Job Characteristics and Serum Lipid Profile in Japanese Rural Workers: The Jichi Medical School Cohort Study

To observe the association between adverse psychosocial job characteristics, measured by the Karasek job demand-control questionnaire, and a lipid profile, cross-sectional analyses were performed for a Japanese rural working population. The study population comprised 3,333 male and 3,596 female actively employed workers, aged 65 years and under. Among men, higher psychological demands were associated with high total cholesterol levels, with an adjusted difference from the top to bottom tertiles of 3.3 mg/dl (F = 3.03; p = 0.048). High demands were also positively associated with the total/HDL cholesterol ratio (F = 3.94; p = 0.020). Neither job control nor job strain (the ratio of demands to control) was associated with any of the lipid levels in either gender. A psychologically demanding job may be associated with an unfavorable lipid profile, but the impact of job strain on atherogenic lipids is negligible.

Evidence has been accumulating that adverse psychosocial job characteristics may predict the onset of cardiovascular diseases.In occupational stress research, the job demand-control model is currently most prevalent among those that demonstrate such evidence.1-3The hypothesis of the job demand-control model proposes that employees with a combination of high psychological job demand and low control over their job (i.e., job strain) are at risk of developing illnesses.4 Although the combined effect of the demand and control components is of central interest to this model, its individual components may also predict a worker's stress-related health outcome.
Despite the numerous empirical studies that show an association between job strain or components and cardiovascular diseases, the physiologically mediating mechanisms remain unre-solved.Metabolic disorders, such as an unfavorable lipid profile, might be one of the mediators,5-6 but the evidence on the association between job strain and atherogenic lipids is far from definite.Theorell et al. 7 showed that there was a significant correlation between the ratio of job demands to influence over the job and total cholesterol level.Niedhammer et al.8 showed a high prevalence of reported hyperlipidemia in men exposed to both high job demands and low support, and in women with low decision latitude (low job control).This finding about women was replicated in other studies.9,1011 However, there are many other studies in which no association between job strain or its components and atherogenic lipids was demonstrated.11-17 Outside of Europe and the United States, very few studies have been conducted to investigate this association.2,13,14Furthermore, most of the earlier studies were conducted on workers employed in big enterprises or in urban settings.To the best of our knowledge, except for some epidemiologic studies of which study population was national representative sample and presumably included pre-industrial workers,16 studies that investigated occupational stress involving large populations of workers in farming, forestry and fisheries are scarce.Even in such a study, attempts have not been made to explore the association between job characteristics and atherogenic lipids specifically for pre-industrial workers.Accordingly, the aim here was to observe the association in Japanese workers from a large community-dwelling sample in non-urban areas.

METHODS
The aim of the Jichi Medical School Cohort Study was to investigate the risk factors of cardiovascular disease in Japan.Data were collected between 1992 and 1995.Ultimately, 12,490 Japanese (4,911 men and 7,579 women) from 12 rural communities located across Japan participated.18 In accordance with the provisions of the Health and Medical Service Law for the Aged, a mass screening examination program concerned with the risk factors for cardio-cerebrovascular disease have been conducted since 1983: The law requires municipal governments to manage the programs efficiently and offer them to all residents who are willing to participate.The target subjects vary according to each community, from all the residents to those who are not offered physical examinations at their workplaces or elsewhere including the subjects of the National Health Insurance.Residents aged 40-69 years were the subjects for the mass screening examination program in 8 of the 12 communities, those aged 30 years and older in one, and adults for other age groups were included in the rest.Accordingly, the cohort for the study was to include residents between 40 and 69 years of age living in the first 8 communities and between 30 and 69 years in the other 4 initially.In each community, the local government office sent letters to all potential participants inviting them to take part in the program.The invitation mentioned that persons who were visiting hospitals or clinics because of cardiovascular disease did not have to take the examination.People other than those in the above-defined age groups (n = 282 for the younger age group and 696 for those over 69) as well as those who voluntarily applied for the program participated in the study and are included in the database.The overall response rate was 65.4%, whereas the response rate based on the intended subjects between 40 and 69 years was 62.7%.
In this analysis, the study population was limited to actively working men and women who were not older than 65 because the aim of this analysis was to observe the association between job characteristics and serum lipid profiles.A total of 6,929 participants (3,333 men and 3,596 women) who were employed and had filled out the questionnaire were subjected to the present study.The following occupations were included: farming and forestry (n = 1,074 men, 1,253 women), fisheries (248, 37), secu-rity (21, 1), transportation (94, 4), construction (629, 86), production (345, 680), business (258, 378), office work (200, 352), professional (201, 207), and the service industry (263, 598).The first six were designated blue-collar jobs and the next four, white-collar.With regard to employment status, workers who classified themselves as administrators or self-employed were labeled managerial.More than 99% of the workers were either self-employed or employed by companies with less than 300 employees, which may reflect the current industrial structure in Japan.19However, it should be noted that the under-representativeness for younger working population precludes the generalizability of the results.Some who were employed part-time may have been included in the study population but this was not ascertained.

Procedure
Socio-demographic and behavioral variables were obtained by using a standardized questionnaire, which included information on occupational environment, age, marital status, educational attainment, dietary habits, smoking habits, alcohol consumption, physical activity, and medical history.The questionnaire was distributed to the subjects beforehand to complete on their own.Informed consent was obtained from all prospective participants.Job characteristics were derived by using a Japanese version of the demand-control questionnaire20 originally developed by Theorell et al.21 The psychometric property of the questionnaire has been reported elsewhere.22 Those job characteristics, job control and psychological demands, were defined on two scales.The job control was defined as the sum of two subscales that were given equal weight: (1) skill discretion, measured by four parameters (possibility for learning new things, skills required by the job, requirement for creativity, and repetitious nature of the work); and (2) autonomy for decision making, measured by two parameters (right to make one's own decisions and freedom in choosing the manner by which the work is performed).The second scale, psychological job demands, was defined by five parameters (speed in completing work, degree of difficulty of the work, excessive workload, insufficient time allowed to complete work, and conflicting demands).All questions were scored on a Likert scale of 1 to 4. Cronbach's coefficient alpha for the job control index and for the psychological demands index was 0.64 and 0.70, respectively.Job strain was defined as the ratio of demands to job control.The participants were grouped into one of three strata for each job characteristic index (low, medium or high) based on tertiles, defined according to the distribution of scores in the total working population, then separately for men and women.
The following socio-demographic and behavioral variables were used as possible covariates.The participants were grouped by age into four categories: under 39 years, 40-49, 50-59, and 60-65.Marital status was coded as currently married or unmarried.Educational attainment was categorized into two strata: lower or higher than the level of compulsory education.
The dietary habits of the participants were ascertained by employing a food frequency questionnaire, which was composed of 30 different foods most likely to be consumed , each on a graded five-point Likert scale.23 The items were subjected to a factor analysis with Varimax rotation; and with regard to the dietary pattern, three additive scales that were composed of factors with eigenvalues greater than 1.5 were created.An item was included in a scale if it had a loading of 0.40 or above on one factor but less than 0.39 on any other factor.These cutoff points were selected to ensure that no item was included on more than one scale.They were: vegetable type (food preference for green vegetable, yellow-green vegetables, potatoes, fruits, tofu, seaweed, oranges, beans, and dried fish); meat type (ham, pork, beef, chicken, and fish-paste); and Western type food (bread, butter, rice (reversed score), salty food (reversed score), miso soup (reversed score), and yogurt).The alpha coefficients for the three scales ranged from 0.55 to 0.76.A participant was placed in one of three linear strata for the food frequency pattern based on the tertile in which his/her score fell.Analysis of covariance, adjusted for possible confounders, revealed that frequent consumption of vegetables was associated with a lower blood pressure (systolic and diastolic, F = 11.4,p< 0.001, and F = 6.8, p = 0.001, respectively), while frequent consumption of Western foods was associated with a higher diastolic blood pressure (F = 4.9, p = 0.008).There were also positive associations between frequent consumption of Western foods and total cholesterol (F = 18.0, p < 0.001) and between frequent consumption of vegetables and HDL cholesterol levels (F = 3.6, p = 0.028).The group with the lowest frequency of meat consumption exhibited the lowest total cholesterol levels (F = 7.5, p = 0.001) but the pattern was nonlinear (M.Yoshimura, personal communication).
Smoking habits were classified as "lifetime non-smoker", "exsmoker", "1-20 cigarettes per day", or "21+ cigarettes per day" for men; and "lifetime non-smoker", "ex-smoker", or "current smoker" for women.The total average amount of alcohol consumed was calculated in grams per day, after taking into account the frequency, amount, and alcohol content for specific beverages.Alcohol consumption was categorized into "non-drinker", "<1 go daily" (go; a traditional Japanese alcohol unit, 1 go = 28.9grams of alcohol), "1-3 go daily" (28.9-86.6 g), or "3+ go daily" (>=86.7 g) for men; and "non-drinker", "<1 go daily", or "1+ go daily" for women.The physical activity index, which was developed in the Framingham Study,24 was calculated by totaling the hours at each level of activity and multiplying this by a weight that was based on the oxygen consumption required for that activity.We computed the physical activity indices based on reported activity during an ordinary day taking into account physical activity at work.The index was categorized into three strata: low (<29), medium (29)(30)(31)(32)(33)(34)(35)(36), and high (37+).
Women who reported that they were in a postmenopausal stage were defined as such, whether it was natural or surgically induced.Although contraceptive use and postmenopausal hormone replacement therapy are possible modifiers of serum lipid profiles,25,26 they were not taken into consideration here because of their low prevalence among Japanese .27,28Other than those vari-ables, seasonal variations can occur in cholesterol levels.29 Only 8% (n = 409) of the blood sample was drawn during the winter season (November through February): the effect of adjusting for this was negligible.The prevalence of hypertension in those under antihypertensive therapy was low (7% for men, 8% for women); and so was that for those under medication for hyperlipidemia (1% for both men and women).Adjusting for those under medication (possible lipid-altering medications) did not alter the results; thus these variables were not considered in the analyses.
The physical examinations took place in each community.Height was measured without shoes;body weight was recorded with the subject clothed; and 0.5 kg in summer or 1 kg in the other seasons was subtracted from the recorded weight.The body mass index (BMI) was calculated as weight (kg)/height (m)2.BMI readings were categorized into tertiles, based on the total sample distribution (<21.6 kg/m2, 21.6-23.9kg/m2, or 24.0+ kg/m2).
While the subject was seated, blood samples were drawn from the antecubital vein with the minimal use of a tourniquet.Specimens were collected in siliconized vacuum glass tubes containing no additives.The tubes were centrifuged at 3,000 G for 15 minutes at room temperature.Total cholesterol was measured by an enzymatic method (Wako, Osaka, Japan).High-density lipoprotein (HDL) cholesterol was measured by the phosphotungstate precipitation method (Wako).Lipoprotein (a) levels were measured by using an enzyme-linked immunosorbent assay kit (Biopool, Uppsala, Sweden).Blood variables were measured at the Central Laboratory of SRL (Tokyo, Japan), a commercial hematology laboratory, where the measurements were all standardized by the Lipid Standardization Program, Center for Disease Control and Prevention, Atlanta, Georgia.All the lipid profiles were determined except in a single community (n = 450), where lipoprotein (a) was not measured.
As another outcome, a ratio of serum total cholesterol divided by HDL cholesterol was computed.According to Criqui and Golomb,30 this ratio provides a more precise estimate of coronary risk than each variable considered individually; and this ratio has been used in several studies as a coronary risk indicator for both men and women .25,31

Statistics
All analyses were performed separately for men and women.Means, standard deviations, and 95% confidence intervals are given for the serum lipid profile data.Then a series of cross-tabulations of job-characteristics and socio-demographic/behavioral variables was performed.The chi-square test was conducted and a linear trend for the effect of job characteristics was assessed by computing the p value for trend.Associations between job characteristics and serum lipid levels were assessed by adjusting for possible covariates, and the adjusted means for three job characteristic levels are shown.In the statistical tests, the distributions for the lipoprotein (a) and the total/HDL cholesterol ratio levels were skewed.Therefore natural logarithms were used for these variables.For comparison, however, arithmetic means are displayed.
The multivariate analyses were repeated on stratified subgroups; the pre-industrial occupations (farming, forestry and fisheries) and the other post-industrial occupations, to see if the patterns and findings observed for the full study population were the same across the occupations of interest.
Additional multivariate tests were done by using cutoff points, which were established in a guideline for Japanese that indicated the thresholds of increased risk for cardiovascular diseases .32 Logistic regression analyses were conducted using the following criteria as dependent variables: total cholesterol 220+ mg/dl, HDL cholesterol < 40 mg/dl, and lipoprotein (a) 40+ mg/dl.For the total/HDL cholesterol ratio, the upper tertile was chosen as a cutoff point related to the distribution of the data.33Categorical variables including the job characteristics were represented by dummy indicators in the analyses.
All tests were two-tailed and a value where p < 0.05 was considered statistically significant.SPSS* for Windows, release 6.1, was used for the statistical analyses.

RESULTS
The serum lipid levels of the study population are displayed in Table 1.
Table 2 shows the sex-specific population profiles according to job characteristics.In both sexes, psychological demands were higher in the younger age groups and managers than in the respective others.High demands were associated with higher levels of alcohol consumption and physical activity.Men with high demands were married and frequent meat consumers.As for women, psychological demands were more prevalent in the bluecollar and the pre-menopausal women than in the respective counterparts.Job control was lower in the less educated, blue-col-lar, and subordinates, and low control was associated with low levels of vegetable consumption, physical activity, and BMI.Men with high job control were younger, married, and had high consumption of cigarettes and western foods, while women with high control had meat frequently.Both male and female blue-collar workers were exposed to job strain more frequently than whitecollar workers.Men exposed to job strain were less educated and subordinate employees, and with high levels of meat consumption and physical activity.Women with job strain were younger and had vegetables less frequently.The analysis of covariance showed significant differences across demand levels for total cholesterol and the total/HDL cholesterol ratio in men.Higher psychological demands were associated with a higher total cholesterol level, with an adjusted difference from the top to bottom tertile of 3.3 mg/dl (F = 3.03, p = 0.048).Higher demands were also associated with a higher total/HDL cholesterol ratio (F = 3.94, p = 0.020).The linear associations were confirmed by multiple regression analyses.However, neither job control nor job strain was associated with any lipid levels.No significant associations were found in women (Table 3).
Observed findings of job demands and the both levels of total cholesterol and total/HDL cholesterol ratio in men were stronger in the post-industrial occupations (F = 2.90, p = 0.055 for total cholesterol, and F = 3.71, p = 0.025 for total/HDL cholesterol ratio, respectively).However, unexpected negative findings were also observed; male post-industrial workers with the lowest demands had the highest level of lipoprotein (a) (F = 3.38, p = 0.034) and female pre-industrial workers exposed to high strain had the highest level of HDL cholesterol (F = 5.20, p = 0.006).
Neither in the full study population nor in the stratified groups , did the logistic regression analyses reveal any significantly elevat-   Due to selective missing values, the total number of subjects included in multivariate analysis was somewhat lower.Analysis of covariance was performed adjusting for age, marital status, educational attainment, type of job, employment status, smoking hal alcohol intake, dietary patterns (vegetable pattern, meat pattern, and western pattern), physical activity index, and BMI.In women, menopausal status was also adjusted for.
ed risk of adverse job characteristics in unfavorable lipid categories (data not shown).

DISCUSSION
In a Japanese male working population, a statistically significant difference of the total cholesterol levels was observed depending on the psychological job demands.High demands also had a significant association with the total/HDL cholesterol ratio.The well-known effects of dietary habits, smoking, alcohol drinking, physical activity, and relative weight on atherogenic lipids 25,34-37 were taken into account in the statistical model.However, because the sample analyzed here appeared to have favorable lipid profiles,30,38 the magnitude of the observed effect, up to a 3% average difference of the total cholesterol levels, is not biologically meaningful.Besides, neither job control nor a combination of high demands and low control, represented by their ratio in this study, had expected associations with any lipid profile.Stratified analyses provided inconsistent results.These findings replicated most of those that appeared in previous studies,11-17 which show no significant associations between job strain or its major components and atherogenic lipids.
Our study was unique in that the study population was recruited from rural residents and included large numbers of pre-industrial workers.However, because of the sampling method, relatively older workers represented the study population.It was possible that the older workers, whose prevalence for hyperlipidemia were presumably high, accounted for the relation between psychological demands and high total cholesterol.In fact , the magnitude of the association between psychological demands and total cholesterol in men was stronger for the older workers (F = 3 .46,p = 0.032 for workers aged 50-65 years, and F = 2 .56,p = 0.078 but the pattern was nonlinear for workers aged 49 years and under , respectively).However, except for lipoprotein (a) , mean lipid levels were generally better in the older men than in the others (the opposite was the case with female workers; data were not shown) .Thus, at least in this male working population , the effect of age distribution for the lipid profiles on the results is not considered so large.However, to fill the gap of the knowledge , further studies are necessary including younger workers who appeared to be exposed to severer job characteristics (except for job control) in this study.
Failure to examine some possible biological variability in an individual's cholesterol response could be a reason for the lack of significant associations in this study.First, individuals with cardiovascular problems were under-represented .38 This means that the association, if any, between adverse job characteristics and unfavorable lipid profiles may have been underestimated .Cholesterol levels seem highly labile in some individuals5 and those with a high coronary risk may be susceptible to stressinduced alterations in their lipid metabolism.6,39Second .lipid elevation appears to be more significantly affected by perceived stress than an objective stressful situation,40 whereas the participants were questioned about their job characteristics, not about their feelings about stress.41Another possible explanation for the null association is relatively lower job demands level of our selfemployed sample in the pre-industrial occupations.22,24The job demands and job control for such an occupational group has been reported to be greater when compared to other occupations.4 It should be noted that the job control dimension had an unexpectedly low Chronbach's alpha of 0.64, which may also have affected the results negatively.Lastly, because research has suggested that health behavior is in the causal pathway between job strain and cardiovascular disease and/or cardiovascular disease risks,42 adjusting for all the covariates would represent over-adjustment and an underestimating of associations.Such an effect was unlikely, however, since the univariate and age-adjusted analyses revealed almost the same positive associations as in the final results.
Contrary to the men whose job characteristics were found to be associated, though weakly, with poor lipid profiles, no positive associations were found for the women.Attitudes toward work and/or perception/expression of job characteristics may be different between genders.The proportion of female employees from these rural settings who sought to develop a career in their occupational activities was not ascertained.Another explanation is that there may be physiological differences related to sex in one's lipid response to a stressful situation .41 We could not take into account the part-time employment, too.It was plausible that more women than men were employed as part time, since a quarter of the women and 11% of the men reported (the questionnaire on the daily activity) to work totally less than 8 hours per day, and this uneven distribution caused, at least in part, gender difference of the findings.We do not know any reports showing the psychometric properties of the demandcontrol questionnaire in part-time workers.Based on our proxy categories by self-reported working hours, inferred reliability of the control scale was slightly lower in those working less than 8 hours per day than in the counterparts (Chronbach's alpha = 0.63 for the former, and 0.65 for the latter, respectively).The coefficient levels of psychological demands were not different between the groups (0.67 for both).Since the demand-control questionnaire tries to capture objective work environment,41 it is unlikely that the measures produce considerable deviation from the constructs.However, we cannot deny that social roles outside of work interact with job characteristics differently for the employment status.
Although job strain or other forms of chronic work stress seem to have little impact on lipid profiles,6 more precise study designs should be adopted before definite conclusions can be drawn for this study question.Prospective studies using lipid levels as outcomes are necessary.Most epidemiologic studies investigating the association (such as ours) were based on a cross-sectional design.With regard to this study question, prospective evidence exists showing that workplace demands are a predictor for the progression of carotid atherosclerosis.39It was also plausible that the stronger effect observed in older men was due to a cumulative effect of adverse job characteristics on athelogenic lipid, because changing job in rural communities was not considered so often as in urban areas.Such an effect can be scrutinized only by repeated measures both of job characteristics and lipid profiles.Interventional studies would also provide a clearer picture.That work stress intervention (i.e., increasing job control) may have a lipid lower effect has been demonstrated.44 In conclusion, this cross-sectional analysis indicates a possible association between a psychologically demanding job and an unfavorable lipid profile in Japanese male workers.However, at least in the rural setting, the logical basis for the application of the demand-control model to atherogenic lipids is weak.
SD: standard deviation CI: confidence interval.

Table 2 .
Socio-demographic and behavioral profiles according to job characteristics, active workers aged 65 and under , J MS Cohort Study,1992-1995.Chi-square test was conducted for linear trends.The numbers do not add up to the total number of subjects due to missing values.

Table 3 .
Associations between job characteristics and serum lipid profiles, active workers aged 65 and under, JMS Cohort Study , 1992-1995.