This Week in BMJ | Editor's Choice | Press releases | Advertisement details

Paper Saturday 22 February 1997

Low job control and risk of coronary heart disease in Whitehall II (prospective cohort) study

Hans Bosma, Michael G Marmot, Harry Hemingway, Amanda C Nicholson, Eric Brunner, Stephen A Stansfeld

See editorial by Haines and Smith

Abstract

Objective: To determine the association between adverse psychosocial characteristics at work and risk of heart disease among male and female civil servants

Design: Prospective cohort study (Whitehall II study). At the baseline examination (1985-8) and twice during follow up a self report questionnaire provided information on psychosocial factors of the work environment and coronary heart disease. Independent assessments of the work environment were obtained from personnel managers at baseline. Mean length of follow up was 5.3 years.

Setting: London based office staff in 20 civil service departments.

Subjects: 10,308 civil servants aged 35-55 were examined - 6,985 men (67%) and 3,413 women (33%).

Main outcome measures: New cases of angina (Rose questionnaire), sever pain across the chest, diagnosed ischaemic heart disease, and any coronary event.

Results: Men and women with low job control, either self reported or independently assessed, had a higher risk of newly reported coronary heart disease during follow up. Job control assessed on two occasions three years apart although intercorrelated, had cumulative effects on newly reported disease. Subjects with low job control on both occasions had an odds ratio for any subsequent coronary event of 1.93 (95% confidence interval 1.34 to 2.77) compared with subjects with high job control at both occasions. This association could not be explained by employment grade, negative affectivity, or classic coronary risk factors. Job demands and social support at work were not related to the risk of coronary heart disease.

Conclusions: Low control in the work environment is associated with an increased risk of future coronary heart disease among men and women employed in government offices. The cumulative effect of low job control assessed on two occasions indicates that giving employees more variety in tasks and a stronger say in decisions about work may decrease the risk of coronary heart disease.

Introduction

In the first Whitehall study men in the lower grades of the British civil service had nearly three times the 10 year risk of mortality from coronary heart disease compared with men in the higher grades.1 Less than half of this gradient could be explained by accepted coronary risk factors, and we speculated that psychosocial factors may provide some of the missing predictive power. As we examined differences in employment grade with coronary heart disease, psychosocial work characteristics became a logical subject to study.2

Studies using Karasek's job strain model have shown higher cardiovascular risk in jobs characterised by high demands and low control.3-5 An analysis of British occupational mortality suggested that high demands were less important predictors of mortality than low control.2 In the Whitehall II study people in the high grades, who have lower coronary risk, have higher demands than the low grades. Low control therefore seemed to be more promising.6-7

The longitudinal phase of the Whitehall II study allowed us to examine the relation between the psychosocial work environment and newly reported cases of coronary heart disease. In particular, it enabled us to test whether low control is more strongly related to coronary heart disease than is the full job strain model. Furthermore, this study allowed us to examine both men and women5; the possible role of reporting bias (using a measure of negative affectivity)8-11; and the cumulative effects of adverse work characteristics by using measures of low control assessed on two occasions.12-13 A methodological advance is the use of an independent measure of control. By having an independent assessment as well as a self report of the work environment, the study could address the question of whether job stress is influenced by (subjective) perceptions or by more objective appraisals of the work, or by both.14

Subjects and methods

Study population
The Whitehall II study is the sequel to the first Whitehall study, which began in 1967.1 15 The Whitehall II study was set up primarily to investigate the degree and causes of the social gradient in illness rates; to study work characteristics, social support, and additional factors related to the gradient in mortality; and, importantly, to include women. In the study a new cohort of civil servants was established between 1985 and 1988 (phase 1). All male and female civil servants aged between 35 and 55 in 20 London based civil service departments were sent an introductory letter and screening questionnaire and were offered a screening examination for cardiovascular diseases. The response rate was 73%. The true response rate would probably have been higher, however, because about 4% of the civil servants on the lists provided by the civil service had moved before the study and were therefore not eligible for inclusion. In total, 10,308 civil servants were examined - 6895 men (67%) and 3,413 women (33%).

After the initial participation in phase 1 the participants were approached again in 1989-90 (phase 2: postal questionnaire) and in 1991-3 (phase 3: postal questionnaire and screening examination). The participation rates at these two phases were 79% and 83% respectively; 7,372 subjects (72%) participated in all three phases and 9302 subjects (90.2%) participated in either phase 2 or 3. Furthermore, although still eligible for participation, 1286 subjects (12.5%) had left the civil service before phase 3. The length of follow up was 5.3 years on average, with a range of 3.7 to 7.6 years. Full details of the screening examinations are reported elsewhere.7 16

Coronary heart disease
Four indicators of coronary heart disease were analysed: angina, severe pain across the chest, diagnosed ischaemic heart disease, and any coronary event. Angina was measured by the Rose questionnaire and defined as pain located over the sternum or in both the left chest and the left arm that is precipitated by exertion, that causes the person to stop, and that goes away in 10 minutes or less.17 Severe pain across the chest was defined as having ever had a severe pain across the front of the chest lasting half an hour or more. Diagnosed ischaemic heart disease depended on whether the subject reported that a general practitioner or hospital doctor had ever suspected or confirmed myocardial infarction or angina. Subjects could report more than one of these outcomes. Reports of any of these three outcomes were coded as any coronary event. The outcomes were assessed at all three phases.

Work characteristics
Two methods were used to assess the psychosocial work environment: self reports by the civil servants and ratings by personnel managers (independent assessments). The self report items were derived from well known questionnaires for the central components of the job strain model - that is, job demands, job control, and social support (appendix).3-4 18-19 As the results with specific subconstructs, such as skill discretion and social support from colleagues, did not differ from the results with the three main constructs, we focused on the main constructs (job control, job demands, social support). After all items were recoded in the same direction scores for each scale were calculated as the sum of the item scores. Subjects who did not answer one item in a scale were assigned an average score based on the items that they did answer. Self reported work characteristics were available at all three phases. In addition, at phase 1 personnel managers assessed each job in terms of work pace, the importance of not making mistakes, conflicting demands, and the degree of control (appendix). In 18 out of 20 departments 140 well informed personnel managers undertook these ratings. Detailed information was obtained on individual jobs because 5766 different jobs were filled by 8838 subjects.

Statistical analysis
The logistic regression model was used to estimate the strength of the association between the psychosocial work characteristics in thirds at phase 1 and the occurrence of newly reported coronary heart disease at phase 2 or 3. Tests for trends were also performed by modelling the group scores of each work characteristic (1, 2, 3) as one variable. The corresponding odds ratio reflects the increase in the odds of new coronary heart disease per unit increase in this variable. To examine their independent impact on new coronary heart disease, self reported and independently assessed work characteristics were simultaneously controlled. By using both phase 1 and 2 self reported work characteristics as predictive factors, further insight was obtained into their cumulative effects on coronary heart disease reports at phase 3 (phase 1 and 2 were about three years apart). Several multivariate logistic regression models provided further insight into the extent to which adverse work characteristics affected the future reporting of coronary heart disease, independent of employment grade, negative affectivity, and classic coronary risk factors.

Negative affectivity is the disposition to respond negatively to questionnaires and may inflate correlations between self reported work characteristics and self reported disease.8-11 Negative affectivity was measured with the negative affect subscale of the affect balance scale.20-21

Coronary risk factors included smoking (never smoked, stopped smoking, smoked 1-10, 11-20, or 21 or more cigarettes daily), cholesterol concentration (mmol/l), diastolic blood pressure (mm Hg), drug treatment for high blood pressure, and body mass index (kg/m²). Ordinal variables, such as employment grade, were represented by dummy indicators in the analyses. All logistic regression analyses were adjusted for age and length of follow up. Baseline cases were excluded in all analyses. In the analyses using work characteristics during phases 1 and 2 cases of coronary heart disease at both phase 1 and 2 were excluded.

Results

Table 1 presents the number of baseline and new cases of all four outcomes. Women reported angina and any coronary event more often than men. Men more often reported that a doctor had confirmed coronary heart disease. Women reported severe chest pain less often than men at phase 1. During follow up they reported severe chest pain more often than men.

Table 1 - Numbers (percentages) of baseline and new cases of angina, severe chest pain, diagnosed ischaemic heart disease, and any coronary event among 6,895 men and 3,413 women
	Phase 1	Phase 2 or 3*
Angina:
Men	164 (2.4)	177 (3.7)
Women	134 (4.0)	151 (7.0)
Severe chest pain:
Men	433 (6.3)	258 (5.5)
Women	193 (5.7)	145 (6.8)
Diagnosed ischaemic heart disease:
Men	84 (1.2)	124 (2.5)
Women	15 (0.4)	42 (1.8)
Any coronary event:
Men	595 (8.7)	401 (8.8)
Women	319 (9.5)	253 (12.3)
*Excluding baseline cases at phase 1 and non-participants at phase 2 or 3.

Job demands and social support were not related to any of the outcomes. The odds ratio of any coronary event (trend test) for job demands (low, intermediate, high) was 0.97 (95% confidence interval 0.85 to 1.12) and 1.17 (0.98 to 1.41) for men and women, respectively. The independent assessments of job demands showed similar negative findings. Work support (high, intermediate, low) had the following odds ratios: 1.11 (0.96 to 1.28) and 1.15 (0.95 to 1.38). Furthermore, there was no consistent pattern across the outcomes. Multiplicative interactions between job demands, job control, and social support did not add to the prediction of new coronary heart disease. Therefore, only the results for job control are presented.

Table 2 shows that there were striking grade related differences in job control. The mean self reported job control was 47 and 78 for the lowest and highest employment grade, respectively. Men had higher scores on self reported job control than women. However, within employment grades there were hardly any differences between men and women. Similar findings were found for independently assessed job control.

Table 2 - Mean self reported job control by sex and employment grade at phase 1
	Men		Women
	No	Mean (SD) score*	No	Mean (SD) score*
Employment grade:
1 (High)	1,007	78.3 (9.9)	117	77.3 (10.6)
2	1,627	73.8 (11.1)	261	73.4 (11.6)
3	1,223	71.5 (11.1)	197	71.8 (11.8)
4	1,490	66.6 (12.3)	475	67.2 (12.5)
5	879	61.2 (13.8)	655	59.5 (14.0)
6 (Low)	634	46.5 (17.0)	1,635	47.7 (16.1)
Total	6,860	68.4 (15.0)	3,340	57.2 (18.0)
*Continuous self reported job control was rescaled to variable ranging from 0 (low) to 100 (high).

The odds ratios indicated about 50% higher risks of any new report of coronary heart disease report at phase 2 or 3 for men with low job control compared with men with high job control at phase 1 (table 3). The odds ratio for women was about 1.70. The odds ratio of low self reported job control and doctor diagnosed ischaemia among women (odds ratio 0.85) was an exception to the general inverse association. This may reflect the small number of newly reported diagnosed ischaemic heart disease among women (n=36). The associations of self reported job control and independently assessed job control with reported coronary heart disease were about equally strong.

Table 3 - Odds ratios (95% confidence intervals)* of newly reported coronary heart disease at phase 2 or 3 by self reported and independently assessed job control at phase 1 in men and women
	Angina	Severe chest pain	Diagnosed ischaemic heart disease	Any coronary event
Self report
Men:
High	1.00	1.00	1.00	1.00
Intermediate	1.16 (0.81 to 1.68)	0.89 (0.65 to 1.20)	1.37 (0.89 to 2.11)	1.05 (0.81 to 1.34)
Low	1.54 (1.05 to 2.26)	1.33 (0.97 to 1.82)	1.60 (1.01 to 2.55)	1.55 (1.20 to 2.01)
Test for trend+	1.24 (1.02 to 1.50)	1.13 (0.96 to 1.33)	1.27 (1.01 to 1.60)	1.23 (1.08 to 1.41)
No of men (events)	4,812 (168)	4,683 (250)	5,021 (118)	4,522 (384)
Women:
High	1.00	1.00	1.00	1.00
Intermediate	1.16 (0.70 to 1.94)	2.01 (1.12 to 3.61)	0.35 (0.11 to 1.04)	1.80 (1.16 to 2.79)
Low	1.20 (0.74 to 1.92)	1.80 (1.02 to 3.16)	0.85 (0.38 to 1.87)	1.74 (1.15 to 2.64)
Test for trend+	1.08 (0.86 to 1.36)	1.23 (0.96 to 1.56)	1.03 (0.67 to 1.59)	1.23 (1.03 to 1.49)
No of women (events)	2,107 (138)	2,093 (123)	2,240 (36)	2,001 (225)
Independent assessment
Men:
High	1.00	1.00	1.00	1.00
Intermediate	1.43 (0.94 to 2.18)	0.87 (0.61 to 1.24)	1.15 (0.70 to 1.89)	1.16 (0.87 to 1.55)
Low	1.50 (1.00 to 2.25)	1.10 (0.79 to 1.53)	1.33 (0.83 to 2.12)	1.43 (1.09 to 1.88)
Test for trend+	1.21 (0.99 to 1.48)	1.05 (0.89 to 1.24)	1.15 (0.91 to 1.45)	1.20 (1.04 to 1.37)
No of men (events)	4,165 (149)	4,054 (212)	4,347 (106)	3,913 (331)
Women:
High	1.00	1.00	1.00	1.00
Intermediate	1.08 (0.58 to 1.99)	1.03 (0.53 to 1.99)	0.68 (0.18 to 2.56)	1.15 (0.70 to 1.88)
Low	1.46 (0.87 to 2.43)	1.70 (0.98 to 2.93)	1.43 (0.53 to 3.85)	1.73 (1.14 to 2.62)
Test for trend+	1.23 (0.96 to 1.57)	1.36 (1.04 to 1.78)	1.31 (0.79 to 2.17)	1.35 (1.10 to 1.65)
No of women (events)	1,838 (122)	1,833 (107)	1,956 (31)	1,756 (200)
*Adjusted for age and length of period between phase 1 and 3; baseline cases at phase 1 were excluded. +Group scores of job control (1, 2, 3) were modelled as one (ordinal) variable; odds ratio reflects increase in odds of new coronary heart disease per unit increase in this variable.

Because the odds ratios for men and women did not differ significantly, further analyses were based on  the total sample (sex was controlled for in each model). Self reported job control and independently assessed job control were not highly correlated (Pearson correlation coefficient 0.41). The associations of self reported and independently assessed job control at phase 1 with future reported coronary heart disease seemed to be independent from each other (table 4). Furthermore, both assessment instruments had similarly strong associations with the outcomes. Simultaneously adjusted, the odds ratios of any new coronary event for subjects with low self reported job control or low independently assessed job control were both 1.40. There was no indication that subjects with discrepancies between self reported and independently assessed job control had any particular excess risk. The presented odds ratios may be underestimated because both measurements probably reflected the same work characteristic.

Table 4 - Odds ratios (95% confidence intervals)* of newly reported coronary heart disease at phase 2 or 3 by self reported and independently assessed job control at phase 1 in whole sample (men and women combined), unadjusted (model I) and adjusted for other assessment instrument (model II)
	Angina	Severe chest pain	Diagnosed ischaemic heart disease	Any coronary event
Self report
Model I:
High	1.00	1.00	1.00	1.00
Intermediate	1.16 (0.87 to 1.56)	1.07 (0.82 to 1.39)	1.13 (0.75 to 1.68)	1.21 (0.98 to 1.49)
Low	1.33 (0.99 to 1.79)	1.32 (1.02 to 1.72)	1.48 (0.99 to 2.22)	1.50 (1.21 to 1.85)
Test for trend+	1.15 (1.00 to 1.34)	1.15 (1.01 to 1.32)	1.22 (0.99 to 1.49)	1.22 (1.10 to 1.36)
No of subjects (events)	6,919 (306)	6,776 (373)	7,261 (154)	6,523 (609)
Model II:
High	1.00	1.00	1.00	1.00
Intermediate	1.18 (0.86 to 1.63)	1.08 (0.81 to 1.44)	0.99 (0.54 to 1.53)	1.19 (0.95 to 1.50)
Low	1.24 (0.89 to 1.73)	1.27 (0.94 to 1.71)	1.28 (0.82 to 2.00)	1.39 (1.10 to 1.77)
Test for trend+	1.11 (0.94 to 1.30)	1.14 (0.98 to 1.32)	1.14 (0.91 to 1.42)	1.18 (1.05 to 1.33)
No of subjects (events)	5,952 (266)	5,838 (315)	6,247 (136)	5,620 (523)
Independent assessment
Model I:
High	1.00	1.00	1.00	1.00
Intermediate	1.28 (0.91 to 1.81)	0.91 (0.67 to 1.24)	1.08 (0.68 to 1.71)	1.15 (0.89 to 1.47)
Low	1.47 (1.07 to 2.02)	1.23 (0.94 to 1.62)	1.38 (0.91 to 2.09)	1.51 (1.21 to 1.89)
Test for trend+	1.21 (1.03 to 1.41)	1.12 (0.98 to 1.29)	1.18 (0.96 to 1.46)	1.24 (1.11 to 1.38)
No of subjects (events)	6,003 (271)	5,887 (319)	6,303 (137)	5,669 (531)
Model II:
High	1.00	1.00	1.00	1.00
Intermediate	1.26 (0.89 to 1.78)	0.90 (0.66 to 1.23)	1.06 (0.67 to 1.69)	1.14 (0.88 to 1.46)
Low	1.38 (1.00 to 1.92)	1.18 (0.89 to 1.56)	1.27 (0.82 to 1.96)	1.40 (1.11 to 1.76)
Test for trend+	1.17 (1.00 to 1.37)	1.09 (0.95 to 1.27)	1.13 (0.91 to 1.41)	1.18 (1.05 to 1.33)
No of subjects (events)	5,952 (266)	5,838 (315)	6,247 (136)	5,620 (523)
*Adjusted for age, sex, and length of period between phase 1 and 3; baseline cases at phase 1 were excluded. +Group scores of job control (1, 2, 3) were modelled as one (ordinal) variable; odds ratio reflects increase in odds of new coronary heart disease per unit increase in this variable.

Self reported job control at phase 1 and 2 were independently associated with newly reported coronary heart disease at phase 3 (table 5). Although the odds ratios were not significant, the findings indicate that job control in phases 1 and 2 had cumulative effects on new coronary heart disease. This implies that subjects with low job control on both occasions had the highest risks of new coronary heart disease, whereas subjects with high job control on both occasions had the lowest risks. Subjects with stable intermediate job control or subjects who changed from low to high job control or vice versa generally had intermediate risks of coronary heart disease.

Table 5 - Odds ratios (95% confidence intervals)* of newly reported coronary heart disease at phase 3 by self reported job control at phase 1 and phase 2 (simultaneously controlled) and by average job control at phase 1 and 2 in total sample (men and women combined)
	Angina	Severe chest pain	Diagnosed ischaemic heart disease	Any coronary event
Job control at phase 1:
High	1.00	1.00	1.00	1.00
Intermediate	1.29 (0.78 to 2.15)	0.90 (0.57 to 1.42)	1.09 (0.59 to 2.02)	1.22 (0.84 to 1.75)
Low	1.48 (0.83 to 2.64)	1.20 (0.70 to 2.04)	1.16 (0.54 to 2.48)	1.40 (0.92 to 2.13)
Test for trend+	1.21 (0.91 to 1.61)	1.09 (0.83 to 1.43)	1.08 (0.74 to 1.57)	1.18 (0.95 to 1.46)
Job control at phase 2:
High	1.00	1.00	1.00	1.00
Intermediate	1.27 (0.78 to 2.08)	1.34 (0.86 to 2.09)	1.09 (0.59 to 2.02)	1.57 (1.10 to 2.24)
Low	1.45 (0.82 to 2.56)	1.18 (0.67 to 2.08)	1.41 (0.54 to 2.48)	1.36 (0.88 to 2.10)
Test for trend+	1.20 (0.90 to 1.59)	1.10 (0.84 to 1.45)	1.18 (0.81 to 1.73)	1.16 (0.94 to 1.44)
Average job control at phases 1 and 2‡:
High	1.00	1.00	1.00	1.00
Intermediate	1.36 (0.83 to 2.23)	1.36 (0.88 to 2.10)	1.39 (0.79 to 2.45)	1.71 (1.20 to 2.43)
Low	2.09 (1.29 to 3.37)	1.52 (0.96 to 2.38)	1.49 (0.81 to 2.74)	1.93 (1.34 to 2.77)
Test for trend+	1.46 (1.15 to 1.85)	1.23 (0.98 to 1.53)	1.22 (0.91 to 1.65)	1.36 (1.15 to 1.62)
No of subjects (events)	6,565 (132)	6,372 (136)	6,982 (73)	5,999 (231)
*Adjusted for age, sex, and length of period between phase 1 and 3; baseline cases at phase 1 and phase 2 were excluded. +Group scores of job control (1, 2, 3) were modelled as one (ordinal) variable; odds ratio reflects increase in odds of new coronary heart disease per unit increase in this variable. ‡Sum of continuous job control scales at phase 1 and 2, with resulting score divided into thirds.

The continuous job control scales at phase 1 and 2 were summed and divided into thirds to reflect the cumulative effects of low job control during phases 1 and 2. Subjects with low job control on average had an odds ratio for any subsequent coronary event of 1.93 (1.34 to 2.77) compared with subjects with high job control on average. This is primarily caused by the high odds ratios for angina. The odds ratios for severe chest pain and diagnosed ischaemic heart disease were smaller (about 1.50).

The odds ratios for average job control slightly decreased after adjustment for employment grade, negative affectivity, and classic coronary risk factors, but they continued to be significant for angina and any coronary event (table 6). The decrease in the odds ratios for diagnosed ischaemic heart disease when the coronary risk factors were taken into account (odds ratio 1.26) was primarily caused by the higher prevalence of subjects having drug treatment for hypertension in the group reporting low job control. When employment grade at phase 2 was additionally controlled for in the model already controlling for employment grade at phase 1 the odds ratios did not change substantially.

Table 6 - Odds ratios (95% confidence intervals)* of newly reported coronary heart disease at phase 3 by average self reported job control at phase 1 and 2+ in total sample (men and women combined) adjusted for age, sex, and length of period between phase 1 and phase 3 (model I), and separately adjusted for employment grade (model II), negative affectivity (model III), and coronary risk factors (model IV)
	Model I	Model II	Model III	Model IV
Angina:
High	1.00	1.00	1.00	1.00
Intermediate	1.36 (0.83 to 2.23)	1.20 (0.71 to 2.03)	1.33 (0.81 to 2.18)	1.44 (0.86 to 2.39)
Low	2.09 (1.29 to 3.37)	1.80 (1.03 to 3.14)	2.02 (1.25 to 3.27)	2.02 (1.22 to 3.34)
Test for trend‡	1.46 (1.15 to 1.85)	1.36 (1.03 to 1.80)	1.44 (1.13 to 1.82)	1.42 (1.11 to 1.82)
No of subjects (events)	6,565 (132)	6,565 (132)	6,565 (132)	6,228 (123)
Severe chest pain:
High	1.00	1.00	1.00	1.00
Intermediate	1.36 (0.88 to 2.10)	1.31 (0.83 to 2.06)	1.28 (0.83 to 1.98)	1.48 (0.94 to 2.31)
Low	1.52 (0.96 to 2.38)	1.38 (0.81 to 2.34)	1.38 (0.87 to 2.17)	1.52 (0.95 to 2.45)
Test for trend‡	1.23 (0.98 to 1.53)	1.17 (0.90 to 1.53)	1.17 (0.94 to 1.46)	1.23 (0.97 to 1.54)
No of subjects (events)	6,372 (136)	6,372 (136)	6,372 (136)	6,028 (128)
Diagnosed ischaemic heart disease:
High	1.00	1.00	1.00	1.00
Intermediate	1.39 (0.79 to 2.45)	1.42 (0.78 to 2.56)	1.35 (0.76 to 2.38)	1.36 (0.77 to 2.41)
Low	1.49 (0.81 to 2.74)	1.39 (0.67 to 2.86)	1.42 (0.77 to 2.62)	1.26 (0.67 to 2.39)
Test for trend‡	1.22 (0.91 to 1.65)	1.19 (0.84 to 1.69)	1.19 (0.88 to 1.61)	1.13 (0.83 to 1.54)
No of subjects (events)	6,982 (73)	6,982 (73)	6,982 (73)	6,617 (69)
Any coronary event:
High	1.00	1.00	1.00	1.00
Intermediate	1.71 (1.20 to 2.43)	1.59 (1.10 to 2.30)	1.64 (1.15 to 2.34)	1.81 (1.26 to 2.60)
Low	1.93 (1.34 to 2.77)	1.81 (1.18 to 2.73)	1.82 (1.26 to 2.63)	1.99 (1.36 to 2.91)
Test for trend‡	1.36 (1.15 to 1.62)	1.32 (1.08 to 1.62)	1.33 (1.11 to 1.58)	1.38 (1.15 to 1.65)
No of subjects (events)	5,999 (231)	5,999 (231)	5,999 (231)	5,681 (218)
*Cases of coronary heart disease at phase 1 and 2 were excluded. +Sum of continuous job control scales at phase 1 and 2, with resulting score divided into thirds. ‡Group scores of job control (1, 2, 3) were modelled as one (ordinal) variable; odds ratio reflects increase in odds of new coronary heart disease per unit increase in this variable.

Discussion

The results of the Whitehall II study provide evidence that particular psychosocial factors may account for some of the missing predictive power for coronary heart disease. Excess risks of coronary heart disease were found for male and female British civil servants in jobs characterised by low control. Self reported and independently assessed job control showed roughly similar associations with coronary heart disease. Self reported job control assessed on two occasions had cumulative effects on new coronary heart disease. The association was independent of employment grade, negative affectivity, and conventional coronary risk factors. Although small numbers did not permit extensive analyses of the associations between employment grade, job control, and coronary heart disease, the findings showed that low job control had adverse effects in all employment grades (data not shown). This implies that the association between job control and coronary heart disease was not confounded by employment grade. The relative contribution of work and other factors to the association between grade and coronary heart disease will be investigated in future analyses.

High job demands, low social support, and the interactions between work characteristics (job strain) were not related to the coronary outcomes. The findings correspond to the review by Schnall et al, in which 17 out of 25 studies found significant associations between job control and cardiovascular outcome, whereas associations with job demands were significant in only eight out of 23 studies.5 The importance of job control was further elaborated by several other investigators.22-24 Specific characteristics of our sample of white collar workers may have contributed to the negative findings for high job demands and high job strain. High job demands were more common in the higher employment grades, and high job demands and high job control were positively associated, resulting in comparatively few high strain jobs.

The similar associations between job control and coronary heart disease in men and women support the conclusion by Schnall et al that psychosocial work characteristics as yet seem to have no sex specific effects on coronary heart disease.5 However, this should be interpreted cautiously because angina and severe chest pain reported with the Rose questionnaire may be differently related to underlying coronary heart disease in men and women.25-26 Given that people with low job control, angina, severe chest pain, or diagnosed ischaemic heart disease at phase 1 had lower participation rates at phase 2 or 3, the impact of job control on newly reported coronary heart disease is probably somewhat underestimated in the analyses. Controlling for whether people had left the civil service did not affect the results.

Changing job control
It is still unclear how long, with what intensity, and how frequently subjects have to be exposed to stress at work before their health becomes damaged.12 Whether previous exposure to adverse work environments is mitigated when people move to positive work environments is not known. The Whitehall II study provides evidence that the risk profile may change during individual job trajectories. In general, subjects who changed from a work environment characterised by high job control to a work environment with low job control or vice versa had intermediate risks of coronary heart disease. Subjects who had stable work environments with low job control had the highest risks, while subjects with stable high job control tended to have the lowest risk. These findings may point to the usefulness of measuring job stress as a function of intensity, frequency, and possibly duration of adverse work characteristics during the whole occupational career.12

Methodological issues
Given the known variability in reporting of angina,27 which has been replicated in our data set, the use of new reports of angina or severe chest pain as an indicator of incident coronary disease may be problematic. The underlying condition may not have altered but the tendency to report might have changed. A new report of a diagnosed disease is likely to be a better indicator of new disease, although other factors may also influence both recall of diagnosis or access to medical care. Preliminary results show that 87% of the 188 subjects reporting a myocardial infarction at phase 3 had documented coronary heart disease. Furthermore, the classic coronary risk factors were related to all four outcomes, suggesting that the outcomes reflect coronary heart disease and not only reporting bias. Despite the different sensitivity and specificity of the outcomes, job control shows consistent effects. This supports an aetiological hypothesis. The credibility of this hypothesis is further strengthened by the finding in baseline data of the Whitehall II study that fibrinogen concentrations were raised in men and women with independently assessed low job control.28 This finding points to a possible psychophysiological mechanism relating low job control to coronary heart disease. Future analyses will examine the effects of low job control on fatal and non-fatal myocardial infarction.

Key messages

Low job control in the work environment contributes to the development of coronary heart disease among British male and female civil servants The risk of heart disease is associated with both objective low job control and perceived low job control. Increase in job control over time decreases the risk of coronary heart disease. This suggests that policies giving people a stronger say in decisions about their work or providing them with more variety in work tasks may contribute to better cardiovascular health

Information bias is a potential source of bias in our study, because information on both job control and coronary heart disease was obtained from self reports.29-32 This bias may have caused overestimated odds ratios because a complaining attitude towards work and health (negative affectivity) may have resulted in negative reports about both job control and coronary heart disease. However, because baseline cases were excluded in the longitudinal study and a measure of negative affectivity was controlled for, negative affectivity was unlikely to have biased the results for the participants' reports.8-11 Furthermore, the association between the independent assessment of job control and newly reported coronary heart disease was of about equal strength. The only partial agreement between self reported and independently assessed job control suggests that these assessment instruments provide partially complementary information on objective job control. However, the possibility that perceived job control was influenced by need for control or indicators of low status control, such as job insecurity, cannot be excluded.33-34

Conclusion
Low control in the work environment is associated with an increased risk of coronary heart disease among men and women employed in government offices. The fact that independently assessed low job control is as important as self reported low job control points to the relevance of objective low job control, not just appraisal or perception of low job control. The cumulative effect of low job control assessed on two occasions implies that giving subjects more variety in tasks and a stronger say in decisions related to work could have benefits for public health. The results add to the body of work linking psychosocial work characteristics to the risk of coronary heart disease.

We thank all participating civil service departments and their welfare and personnel officers, the Civil Service Occupational Health Service and their directors, Dr Elizabeth McCloy, Dr George Sorrie, Dr Adrian Semmence, and all participating civil servants.

Funding: The work presented in this paper was supported by grants from the Medical Research Council, British Heart Foundation, National Heart Lung and Blood Institute (2 RO1 HL36310), Agency for Health Care Policy Research (5 RO1 HS06516), Health and Safety Executive, the Institute for Work and Health, Toronto, Ontario, and the John D and Catherine T MacArthur Foundation Research Network on Successful Midlife Development. MGM is supported by a MRC research professorship. HB is supported by grants from the EU BIOMED network Socioeconomic variations in cardiovascular disease in Europe: the impact of the work environment (heart at work).

Conflict of interest: None.

Appendix

Self reports of the work environment
Three characteristics of the work environment - job control, job demands, and social support - were assessed by means of 25 items. Response categories ranged from 1 (often) to 4 (never).

Job control - Nine of the 15 items for job control covered decision authority and six covered skill discretion; these subscales were equally weighted. Cronbach's |ga=0.84 (measure of internal consistency). The nine items for decision authority were Do you have a choice in deciding how you do your job? Do you have a choice in deciding what you do at work? Others take decisions concerning my work; I have a good deal of say in decisions about work; I have a say in my own work speed; my working time can be flexible; I can decide when to take a break; I have a say in choosing with whom I work; and I have a great deal of say in planning my work environment. The six items for skill discretion were Do you have to do the same thing over and over again? Does your job provide you with a variety of interesting things? Is your job boring? Do you have the possibility of learning new things through your work? Does your work demand a high level of skill or expertise? Does your job require you to take the initiative?

Job demands - Cronbach's alpha=0.67 for job demands, which had four items: Do you have to work very fast? Do you have to work very intensively? Do you have enough time to do everything? Do different groups at work demand things from you that you think are hard to combine?

Social support - Cronbach's alpha=0.79 for social support, which had six items: How often do you get help and support from your colleagues? How often are your colleagues willing to listen to your work related problems? How often do you get help and support from your immediate superior? How often is your immediate superior willing to listen to your problems? Do you get sufficient information from line management (your superiors)? Do you get consistent information from line management (your superiors)?

Independent assessments of the work environment
Response categories for independent assessments of the work environment ranged from 1 (often) to 12 (never). There were four items: How often does the job involve working very fast? How often is it extremely important to do the work without mistakes? How often do different groups at work demand things which are difficult to combine? How often does the job permit complete discretion and independence in determining how, and when, the work is to be done?

References

1 Marmot M G, Shipley M J, Rose G. Inequalities in death - specific explanations of a general pattern? Lancet 1984;i:1003-6.

2 Marmot M, Theorell T. Social class and cardiovascular disease. The contribution of work. Int J Health Services 1988;18:659-74.

3 Karasek R A. Job demands, job decision latitude, and mental strain: implications for job redesign. Administrative Science Quarterly 1979;24:285-308.

4 Karasek R A, Theorell T. Healthy work: stress, productivity and the reconstruction of working life. New York: Basic Books, 1990.

5 Schnall P L, Landsbergis P A, Baker D. Job strain and cardiovascular disease. Ann Rev Public Health 1994;15:381-411.

6 Johnson J V, Hall E M. Class, work, and health. In: Amick III B C, Levine S, Tarlov A R, Chapman Walsh D, eds. Society and health. New York: Oxford University Press, 1995:247-71.

7 Marmot M G, Davey Smith G, Stansfeld S, Patel C, North F, Head J, et al. Health inequalities among British civil servants: the Whitehall II study. Lancet 1991;337:1387-93.

8 Brief A P, Burke M J, George J M, Robinson B S, Webster J. Should negative affectivity remain an unmeasured variable in the study of job stress? J Appl Psychol 1988;73:193-8.

9 Chen P Y, Spector P E. Negative affectivity as the underlying cause of correlations between stressors and strains. J Appl Psychol 1991;76:398-407.

10 McCrea G. Controlling neuroticism in the measurement of stress. Stress Medicine 1990;6:237-41.

11 Spector P E. A consideration of the validity and meaning of self-report measures of job conditions. International Review of Industrial and Organizational Psychology 1992;7:123-51.

12 Johnson JV, Stewart WF. Measuring work organization exposure over the life course with a job-exposure matrix. Scand J Work Environ Health 1993;19:21-8.

13 House J S, Strecher V, Metzner H L, Robbins C A. Occupational stress and health among men and women in the Tecumseh community health study. Journal of Health and Social Behavior 1986;27:62-77.

14 Kristensen T S. The demand-control-support model: methodological challenges for future research. Stress Medicine 1995;11:17-26.

15 Reid D D, Brett G Z, Hamilton P J S, Jarret R J, Keen H, Rose G. Cardiorespiratory disease and diabetes among middle-aged male civil servants. Lancet 1974;i:469-73.

16 Beksinska M, Yea L, Brunner E. Whitehall II study. Manual for screening examination 1991-93. London: University College London, Department of Epidemiology and Public Health, 1995.

17 Rose G, Reid D D, Hamilton P S, McCartney P, Jarrett R J. Myocardial ischaemia risk factors and death from coronary heart disease. Lancet 1977;i:105-9.

18 Karasek R A, Baker D, Marxer F, Ahlbom A, Theorell T. Job decision latitude, job demands and cardiovascular disease: a prospective study of Swedish men. Am J Public Health 1981;71:694-705.

19 Quinn RP, Staines GL. The 1977 quality of employment survey. Ann Arbor, MI: Survey Research Center, Institute for Social Research, University of Michigan, 1979.

20 Bradburn N M. The structure of psychological wellbeing. Chicago: Aldine, 1969.

21 Stansfeld S, North F M, White I, Marmot M G. Work characteristics and psychiatric disorder in civil servants in London. J Epidemiol Community Health 1995;49:48-53.

22 De Jonge J. Job autonomy, well-being, and health [thesis]. Maastricht: University of Maastricht, 1996.

23 Sauter S L, Hurrell J J, Cooper C L, eds. Job control and worker health. Chichester: Wiley, 1989.

24 Steptoe A, Appels A, eds. Stress, personal control and health. Chichester: Wiley, 1989.

25 Wenger N K. Gender, coronary artery disease, and coronary bypass surgery. Ann Intern Med 1990;112:557-8.

26 Wilcosky T, Harris R, Weissfeld L. The prevalence and correlates of Rose questionnaire angina among women and men in the lipid research clinics program prevalence study population. Am J Epidemiol 1987;125:400-9.

27 Rose G. Variability of angina. Some implications for epidemiology. Br J Prev Soc Med 1968;22:12-5.

28 Brunner E, Davey Smith G, Marmot M, Canner R, Beksinska M, O'Brien J. Childhood social circumstances and psychosocial and behavioural factors as determinants of plasma fibrinogen. Lancet 1996;347:1008-13.

29 Frese M. Stress at work and psychosomatic complaints. A causal interpretation. J Appl Psychol 1985;70:314-28.

30 Frese M, Zapf D. Methodological issues in the study of work stress: objective vs subjective measurement of work stress and the question of longitudinal studies. In: Cooper C L, Payne R, eds. Causes, coping and consequences of stress at work. Chichester: Wiley, 1990:375-411.

31 Kasl S V. The challenge of studying the disease effects of stressful work conditions. Am J Public Health 1981;71:682-4.

32 Kasl S V. Methodologies in stress and health: past difficulties, present dilemma's, future directions. In: Kasl S V, Cooper C L, eds. Stress and health: issues in research methodology. Chichester: Wiley, 1987:307-18.

33 Siegrist J, Peter R, Junge A, Cremer P, Seidel D. Low status control, high effort at work and ischemic heart disease: prospective evidence from blue-collar men. Soc Sci Med 1990;31:1127-34.

34 Ferrie J E, Shipley M J, Marmot M G, Stansfeld S, Davey Smith G. Health effects of anticipation of job change and non-employment: longitudinal data from the Whitehall II study. BMJ 1995;311:1264-9.

(Accepted 24 January 1997)

International Centre for Health and Society,
Department of Epidemiology and Public Health,
University College London Medical School,
London WC1E 6BT
Hans Bosma, senior research fellow
Michael G Marmot, director, Whitehall II study
Harry Hemingway, clinical lecturer in epidemiology
Amanda C Nicholson, clinical lecturer in epidemiology and public health
Eric Brunner, senior research fellow
Stephen A Stansfeld, codirector, Whitehall II study

Correspondence to: Dr H Bosma,
Faculty of Medicine and Health Sciences,
Erasmus University,
Box 1738,
3000DR Rotterdam,
Netherlands.