1.84 1.55 2.41

1.00 1.77

(ref) 0.44

7.05

1.00 1.69

(ref) 0.40

7.20

1.00 1.47

(ref) 0.34

6.33

1.00 0.72

(ref) 0.54

0.95

1.00 0.71

(ref) 0.53

0.95

1.00 0.67

(ref) 0.49

0.90

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

London Job strain Low strain Active Passive High strain

Model 2 CI 95%

Table 6 Associations between working conditions and obesitya: logistic regression analyses among public sector employees of the Whitehall II Study (London), Helsinki Health Study, and the Japanese Civil Servants Study Women

Men

Model 1 OR

Working overtime No Yes (over 40 h per week) Helsinki Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week) Japanese province Job strain Low strain Active Passive High strain Working overtime No Yes (over 40 h per week)

CI 95%

OR

Model 3 CI 95%

OR

Model 1 CI 95%

OR

n ¼ 640

Model 2 CI 95%

Model 3

OR

CI 95%

OR

CI 95%

n ¼ 1747

1.00 1.00 1.44 1.46

(ref) 0.58 0.86 0.76

1.71 2.41 2.82

1.00 1.05 1.36 1.39

(ref) 0.59 0.78 0.71

1.85 2.37 2.72

1.00 0.96 1.37 1.33

(ref) 0.54 0.78 0.68

1.70 2.40 2.61

1.00 0.96 1.15 1.37

(ref) 0.65 0.80 0.88

1.42 1.66 2.13

1.00 0.99 1.10 1.35

(ref) 0.67 0.75 0.87

1.46 1.62 2.11

1.00 0.96 1.12 1.33

(ref) 0.65 0.76 0.85

1.44 1.65 2.07

1.00 1.45

(ref) 0.97

2.16

1.00 1.68

(ref) 1.09

2.59

1.00 1.71

(ref) 1.11

2.65

1.00 1.02

(ref) 0.76

1.36

1.00 1.10

(ref) 0.81

1.49

1.00 1.11

(ref) 0.81

1.52

n ¼ 4667

n ¼ 1015

1.00 0.96 1.04 0.91

(ref) 0.76 0.84 0.71

1.20 1.30 1.15

1.00 1.03 0.93 0.84

(ref) 0.82 0.74 0.66

1.31 1.16 1.07

1.00 1.01 0.92 0.84

(ref) 0.80 0.74 0.66

1.28 1.16 1.07

1.00 0.86 0.90 1.22

(ref) 0.53 0.55 0.75

1.42 1.48 1.98

1.00 0.91 0.85 1.22

(ref) 0.55 0.51 0.74

1.50 1.39 1.98

1.00 0.85 0.85 1.18

(ref) 0.51 0.52 0.72

1.42 1.41 1.94

1.00 1.04

(ref) 0.83

1.31

1.00 1.09

(ref) 0.86

1.38

1.00 1.08

(ref) 0.85

1.36

1.00 1.17

(ref) 0.77

1.78

1.00 1.27

(ref) 0.83

1.95

1.00 1.28

(ref) 0.82

1.99

n ¼ 307

n ¼ 804

1.00 1.69 0.90 1.67

(ref) 0.80 0.42 0.67

3.57 1.94 4.18

1.00 1.58 1.01 1.74

(ref) 0.74 0.45 0.69

3.38 2.27 4.40

1.00 1.71 0.95 1.82

(ref) 0.79 0.42 0.71

3.70 2.16 4.63

1.00 1.09 1.08 0.84

(ref) 0.72 0.71 0.51

1.65 1.65 1.39

1.00 1.03 1.10 0.83

(ref) 0.67 0.71 0.50

1.57 1.70 1.38

1.00 1.07 1.08 0.85

(ref) 0.69 0.70 0.51

1.64 1.68 1.42

1.00 0.85

(ref) 0.49

1.47

1.00 0.72

(ref) 0.40

1.29

1.00 0.64

(ref) 0.35

1.17

1.00 0.87

(ref) 0.63

1.21

1.00 0.85

(ref) 0.61

1.18

1.00 0.86

(ref) 0.61

1.21

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

London Job strain Low strain Active Passive High strain

Model 2

a

Body mass index (BMI) 30 or more; for Japan 25 or more; Model: 1 Age-adjusted; Model: 2 Age, occupational class, and marital status adjusted for; Model: 3 All working conditions and covariates mutually adjusted for. 1691

1692

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

London, about a sixth of women and men were current smokers, whereas in Helsinki 21% of women and 25% of men reported smoking. In Japan, nearly half of men, but only 4% of women were smokers. In London, 19% of women and 13% of men were obese, while the percentage of obese was 15 in Helsinki among both women and men. In Japan, the obesity rates were somewhat higher, but the prevalence is not directly comparable due to different criteria. Multivariate models In London, men with passive work were more likely to have unhealthy food habits, while those working overtime had healthier food habits than those with normal working hours (Table 2). However, the associations were statistically significant only in the age-adjusted Model 1. In Helsinki, working conditions were unassociated with food habits. In London, women working overtime were less likely to be physically inactive (Table 3). However, when adjusting for occupational class, marital status, and other working conditions, this association attenuated. A similar pattern was seen for men in London, but it did not reach statistical significance. Instead, men reporting passive work or high job strain were more often physically inactive than men in low-strain jobs. These associations remained after mutual adjustments. Similar associations were observed among women in Helsinki, but not among men, or in Japan. Women in passive work were less likely to report heavy drinking in London, and the association only somewhat attenuated after mutual adjustments (Table 4). Otherwise, no associations were found between working conditions and heavy drinking in London, Helsinki or Japan. We also conducted subsidiary analysis with a moderate cut-off point, i.e., consumption of more than 140 g of pure alcohol per week for men and 105 g for women. Moderate drinking remained inversely associated with passive work among women in London. Additionally, working overtime showed a bidirectional effect in men. In Helsinki, men who worked overtime were more likely to report moderate drinking, while the opposite activity was observed among Japanese men (data not shown). In London, women in passive work reported less current smoking (Table 5). The association even strengthened after mutual adjustment for occupational class and marital status. Also among men in London, those reporting passive work were less likely to be smokers. However, the association was of borderline significance only. Working overtime was associated with non-

smoking in men in Helsinki in the age-adjusted model, but in Japanese men in all three models. Additionally, high job strain was associated with smoking among Japanese men. The pattern seemed to be similar among women, but due to a very low number of smoking women, the associations cannot be ascertained in this study. We also conducted some additional analyses for smoking cessation (data not shown). First, non-smokers were compared to ex-smokers. Working overtime among London men was weakly related to ex-smoking. Secondly, smokers were compared to ex-smokers. Only high job strain was weakly associated with smoking cessation among Japanese men, i.e., those reporting high job strain were less likely to have stopped smoking. Associations between obesity and working conditions were found only among women in London (Table 6). Women who worked overtime were more obese than their counterparts with normal working hours. This association remained even after mutual adjustments. Otherwise, the studied psychosocial working conditions were unassociated with obesity. Some additional analyses were also conducted using BMI as a continuous variable (data not shown). Weak and non-existent associations with job strain and working overtime were found, as with dichotomized BMI reflecting obesity. BMI was used both as such and a logarithmic BMI. The results were similar with both these measures. Finally, interaction analyses examining nested models stratified by gender were conducted with the pooled data to confirm whether the effects of the studied working conditions on adverse health behaviors and obesity differ by cohort (data not shown). These interactions were mostly statistically non-significant suggesting that the associations of working conditions with these outcomes were similar in the different cohorts. The effects of working overtime on food habits and smoking, however, differed between the cohorts among men only. Overall, these interactions were quite inconsistent and difficult to interpret explicitly. Thus, there may be chance findings that need to be confirmed in future analyses. Discussion Main findings This study sought to examine associations of job strain and working time with unhealthy food habits, physical inactivity, heavy drinking, smoking and obesity among employee cohorts from Britain, Finland, and Japan using comparable data. The main overall finding is that associations of these working conditions with

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

adverse health behaviors and obesity are rare and weak in each examined cohort. Nevertheless, some associations were found. Similarities and dissimilarities, as well as inconsistencies in the results among the cohorts and between genders exist. Partly, dissimilarities were expected since the three countries differ in terms of social context and working environment in general. However, some work characteristics may be equally important to health behaviors despite national differences. Working overtime showed associations with non-smoking among men in Helsinki and Japan, for example. Previous studies Previous evidence on working conditions and health behaviors derives mostly from country-specific studies. First, findings for the job strain model will be discussed, followed by results concerning working hours. Our results are in line with a recent review suggesting that job strain has some, albeit limited, associations with health behaviors (Siegrist & Ro¨del, 2006). The associations between passive work, high job strain and physical inactivity among men in London and women in Helsinki agree with previous findings from Finland (Kouvonen, Kivima¨ki, Elovainio, et al., 2005), as well as from Sweden (Ali & Lindstro¨m, 2006). A previously found inverse association between high job demands and alcohol consumption (Kouvonen, Kivima¨ki, Cox, Poikolainen, et al., 2005) was shown in the London cohort for women, but the pattern was statistically significant only with a somewhat lower cut-off point (data not shown). As alcohol dependence has been found to be linked with the psychosocial work environment in terms of effort-reward imbalance among men in London (Head, Stansfeld, & Siegrist, 2004), it is likely that the absence of an association between examined high job strain and heavy drinking reflects somewhat different phenomena. In other words, working conditions may predict alcohol problems, but not necessarily the level of alcohol consumption. While we neither did find any associations between job strain and current smoking in Finland, job strain was, however, associated with smoking in another Finnish public sector study (Kouvonen, Kivima¨ki, Virtanen, et al., 2005). Consistent with this result, smoking showed associations with high job strain within the Japanese cohort. Instead, passive work was inversely associated with smoking among women in London. Previous public sector studies, however, included manual employees in contrast to the current comparisons, which included white-collar employees only. It is possible that the previous findings are, therefore, related to

1693

occupational class variation in health behaviors and risk factors (Hulshof et al., 1991). Thus, if we had included manual employees as well, these results might be more similar. Unique results among the Japanese employees may also relate to the social patterning of these risk factors, which has been found to be different in Japan as compared to Western Europe (Martikainen, Ishizaki, Marmot, Nakagawa, & Kagamimori, 2001). Finally, by focusing on current smoking, we cannot exclude the possibility that working conditions might, however, influence smoking intensity or nicotine dependence more specifically. Lack of information about smoking intensity is a serious limitation as job strain may contribute to the smoking intensity in particular. While we did not observe any associations between job strain and obesity, weak associations have been previously found between job strain and body mass index among Finnish public sector employees (Kouvonen, Kivima¨ki, Cox, Cox, & Vahtera, 2005). Another study concluded that job strain may have differential effects, i.e., its contribution to body mass index may vary according to baseline weight (Kivima¨ki et al., 2006). Additionally, recent prospective evidence within the Whitehall II cohort implies that work stress shows a dose-response effect on obesity that is only modestly attenuated after excluding obese employees at baseline and after adjusting for health behaviors (Brunner, Chandola, & Marmot, 2007). Since the study was prospective, comprised of all age groups, and work stress was assessed using the isolated job strain (iso-strain model) including items about work-place social support, this may partly explain why the findings with the present study are not fully comparable. In contrast to an inverse association between working overtime and smoking found among men from both Japan and Helsinki, a shift from normal hours to working overtime has been shown to increase the likelihood of smoking in Canada (Shields, 1999). A similar shift is also associated with an increase in drinking for women. In our study, working overtime was, however, unassociated with heavy drinking. Additionally, the previous study measured drinking differently, and assessed a change (increase) in the number of drinks during follow-up. Thus, the current results are not directly comparable to the previous ones. As our results are derived from cross-sectional studies, it is not possible to determine whether working long hours had occurred prior to a specific health-related behavior, or if the participant had adopted the health behaviors elsewhere or earlier in life. Furthermore, the information on the duration of time periods spent working overtime was not available.

1694

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

The association between working overtime and obesity among women in London is in line with previous follow-up studies showing increases in body mass index with working overtime among male white-collar employees from Japan (Nakamura et al., 1998) and industrial employees from Finland (Lallukka et al., submitted for publication). Also in agreement with the present finding, a recent Australian study found a positive association between working long hours and higher BMI in men, but not among women (Ostry, Radi, Louie, & LaMontagne, 2006). However, working hours were treated somewhat differently, while normal full-time hours in Australia correspond to 35e40 h a week, and BMI was examined as a continuous variable in contrast to the present study with a dichotomized measure of obesity. Thus, promoting normal working hours may protect against obesity at least for some employee groups. Limitations Some further limitations of the present study need to be acknowledged. These include sample age and job class restriction, use of dichotomized variables potentially contributing to misclassification, different cut-off points for BMI, and a somewhat problematic definition of heavy drinking among women. Additionally, as the data from all three cohorts are cross-sectional, causality cannot be confirmed. Furthermore, most of the information is based on self-reported data, which is prone to personal reporting disposition (Nielsen et al., 2006). Thus, the possibility for under- or over-reporting for some groups cannot be excluded. It is also of note that a rather large number of missing values in London and Japan may have diluted some findings. The response rates, in turn, were high or moderate in Japan and London, while the response rate within Helsinki cohort was somewhat lower. The respondents of the Helsinki Health Study had somewhat higher occupational class and they had somewhat less sickness absence than non-respondents (Laaksonen et al., in press). Moreover, the non-response did not markedly affect the associations between other variables and sickness absence suggesting that even when the data are not fully representative associations between the study variables need not to be seriously biased. However, sickness absence does not equal adverse health behaviors and reflects mainly poor health. Previous studies on other cohorts show that adverse health behaviors are more common among non-respondents to health surveys (Chinn, White, Howel, Harland, & Drinkwater, 2006; Larsen, Grotmol, Almendingen, & Hoff, 2006). Since we did not have information about

health behaviors of the non-respondents, the size of this potential bias remains unknown. Nonetheless, if those with adverse behaviors are less likely to participate and also have more job strain or working long hours, this is likely to have flattened or underestimated our results, i.e., the strength of the associations is likely to be weakened in case of selection bias affecting both the outcomes and the independent variables. Manual workers were not included in the London cohort and, therefore, they were excluded from the Finnish and Japanese cohorts as well. This is likely to truncate the picture on the associations between working conditions and health behaviors. Thus, differences within and between participants and countries are likely to be smaller than if manual workers had been included in the analyses. Furthermore, because all participants were employed, the known beneficial and protective effects of employment on health and on selection into employment, i.e., the ‘‘healthy worker effect’’, are likely to have diluted the results (Wilcosky & Wing, 1987). In other words, those with most adverse health behaviors may have left the workforce for health or other reasons. While on one hand, it is possible that only the healthiest employees are selected to the most strenuous work, on the other hand the most robust employees might have continued in their work despite adverse conditions. In addition to exclusion of manual workers, different gender distributions in the cohorts may have modified the results. Due to the relatively small number of women in the British and Japanese data, statistical significance could not be reached. This does not exclude the possibility that working conditions may, however, affect health behaviors among women in these countries. Additionally, it is possible that associations between working conditions and health behaviors are different or even reverse in younger employees, who have had shorter exposure time to strenuous working conditions (Greenlund et al., 1995). Furthermore, since working in the private sector is associated with a higher risk for myocardial infarction compared to the public sector (Netterstrøm, Nielsen, Kristensen, Bach, & Moller, 1999), it is possible that the relationships between working conditions and health behaviors also differ by sector of employment. Therefore, these results may not apply to the private sector. Generalizability to industrial employees, agriculture, forestry, etc. is likely to be limited as well. It also needs to be acknowledged that with respect to measures of psychosocial work environment, job demands and job control may be perceived differently within Japanese employees as compared to their British and Finnish counterparts. Although the internal

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

consistency of the measures was somewhat lower in Japanese cohorts, identical measures are needed in international comparisons. The Cronbach alphas were, however, similar to a previous Japanese study examining the associations of job strain with health behaviors (Tsutsumi et al., 2003). Job strain has also been examined in several previous studies among Japanese employees and comparable results with western studies have been reached (Tsutsumi et al., 2003). Finally, it is a limitation of this study that both the outcomes and exposure variables were measured at one point in time only. However, longitudinal studies concerning associations of health behaviors and weight with working conditions have equally reported mostly weak associations with similar exposures (Kouvonen, Kivima¨ki, Elovainio, et al., 2008; Lallukka et al., submitted for publication; Shields, 1999). Nonetheless, only with repeated measurements and a prospective design could the causal contribution of work exposures to behavioral risk factors be fully ascertained and the magnitude of the effects established. It is also possible that the cumulative effects of job strain are stronger for other risk factors, such as blood pressure (Landsbergis, Schnall, Pickering, Warren, & Schwartz, 2003). Study strengths This study examined the associations of psychosocial working conditions with adverse health behaviors and obesity. One of the main strengths of the present study is that recent and comparable data from three different countries were included. Additionally, all cohorts were large enabling elaborate statistical analysis, while being representative of the target population. Since the time period between studies is relatively short, the interpretation of the results and comparisons between countries are more valid. Furthermore, data from the studied cohorts are particularly suitable for pooling and comparative analyses, because of similar questions and measures, and the homogeneity of participants in terms of age distribution and employment in the public sector. The dependent and independent variables included in the presented study were harmonized to the best comparability, and only indicators with very similar wording, content and meaning were included in the analyses. Finally, international comparisons provide new information on working conditions and health behaviors moving beyond country-specific assessments to studying variation in the associations across countries and social contexts thereby producing a more comprehensive picture. Additionally, generalizations of the results are improved when findings

1695

can be drawn from three different populations instead of only one. Conclusions Job strain and working overtime have some, albeit limited, associations with health behaviors. Moreover, the associations are mostly weak and rather inconsistent within and between the genders and cohorts. Thus, our results from three cohorts suggest only modest support for the hypothesis that the effect of job strain on employee health might be mediated through health behaviors or obesity. Longitudinal studies are needed to confirm the causal pathways between psychosocial working conditions, health behaviors and subsequent disease and to corroborate whether the mostly nonexistent associations between adverse health behaviors, obesity, and the examined working conditions remain similar in longitudinal settings as well. In other words, repeated measurement for both the outcome measures and strenuous working conditions are needed. Overall, these present results suggest that job strain and working overtime do not have similar effects on all adverse health behaviors, i.e., universally associate with behavioral risk factors. Instead, based on these cross-sectional comparative analyses among white-collar employees, the importance of the examined psychosocial working conditions on adverse health behaviors and obesity is less significant than one might expect considering the time and effort spent in the workplace. Acknowledgment The Whitehall II study has been supported by grants from the Medical Research Council; British Heart Foundation; Health and Safety Executive; Department of Health; National Heart Lung and Blood Institute (HL36310), US, NIH: National Institute on Aging (AG13196), US, NIH; Agency for Health Care Policy Research (HS06516); and the John D and Catherine T MacArthur Foundation Research Networks on Successful Midlife Development and Socio-economic Status and Health. MM is supported by an MRC Research Professorship. The Helsinki Health Study is supported by Academy of Finland (#205588, #70631, #48600, #210435). The Japanese public sector study was granted in part by the Ministry of Health, Labour and Welfare; the Japanese Society for the Promotion of Science; the Occupational Health Promotion Foundation; the Univers Foundation (98.04.017); the Daiwa AngloJapanese Foundation (03/2059); and the Great Britain Sasakawa Foundation (2551).

1696

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

Appendix 1. Karasek comparisons (Karasek 1979) London and Japan: four response alternatives: often - never

Helsinki: five response alternatives: fully agree e fully disagree

Job demands Do you have to work very fast? Do you have to work very intensively? Do you have enough time to do everything (reversed)? Do different groups at work demand things from you that you think are hard to combine?

Job demands I have to work very fast. I have to work very hard. I have enough time to do everything (reversed). Others do not have controversial expectations towards me (reversed).

Job control Others take decisions concerning my work (reversed). Does your job require you to take the initiative? Do you have to do the same thing over and over again (reversed)? I have a good deal of say in decisions about work. Does your work demand a high level of skill or expertise? Does your job provide you with a variety of interesting things? Do you have a possibility of learning new things through your work? Do you have a choice in deciding how you do your work?

Job control I can make decisions concerning my work independently. My work requires creativity. My tasks involve repetition (reversed). I have a great deal of say in my work and tasks. My work demands a high level of skill and expertise. My work involves a lot of different tasks. I have an opportunity to develop my special skills at work. I have very little freedom in deciding how to do my work (reversed)

References Ali, S. M., & Lindstro¨m, M. (2006). Psychosocial work conditions, unemployment, and leisure-time physical activity: a population-based study. Scandinavian Journal of Public Health, 34(2), 209e216. Belkic, K. L., Landsbergis, P. A., Schnall, P. L., & Baker, D. (2004). Is job strain a major source of cardiovascular disease risk? Scandinavian Journal of Work, Environment & Health, 30(2), 85e128. Bhui, K. (2002). Physical activity and stress. In M. S. Stansfeld, & M. Marmot (Eds.), Stress and the heart. Psychosocial pathways to coronary heart disease, (1st ed.). (pp. 158e167). London: BMJ Books. Bobak, M., Pikhart, H., Kubinova, R., Malyutina, S., Pajak, A., & Sebakova, H., et al. (2005). The association between psychosocial characteristics at work and problem drinking: a cross-sectional study of men in three Eastern European urban populations. Occupational & Environmental Medicine, 62(8), 546e550. Brunner, E., Chandola, T., & Marmot, M. G. (2007). Prospective effect of job strain on general and central obesity in the Whitehall II study. American Journal of Epidemiology, 165(7), 828e837. Brunner, E., Shipley, M. J., Blane, D., Davey Smith, G., & Marmot, M. G. (1999). When does cardiovascular risk start? Past and present socioeconomic circumstances and risk factors in adulthood. Journal of Epidemiology and Community Health, 53(12), 757e764. Cargiulo, T. (2007). Understanding the health impact of alcohol dependence. American Journal of Health-System Pharmacy, 64(5 Suppl. 3), S5eS11. Caruso, C. C., Hitchcock, E. M., Dick, R. B., Russo, J. M., & Schmit, J. M. (2004). Overtime and extended work shifts: Recent findings on illnesses, injuries and health behaviors. No. 2004-143. National Institute for Occupational Safety and Health (NIOSH). Chandola, T., Brunner, E., & Marmot, M. (2006). Chronic stress at work and the metabolic syndrome: prospective study. BMJ, 332(7540), 521e525. Chinn, D. J., White, M., Howel, D., Harland, J. O., & Drinkwater, C. K. (2006). Factors associated with non-participation in a physical activity promotion trial. Public Health, 120(4), 309e319.

Dahl, E., Fritzell, J., Lahelma, L., Martikainen, P., Kunst, A., & Mackenbach, J. P. (2006). Welfare state regimes and health inequalities. In J. Siegrist, & M. Marmot (Eds.), Social inequalities in health: New evidence and policy implications, (1st ed.). (pp. 193e222). USA: Oxford University Press. Dynesen, A. W., Haraldsdottir, J., Holm, L., & Astrup, A. (2003). Sociodemographic differences in dietary habits described by food frequency questions e results from Denmark. European Journal of Clinical Nutrition, 57(12), 1586e1597. Esping-Andersen, G. (1990). Three worlds of welfare capitalism. Oxford: Polity Press. Ezzati, M., Henley, S. J., Thun, M. J., & Lopez, A. D. (2005). Role of smoking in global and regional cardiovascular mortality. Circulation, 112(4), 489e497. Ferrie, J. E., Martikainen, P., Shipley, M. J., & Marmot, M. G. (2005). Self-reported economic difficulties and coronary events in men: evidence from the Whitehall II study. International Journal of Epidemiology, 34(3), 640e648. Green, K. L., & Johnson, J. V. (1990). The effects of psychosocial work organization on patterns of cigarette smoking among male chemical plant employees. American Journal of Public Health, 80(11), 1368e1371. Greenlund, K. J., Liu, K., Knox, S., McCreath, H., Dyer, A. R., & Gardin, J. (1995). Psychosocial work characteristics and cardiovascular disease risk factors in young adults: the CARDIA study. Coronary artery risk disease in young adults. Social Science & Medicine, 41(5), 717e723. Head, J., Stansfeld, S. A., & Siegrist, J. (2004). The psychosocial work environment and alcohol dependence: a prospective study. Occupational & Environmental Medicine, 61(3), 219e224. Hellerstedt, W. L., & Jeffery, R. W. (1997). The association of job strain and health behaviours in men and women. International Journal of Epidemiology, 26(3), 575e583. Hu, G., Tuomilehto, J., Silventoinen, K., Barengo, N. C., Peltonen, M., & Jousilahti, P. (2005). The effects of physical activity and body mass index on cardiovascular, cancer and all-cause mortality among 47,212 middle-aged Finnish men and women. International Journal of Obesity, 29(8), 894e902.

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698 Hulshof, K. F., Lowik, M. R., Kok, F. J., Wedel, M., Brants, H. A., & Hermus, R. J., et al. (1991). Diet and other life-style factors in high and low socio-economic groups (Dutch nutrition surveillance system). European Journal of Clinical Nutrition, 45(9), 441e450. Inaba, A., Thoits, P. A., Ueno, K., Gove, W. R., Evenson, R. J., & Sloan, M. (2005). Depression in the United States and Japan: gender, marital status, and SES patterns. Social Science & Medicine, 61(11), 2280e2292. Johansson, G., Johnson, J. V., & Hall, E. M. (1991). Smoking and sedentary behavior as related to work organization. Social Science & Medicine, 32(7), 837e846. Kagamimori, S., Sekine, M., Nasermoaddeli, A., & Hamanisi, S. (2002). Report on stress and health survey in the Japanese civil servants. (in Japanese). Toyama: University of Toyama. Karasek, R. A. (1979). Job demands, job decision latitude, and mental strain: implications for job redesign. Administrative Science Quarterly, 24, 285e308. Karasek, R., Baker, D., Marxer, F., Ahlbom, A., & Theorell, T. (1981). Job decision latitude, job demands, and cardio-vascular disease: a prospective study of Swedish men. American Journal of Public Health, 71(7), 694e705. Kawakami, N., Tsutsumi, A., Haratani, T., Kobayashi, F., Ishizaki, M., & Hayashi, T., et al. (2006). Job strain, worksite support, and nutrient intake among employed Japanese men and women. Journal of Epidemiology, 16(2), 79e89. Kitayama, S., Snibbe, A. C., Markus, H. R., & Suzuki, T. (2004). Is there any ‘‘free’’ choice? Self and dissonance in two cultures. Psychological Science, 15(8), 527e533. Kivima¨ki, M., Head, J., Ferrie, J. E., Shipley, M. J., Brunner, E., & Vahtera, J., et al. (2006). Work stress, weight gain and weight loss: evidence for bidirectional effects of job strain on body mass index in the Whitehall II study. International Journal of Obesity, 30(6), 982e987. Kouvonen, A., Kivima¨ki, M., Cox, S. J., Cox, T., & Vahtera, J. (2005). Relationship between work stress and body mass index among 45,810 female and male employees. Psychosomatic Medicine, 67(4), 577e583. Kouvonen, A., Kivima¨ki, M., Cox, S. J., Poikolainen, K., Cox, T., & Vahtera, J. (2005). Job strain, effort-reward imbalance, and heavy drinking: a study in 40,851 employees. Journal of Occupational and Environmental Medicine, 47(5), 503e513. Kouvonen, A., Kivima¨ki, M., Elovainio, M., Va¨a¨nanen, A. K., De Vogli, R., Heponiemi, T., et al. (2008) Low organisational justice and heavy drinking: a prospective cohort study. Occupational and Environmental Medicine, 65(1), 44–50. Kouvonen, A., Kivima¨ki, M., Elovainio, M., Virtanen, M., Linna, A., & Vahtera, J. (2005). Job strain and leisure-time physical activity in female and male public sector employees. Preventive Medicine, 41(2), 532e539. Kouvonen, A., Kivima¨ki, M., Va¨a¨na¨nen, A., Heponiemi, T., Elovainio, M., & Ala-Mursula, L., et al. (2007). Job strain and adverse health behaviors: the Finnish public sector study. Journal of Occupational and Environmental Medicine, 49(1), 68e74. Kouvonen, A., Kivima¨ki, M., Virtanen, M., Pentti, J., & Vahtera, J. (2005). Work stress, smoking status, and smoking intensity: an observational study of 46,190 employees. Journal of Epidemiology and Community Health, 59(1), 63e69. Laaksonen, M., Aittoma¨ki, A., Lallukka, T., Rahkonen, O., Saastamoinen, P., Silventoinen, K., et al. Non-participation bias in health surveys and check-ups: a register-based study among municipal employees. Journal of Clinical Epidemiology, in press.

1697

Lahelma, E., Martikainen, P., Rahkonen, O., Roos, E., & Saastamoinen, P. (2005). Occupational class inequalities across key domains of health: results from the Helsinki Health Study. European Journal of Public Health, 15(5), 504e510. Lallukka, T., Sarlio-La¨hteenkorva, S., Kaila-Kangas, L., Pitka¨niemi, J., Luukkonen, R., & Leino-Arjas, P. Working conditions and weight gain: a 28-year follow-up study of industrial employees, submitted for publication. Landsbergis, P. A., Schnall, P. L., Pickering, T. G., Warren, K., & Schwartz, J. E. (2003). Life-course exposure to job strain and ambulatory blood pressure in men. American Journal of Epidemiology, 157(11), 998e1006. Larsen, I. K., Grotmol, T., Almendingen, K., & Hoff, G. (2006). Lifestyle characteristics among participants in a Norwegian colorectal cancer screening trial. European Journal of Cancer Prevention, 15(1), 10e19. van Loon, A. J., Tijhuis, M., Schuit, A. J., van Oers, H. A., Surtees, P. G., & Ormel, J. (2004). Are stress related factors associated with alcohol intake? International Journal of Behavioral Medicine, 11(4), 225e235. Markus, H. R., & Kitayama, S. (2003). Models of agency: sociocultural diversity in the construction of action. Nebraska Symposium on Motivation, 49, 1e57. Marmot, M., & Brunner, E. (2005). Cohort profile: the Whitehall II study. International Journal of Epidemiology, 34(2), 251e256. Marmot, M., & Theorell, T. (1988). Social class and cardiovascular disease: the contribution of work. International Journal of Health Services, 18(4), 659e674. Martikainen, P., Ishizaki, M., Marmot, M. G., Nakagawa, H., & Kagamimori, S. (2001). Socioeconomic differences in behavioural and biological risk factors: a comparison of a Japanese and an English cohort of employed men. International Journal of Epidemiology, 30(4), 833e838. Nakamura, K., Shimai, S., Kikuchi, S., Takahashi, H., Tanaka, M., & Nakano, S., et al. (1998). Increases in body mass index and waist circumference as outcomes of working overtime. Occupational Medicine, 48(3), 169e173. Netterstrøm, B., Nielsen, F. E., Kristensen, T. S., Bach, E., & Moller, L. (1999). Relation between job strain and myocardial infarction: a caseecontrol study. Occupational and Environmental Medicine, 56(5), 339e342. Nielsen, N. R., Kristensen, T. S., Prescott, E., Strandberg Larsen, K., Schnohr, P., & Gronbaek, M. (2006). Perceived stress and risk of ischemic heart disease: causation or bias? Epidemiology, 17(4), 391e397. Oliver, G., & Wardle, J. (1999). Perceived effects of stress on food choice. Physiology & Behavior, 66(3), 511e515. Oliver, G., Wardle, J., & Gibson, E. L. (2000). Stress and food choice: a laboratory study. Psychosomatic Medicine, 62(6), 853e865. Ostry, A. S., Radi, S., Louie, A. M., & LaMontagne, A. D. (2006). Psychosocial and other working conditions in relation to body mass index in a representative sample of Australian workers. [Computer File]. BMC Public Health, 6, 53. Paavola, M., Vartiainen, E., & Haukkala, A. (2004). Smoking from adolescence to adulthood: the effects of parental and own socioeconomic status. European Journal of Public Health, 14(4), 417e421. Parrott, A. C. (1999). Does cigarette smoking cause stress? The American Psychologist, 54(10), 817e820. Romelsjo¨, A., Hasin, D., Hilton, M., Bostro¨m, G., Diderichsen, F., & Haglund, B., et al. (1992). The relationship between stressful working conditions and high alcohol consumption and severe

1698

T. Lallukka et al. / Social Science & Medicine 66 (2008) 1681e1698

alcohol problems in an urban general population. British Journal of Addiction, 87(8), 1173e1183. Schneider, S., & Becker, S. (2005). Prevalence of physical activity among the working population and correlation with work-related factors: results from the first German national health survey. Journal of Occupational Health, 47(5), 414e423. Sekine, M., Chandola, T., Martikainen, P., Marmot, M., & Kagamimori, S. (2006). Socioeconomic inequalities in physical and mental functioning of Japanese civil servants: explanations from work and family characteristics. Social Science & Medicine, 63(2), 430e445. Shields, M. (1999). Long working hours and health. Health Reports, 11(2), 33e48. Siegrist, J., & Ro¨del, A. (2006). Work stress and health risk behavior. Scandinavian Journal of Work, Environment & Health, 32(6), 473e481. Singh-Manoux, A., Martikainen, P., Ferrie, J., Zins, M., Marmot, M., & Goldberg, M. (2006). What does self rated health measure? Results from the British Whitehall II and French Gazel cohort studies. Journal of Epidemiology and Community Health, 60(4), 364e372. Smeeding, T. M., & Gottschalk, P. (1999). Cross-national income inequality: how great is it and what can we learn from it? International Journal of Health Services: Planning, Administration, Evaluation, 29(4), 733e741. Stansfeld, S. A., & Marmot, M. G. (2002). In: S. A. Stansfeld, & M. G. Marmot (Eds.), Stress and the heart: Psychosocial pathways to coronary heart disease, (1st ed.). London: BMJ Publishing Group. Tsutsumi, A., Kayaba, K., Hirokawa, K., & Ishikawa, S.The Jichi Medical School Cohort Study Group (2006). Psychosocial job

characteristics and risk of mortality in a Japanese communitybased working population: the Jichi Medical School cohort study. Social Science & Medicine, 63(5), 1276e1288. Tsutsumi, A., Kayaba, K., Theorell, T., & Siegrist, J. (2001). Association between job stress and depression among Japanese employees threatened by job loss in a comparison between two complementary job-stress models. Scandinavian Journal of Work, Environment & Health, 27(2), 146e153. Tsutsumi, A., Kayaba, K., Yoshimura, M., Sawada, M., Ishikawa, S., & Sakai, K., et al. (2003). Association between job characteristics and health behaviors in Japanese rural workers. International Journal of Behavioral Medicine, 10(2), 125e142. Visscher, T. L., Viet, A. L., Kroesbergen, I. H., & Seidell, J. C. (2006). Underreporting of BMI in adults and its effect on obesity prevalence estimations in the period 1998 to 2001. Obesity, 14(11), 2054e2063. Wardle, J., Steptoe, A., Oliver, G., & Lipsey, Z. (2000). Stress, dietary restraint and food intake. Journal of Psychosomatic Research, 48(2), 195e202. Wilcosky, T., & Wing, S. (1987). The healthy worker effect. Selection of workers and work forces. Scandinavian Journal of Work, Environment & Health, 13(1), 70e72. Yang, J. J., Shiwaku, K., Nabika, T., Masuda, J., & Kobayashi, S. (2007). High frequency of cardiovascular risk factors in overweight adult Japanese subjects. Archives of Medical Research, 38(3), 337e344. Zins, M., Carle, F., Bugel, I., Leclerc, A., Di Orio, F., & Goldberg, M. (1999). Predictors of change in alcohol consumption among Frenchmen of the GAZEL study cohort. Addiction, 94(3), 385e395.