Influence of job strain on changes in body mass index and waist circumference—6-year longitudinal study

and Health Cohort Study Group. Influence of job strain on changes in body mass index and waist circumference—6-year longitudinal study. Scand J Work Environ Health 2008;34(4):288–296 . Objectives This study examined the effect of changes in psychosocial workplace characteristics on weight gain and abdominal obesity. Methods Twice, at an interval of 6 years, the authors conducted examinations of job demand–control–support and measurements of body mass index and waist circumference among 2200 men and 1371 women aged 30–53 years and working in a factory. The scores of the psychosocial work characteristics in each examination were dichotomized at the median values for the men and women separately and then categorized into three groups as follows: group I: low score in both the first and second examinations, group II: low score in the first examination and high score in the second (or high at the first and low at the second), and group III: high in both the first and second examinations. Results Although there was no statistically significant association between psychosocial work characteristics and the change in body mass index, for both genders, the change in waist circumference increased more in group III than in group I. Similarly, the odds ratios for the change in waist circumference above the 75th percentile for groups II and III increased more than in group I, being 1.13 [95% confidence interval (95% CI) 0.87–1.46 and 1.39 (95% CI 1.07–1.79) for the men in groups II and III, respectively, and 1.27 (95% CI 0.90–1.78) and 1.78 (95% CI 1.26–2.52) for the women in groups II and III, respectively. Conclusions The results suggest that high job strain is a risk factor for increased abdominal obesity.

Several studies have shown that psychosocial work stress, such as low job control and job strain combined with high job demand and low job control, which are considered to be work stressors in the job demand-control model proposed by Karasek, are related to the development of coronary heart disease (CHD) (1,2). The pathophysiological mechanism underlying the relationship between work stress and CHD has been suggested to be increased ambulatory blood pressure (3,4) and abnormalities of blood coagulation and fibrinolytic function (5).
On the other hand, although obesity is an important CHD risk factor, it is still unclear whether work stress is connected to weight gain. The metabolic syndrome, which is a pathophysiological state in which a cluster of factors such as abdominal obesity, atherogenic dyslipidemia, raised blood pressure, and glucose intolerance are believed to promote CHD, is attracting increasing attention (6). A cross-sectional analysis in the Whitehall II study suggested a biological explanation that socioeconomic inequality in CHD was partly attributable to the metabolic syndrome, because people with a lower Ishizaki et al socioeconomic status have a greater waist circumference (7). A recent Whitehall II study, with an average 14-year-follow-up, showed an association between chronic work stress and the metabolic syndrome, more exposure to a state of low support with job strain being associated with a greater risk of the metabolic syndrome even after adjustment for employment grade and health behavior (8).
Many epidemiologic investigations have used waist circumference or waist-to-hip ratio as an anthropometric measurement index of abdominal obesity, which is closely related to the metabolic syndrome. Rosmond et al (9) and Rosmond & Björntorp (10) reported that inferior work conditions, such as less satisfaction with work management, less influence on work situations, and a lack of attempts to alter work situations, were associated with an increased waist-to-hip ratio.
In our previous cross-sectional investigation on Japanese employees, we found no statistically significant correlation between job demand-control and body mass index (BMI) or waist-to-hip ratio (11). However, we believe that a follow-up study would better clarify the influence of job demand-control on anthropometric measures, waist circumference, or waist-to-hip ratio, since, for example, considerable time must elapse before any changes in anthropometric measurements become apparent after exposure to certain work conditions. In addition, in the past decade, many companies in industrial countries have been trying to dynamically outrun others in the global economy race by introducing various managerial innovations, such as just-in-time production and total quality management (12). As a result, we can expect a rapid change in the work stress perceived by employees.
Therefore, in this study, we compared the results of two psychosocial work characteristics of the same persons in investigations conducted at an interval of 6 years and examined how changes in job demand-control affected the workers' anthropometric measurements.

Study population and methods
In our study, nonmanual and manual employees working for an aluminum-products factory in a rural area of Japan were asked to reply to the Japanese version of the job content questionnaire (13), about the status of job demand-control-support as individual psychosocial work characteristics. The survey was conducted twice, first from April 1996 through March 1997 and then from April 2002 through March 2003. On both occasions, only the persons who had provided their written consent to participate were included in the investigation. The selected workers were aged 30 to 53 years at the time of the first examination so that they were under 60 years of age, namely, the retirement age of the factory, at the time of the second examination. The participation rate of the first examination was 91.4% of the registered workers, or 2821 men and 1701 women, excluding pregnant women, as of 1 May 1996. Altogether 121 men and 39 women out of this population missed the opportunities or refused to consent to having their waist circumference measured. Before the second examination, 186 men and 184 women had resigned and 185 men had been transferred, and consequently they were excluded from the follow-up. Furthermore, 72 men and 97 women who did not reply to the second job content questionnaire or failed to undergo the second waist circumference measurement were also excluded.
Altogether, we included 2200 men and 1371 women as eligible participants whose data from the questionnaire and anthropometric measurements at both examinations were available and who had given complete replies to the questions concerning the confounding factors, such as sedentary job, shift work, and other health behavior, at the first examination. Managers and professionals accounted for 14% of all the men, whereas only 2% of the women were managers or professionals.
Job strain was calculated as a value of job demand divided by job control. The median value of each psychosocial work characteristic of the participants from the age of 30 to 53 years did not change between the two surveys. The median values of the job demand scores, the job control scores, and the scores for worksite support were 66, 32 and 23 for the men and 60, 32, and 22 for the women, respectively. But the median values of the job strain scores slightly changed, from 0.485 for the men and 0.533 for the women in the first examination to 0.500 for the men and 0.536 for the women, respectively, in the second examination. The scores for job control, job demand, worksite support, and job strain in each examination were dichotomized at the median value for the men and women separately and then categorized into three groups as follows: group I: low score in both the first and second examinations, group II: low score in the first examination and high score in the second (or high in the first and low in the second), and group III: high in both the first and second examinations (figure 1).

Anthropometric measurements
Anthropometric measurements were conducted within a month before or after the questionnaire survey on both occasions, the weight, height, and waist circumference of the participants being measured with them wearing light clothes. The BMI (kg/m 2 ) of the participants was calculated by dividing their weight by their height squared. The waist circumference (centimeters) was measured at the umbilicus level by experienced nurses. Underwear worn to correct body shape was removed.

Sedentary job and shift work as other work characteristics
Sedentary job was categorized into three groups according to the average number of sedentary hours per workday in the previous year ("<1 hour", "≥1 and ≤4 hours", and "≥5 hours"). The three-shift workers who worked nights were categorized as a shift work group, as opposed to a nonshift work group. No female worker in the factory was engaged in three-shift work.

Other health behavior
Queries were made about several lifestyle factors. The workers were classified as "non-or ex-smokers" and "current smokers". Alcohol consumption was measured in terms of grams of ethanol consumed per week and was categorized into five groups for the men (ie, no drinking, 1-175 g/week, 176-350 g/week, 351-525 g/week, and ≥526 g/week). As only five women ingested more than 350 grams of alcohol per week, alcohol consumption was re-categorized into three groups as follows: nondrinker, 1-175 g/week, and ≥175 g/week. Exercise for the men was classified as "almost no exercise", "light exercise", "brisk and sweating exercise once or twice a week" and "brisk and sweating exercise more than three times a week". The degree of education for the men was determined by the total years of education and was classified as "<11 years", "11-12 years", "13-14 years", and "≥15 years" of education. Since relatively few women participated in brisk and sweating exercise more than three times a week or had more than 15 years of education, exercise and education were re-categorized into three groups for the women ("almost no exercise", "light exercise", and "brisk and sweating exercise" and "<11 years", "11-12 years", and "≥13 years", respectively). Marital status was divided into "married" and "previously or never married". This study was approved by the Ethics Committee of the Kanazawa Medical University.

Statistical analyses
The data were analyzed separately for the men and women using an SAS program package (SAS Inc, Cary, NC, USA). Changes in BMI and waist circumference were expressed as (second examination value -first examination value) / first examination value. The BMI and waist circumference in the first examination, the change in BMI, and the change in waist circumference were compared among different groups using a general linear model. The data for sedentary job, shift work, smoking habits, alcohol consumption, exercise, education, and marital status (inquired about in the first examination) were adopted as potential confounding factors.
Logistic regression analyses were used to calculate the odds ratios of the change in BMI and the change in waist circumference above the 75th percentile according to changing job strain; they were 0.0458 and 0.0600 for the men and 0.0456 and 0.1080 for the women, respectively.

Results
The mean period from the first to the second examination of the job content questionnaire was 72.0 (SD 2.7) months for the men and 71.2 (SD 1.7) months for the women. Table 1 shows the differences in the mean values for age, BMI, waist circumference, and psychosocial work characteristics in the first examination between the workers with complete data from the two surveys and those for whom only data from the first examination were available. For the men, those who participated in both examinations were younger and had lower job control than those who participated only in the first examination. For the women, the latter group was younger and had [ § = The scores for job control, job demands, worksite support, and job strain (demand and control) in each examination were dichotomized at the median value for 2200 men and 1371 women after those with no job content questionaire or anthropometric measurement data on either occation were excluded or who resigned or transferred after the 1st examination.] (1) group I : low score to low score (2) group II : low score to high score high score to low score (3) § The scores of job control, job demands, worksite support and job strain(demons/control) in each examination were dichotomized at the median value in 2200 males and 1371 females after excluding those with no JCQ or anthropometric measurement data at either occasion or who resigned or transferred after the 1st examination  (14) and Belgians (15).
Regarding the association between psychosocial work characteristics and lifestyle at the first examination, there were no differences in smoking habits or alcohol consumption between the low and high psychosocial Table 1. Mean levels and standard deviations (SD) of the body mass index (BMI), waist circumference, and psychosocial work characteristics at the first examination for those who participated in both the baseline and the follow-up examinations and those who participated only at the baseline.  Table 2. Body mass index (BMI) and waist circumference in the first examination according to the change in the psychosocial work characteristics. (group I = low score in both the first and second examinations, group II = low score in first examination and high score in second examination or high score in first examination but low score in second examination, group III = high score in both the first and second examinations) work characteristic groups of either gender. More men in the high job-strain group had standing work, shift work, no regular exercise, and shorter education and were not married in comparison with the men in the low job-strain group. The women in the high job-strain group had more standing work, less regular exercise, and shorter education than those in the low job-strain group. The men in the low worksite-support group had more standing work and shorter education than those in the high worksitesupport group. No difference between the low and the high worksite-support groups was found for the women (data not shown). Table 2 on page 291 shows the mean BMI levels and waist circumference values according to the subgroups of job control, job demand, worksite support, and job strain in the first examination. For the men, both the BMI and waist circumference were larger in group III than in group I for job control and similarly larger in group I and III than in group II for job demand. As regards job strain, group I had the largest waist circumference among the three groups. For the women, there were no statistically significant differences in BMI or waist circumference among the subgroups of any of the psychosocial work characteristics. After adjustment for potential confounding factors, these differences did not change much for either gender.
The associations of the change in BMI and the change in waist circumference (after adjustment for the confounding factors) with job control, job demand, worksite support, and job strain are shown in table 3  and table 4. No significant difference was found in the change in BMI among the three different groups of the psychosocial work characteristics for either gender. The change in waist circumference was significantly higher in group III for job strain than in groups I and II among the men, and also a marginally significant similarity was found for the women. There was no significant interaction between job strain and the other work conditions (ie, sedentary work and shift work) with respect to the values of the change in BMI or the change in waist circumference. In the analyses of the relationship between job strain and the change in BMI and the change in waist circumference, BMI was categorized into slender, moderate, and overweight groups. Hereupon the moderate group denoted the mean value plus or minus one standard deviation of the BMI in the first examination. The ranges of the slender, moderate, and overweight groups were <20.38 kg/m 2 , 20.38-26.02 kg/m 2 , and ≥26.03 kg/m 2 for the men and <19.18 kg/m 2 , 19.18-25.66 kg/m 2 , and ≥25.67 kg/m 2 for the women, respectively. For the men, the proportions of people in the overweight group at the first examination who gained weight or had an increase in their waist circumference during the period between the two examinations were 53.1% (BMI) and 59.3% (waist). The rates were lower than those in the slender and moderate groups at the first examination. The change in BMI in the overweight group at the first examination was also smaller than that in the slender or moderate group for both genders.
On the other hand, the change in waist circumference increased in all of the BMI categories of slender, moderate, and overweight at the first examination. The change in waist circumference among the men was larger in group III than that in group I or group II. The change in waist circumference in group III was similarly larger than that in group I or group II for both the moderate and overweight women in the first examination.
For those who lost weight during the interval between the first and second examinations, no significant difference in the change in BMI or the change in waist circumference was found among the three job-strain groups (data not shown). Table 5 shows the odds ratios of the change in BMI and the change in waist circumference above the 75th percentile according to the changes in job strain. Regarding the change in waist circumference for both genders, group III showed a significantly higher rate of change than group I. The odds ratios were 1.13 [95% confidence interval (95% CI) 0.87-1.46] in group II and 1.39 (95% CI 1.07-1.79) in group III for the men and 1.27 (95% Table 3. Change in body mass index (BMI) and waist circumference according to changes in job control, job demand, and worksite support. (group I = low score in both the first and second examinations, group II = low score in first examination and high score in second examination or high score in first examination but low score in second examination, group III = high score in both the first and second examinations)

Discussion
This 6-year follow-up study showed that the change in waist circumference in job-strain group III increased more than that in job-strain group I, even though no statistically significant association was noted between the psychosocial work characteristics and the change in BMI. Moreover, the prevalence rate of the change in waist circumference above the 75th percentile among both the men and women increased progressively in order from group I to group II to group III of job strain after adjustment for age, BMI, sedentary job, shift work, smoking, alcohol, exercise, education, and marital status. The results of this study showed that high job strain may contribute to abdominal obesity. Although two previous cross-sectional investigations referred to the association between job strain and abdominal obesity among men (16,17), one review noted that the association was not clear (18). Recently, Table 4. Change in body mass index (BMI) and waist circumference in relation to job strain according to the BMI category in the first examination. (job-strain group I = low score in both the first and second examinations, job-strain group II = low score in first examination and high score in second examination or high score in first examination but low score in second examination, job-strain group III = high score in both the first and second examinations) BMI Table 5. Odds ratios (OR) and their 95% confidence intervals (95% CI) for the change in body mass index (BMI) and waist circumference above the 75th percentile according to job-strain change. (group I = low score in both the first and second examinations, group II = low score in first examination and high score in second examination or high score in first examination but low score in second examination, group III = high score in both the first and second examinations) Brunner et al (19) showed that job strain partly caused abdominal obesity, because a dose-response relationship between work stress and obesity was found in their 19year follow-up study. With an increased number of cases classified as iso-strain (ie, the lowest tertile of worksite support combined with job strain, evaluated on four occasions during the follow-up), the incidences of a high BMI of ≥30 kg/m 2 for both genders and a large waist circumference of >102 cm among the men and >88 cm among the women increased. Rosmond & Björntorp (20) and Rosmond (21) have suggested that, as one of the pathophysiological mechanisms underlying the association between job strain and waist circumference, psychosocial disadvantage pressure affects the activity of the hypothalamic-pituitary-adrenal axis and, as a result, increases the cortisol level. This increase in the cortisol level then causes abdominal fat to accumulate and therefore leads to an increased waist circumference.
The measurement of saliva cortisol has frequently been used to examine the neuroendocrine excretion status in field studies. Those who perceived high chronic work overload (22) and high social stress (23) showed increased cortisol levels on awakening in the morning. The mean cortisol level of workdays was higher in a low job-control group than in a high job-control group among the men, and, among the women with a low socioeconomic status, the mean cortisol level of the workdays was higher in a high job-demand group than in a low job-demand group (24). In addition, for men, a positive association was found between the waist-tohip ratio and the cortisol response to waking (25), and, for women, the urinary cortisol level per 24 hours was increased (26).
It is well known that there is an inverse correlation between socioeconomic status and BMI among people in developed countries (27). Although people with a low socioeconomic status are expected to be under high work stress (ie, low job control), the influence of job demand-control on BMI is obscure. Job strain has not been found to be associated either with BMI in various large cross-sectional population studies among Japanese-Americans, working women, and Canadian white-collar workers (28)(29)(30), nor with weight gain in a 5-year prospective study on civil servants (31). High job demand or low job control was not associated with weight gain in the past year (32). On the other hand, according to the data collected from the 32 worksites in a cross-sectional study, the women in the high-strain group had a higher BMI than those in the other groups, but this trend was not found for the men (33). A study in France found a relationship between high job demand and overweight among women, but not among men (34). Kivimäki et al (31) pointed out the bidirectional effect of work stress on BMI as one reason for the inconsistent correlation between work stress and BMI, because work stress could not only lead to hyperphagia but also to hypophagia. The population of our study may have included some workers who lost weight due to work stress. However, no significant difference in the change in BMI or the change in waist circumference existed among the three job-strain groups for those who lost weight during the intervening period. It is possible that some of the workers who lost weight due to severe anorexia caused by work-stress-induced depression were not able to participate in the examinations because they were not working on the occasions and consequently were excluded from the participants.
Notably, in spite of focusing on the same target population, a 19-year follow-up study found work stress to be related to weight gain, while another 5-year follow-up study found no such relationship (19,31). The former study noted the accumulated effect of work stress, and the observation lasted for a longer period in comparison with that of the latter study. In addition, another study pointed out that the evaluation of job strain at a single point in time possibly underestimated the association between job strain and CHD (35). Thus our previous cross-sectional study may similarly have underestimated the association between job strain and the waist-to-hip ratio.
No changes in the mean scores of job demand, job control, or worksite support were found between the first and second examinations in our study. Prior research showed stability for scores of the job content questionnaire on two occasions, before and after an average interval of 6.6 years among 2490 Europeans who remained in the same job (15). In addition, the scores of the work characteristics for the same persons did not change appreciably over a 5-year interval in Japan (36). However, the Japanese study also found that the scores were less stable when there was a position change even within the same company. Likewise, about one-third of the participants of our study showed some changes in the scores of the job content questionnaire during the 6-year period, shifting from the high group to the low group of psychosocial work characteristics and vice versa.
We categorized the persons with improved psychosocial work characteristics and those who showed deterioration in this respect together as group II, because they were likely to have experienced greater changes in other work conditions, such as workplace, shift work, and sedentary job than group I and group III did. In addition, it is difficult to know exactly when the particular change in psychosocial work characteristics started, as the effect of a change in an anthropometric measurement does not manifest itself immediately but, rather, takes time.
Adopting many factors as potential covariates may weaken the relationship between job strain and the change in BMI or the change in waist circumference.

Ishizaki et al
For instance, many people in the high-strain group had less regular exercise in association with an increased BMI and waist circumference.
Incidentally, the job-strain scores of the women in this study were higher than those of the men. Furthermore, the women rarely changed their occupations and tended to remain in a relatively low employment job. These facts may have affected the results of this follow-up study, making the relationship clearer between job strain and the change in waist circumference.
Some investigations have also shown that work stress, when evaluated in a job demand-control model, was associated with glucose metabolism, blood coagulation, and fibrinolytic function as risk factors of cardiovascular disease in Japan, as well as in other developed countries (37). Obesity (BMI ≥30 kg/m 2 ) is certainly less common in Japan than in western Europe and the United States (38). However, the proportions of overweight (BMI≥≥25 kg/m 2 ) Japanese men are 32.7% for those 40-49 years of age and 30.8% for those 50-59 years of age. The corresponding proportions for Japanese women are 17.9% and 24.1%, respectively. In addition, one in every two men and one in every five women are said to have suspected or potential metabolic syndrome (39). Therefore, the national government has begun to make a concerted effort to tackle the metabolic syndrome by making the measurement of waist circumference mandatory when people aged 40-74 years are screened in medical checkups (40).
There were several potential limitations in this study. First, we evaluated and classified psychosocial work characteristics for the same people using the job content questionnaire twice at an interval of 6 years, but we have no data on the fluctuation of the psychosocial work characteristics during this period. Similarly, although we used the scores of several confounding factors at the first examination, we did not record their subsequent changes. In addition, as about 25% of the men and about 20% of the women of all of the participants either retired, were transferred, did not fill out the questionnaire completely, or refused to participate in this examination, they were excluded from the follow-up survey. This exclusion may have conceivably affected the results, although a large population of both genders was available for continued follow-up. Second, the waist circumference was lower for group III in the first examination. This initial low score may possibly have contributed to the increase in waist circumference in the second examination. However, especially for the women, the difference in the change in waist circumference among the three job-strain groups was larger than the difference affected by the initial potential bias. Third, we did not make a dietary survey with respect to weight gain. However, a large-scale survey of 25 000 Japanese by Kawakami et al (41) did not reveal any evident connection between job strain and total energy intake, even after adjustment for age, educational background, and occupation. Fourth, since the participants of our study were all from a single company, whether our results can be generalized or not will have to be determined in further studies.
In conclusion, we examined psychosocial work characteristic twice for 2200 men and 1371 women with an interval of 6 years between the examinations. We admit that there was a bidirectional influence of work stress on BMI and waist circumference, and yet the results of our study showed that high job strain increased the change in waist circumference even when several potential confounding factors were taken into consideration. This result supports the finding of Brunner et al (19), who reported that chronic work stress may contribute to abdominal obesity. Hence it is important that we take measures to reduce the chronic work stress of workers in terms of preventing atherosclerotic and other diseases triggered by the metabolic syndrome.