This Article Abstract Full Text (PDF) An erratum has been published Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Kivimäki, M. Articles by Shipley, M. J. Search for Related Content PubMed PubMed Citation Articles by Kivimäki, M. Articles by Shipley, M. J. Related Collections Stress and Coping Coronary Artery Disease

Why Is Evidence on Job Strain and Coronary Heart Disease Mixed? An Illustration of Measurement Challenges in the Whitehall II Study

From the University of Helsinki, Helsinki, Finland (M.K.); the University College London Medical School, London, UK (J.H., M.J.S., J.E.F., E.B., M.G.M.); the Finnish Institute of Occupational Health, Helsinki, Finland (M.K., J.V.).

Address correspondence and reprint requests to Mika Kivimäki, Finnish Institute of Occupational Health, Topeliuksenkatu 41 aA, FIN-00250 Helsinki, Finland. E-mail: mika.kivimaki{at}ttl.fi

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 
Objective: Evidence regarding the status of job strain as a risk factor for coronary heart disease (CHD) is mixed, including both results supporting the risk status and null findings. However, previous studies have typically assessed job strain at one point in time only. We examined whether the failure of such measurement to reflect long-term job strain could contribute to false null findings.

Methods: Job strain and its components, as stress indicators, were assessed twice (3-year time lag) for 5043 men and 2210 women who were free of apparent CHD at baseline. Incident CHD after the stress measurement comprised CHD death, a first nonfatal myocardial infarction, or definite angina (mean follow-up, 10.4 years). The data analysis was based on Cox proportional-hazard models adjusted for age, sex, and employment grade and corrected using regression dilution ratios calculated from short-term repeat data in a random subsample.

Results: In the total cohort, incidence of new CHD was higher for higher levels of job strain and demands. For these stress indicators, the corrected excess CHD risk was 30% and 29% higher than the corresponding uncorrected estimates, whereas the corresponding increase for job control was only 13%. Effects of job strain and work demands, but not job control, were stronger for a subgroup, with consistent exposure measurements over time than for the total cohort.

Conclusion: This evidence suggests that use of single-time exposure measures may underestimate the status of long-term job strain as a CHD risk factor.

Key Words: psychosocial factors • stress • coronary heart disease • measurement • epidemiologic studies

Abbreviations: CHD = coronary heart disease; ECG = electrocardiogram.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 
The job strain model is a leading occupational stress model (1,2). It posits that long-term job strain (i.e., a combination of high work demands and low job control at work) is a health risk for employees. Controversy remains regarding the status of job strain as a risk factor for coronary heart disease (CHD) (3,4). Although most published studies support the risk status (5–10), notable exceptions with nonsignificant findings also exist (11,12).

Several issues could lead associations being over- or underestimated in observational studies. Previous studies have typically assessed job strain and its components at one point in time only (4). Thus, a potential source of false null findings may be the lack of accurate indication of a long-term stress exposure.

The present report, from the (Whitehall II study) of British civil servants, used repeated job strain measurements to examine whether measurement problems are likely to contribute to mixed results for job strain and CHD.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 
Participants and Study Design
In the Whitehall II study, the target population was all-London-based office staff, aged 35 to 55 years, and working in 20 civil service departments. Baseline screening (phase 1), including the measurement of job strain and its components, took place between 1985 and 1988; 6895 men and 3413 women participated (response rate, 73%). In 1989 to 1990, 5533 men and 2600 women participated in phase 2 examination (79% of the baseline cohort). Incident CHD, the outcome, was assessed from phase 2 to the end of 1999. We focused on the 5043 men and 2210 women who had data on job strain at both phases and who also were free from apparent CHD at both phases.

Job Strain
Job strain and its components were measured using self-assessment scales of work demands (4 items, Cronbach’s 0.67) and lack of control (15 items, 0.84) in phase 1 and phase 2 (9,13). Scores for both scales were calculated as the sum of item scores, and this sum was expressed as a percentage of the theoretical maximum (range for both scales from 0–100). The job strain score was derived from the equation "work demand score – job control score" (13,14) (range from –87 to 83).

CHD
Incident CHD after phase 2 and before the end of year 1999 (mean follow-up, 10.8 years) comprised CHD death, a first nonfatal myocardial infarction, or definite angina. All of the components of the outcome were confirmed from clinical records. For the assessment of fatal CHD, the participants were flagged at the National Health Service Central Registry, which provided information on the date and cause of death (of the 10,308 male and female employees in the Whitehall II study, 10,300 were successfully flagged). Fatal CHD was defined according to the International Classification of Diseases, Ninth Revision (codes 410 through 414 as underlying causes of death). Potential new cases of nonfatal myocardial infarction were ascertained by questionnaire items on chest pain (15) and the physician’s diagnosis of a heart attack. The confirmation of myocardial infarction according to Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) criteria (16) was based on electrocardiographic findings, markers of myocardial necrosis, and a history of chest pain in the medical records. The assessment of angina was based on the participant’s reports of symptoms, with corroboration in medical records, or abnormalities in a resting electrocardiogram (ECG), an exercise ECG, or a coronary angiogram.

Data Analysis
In the analyses, job strain, work demands, and lack of control in phases 1 and 2 (called stress indicators) were used as standardized z-scores. In addition to single-time measures, we constructed a mean score, a maximum score, and a change score for each stress indicator to examine whether these scores, based on two measurement points, predict CHD more robustly than single-time measures. We fitted Cox proportional-hazard models adjusted for age, sex, and employment grade to study the associations between the stress indicators and the incidence of CHD (employees were censored at the time of the first CHD event or the time of loss to follow-up or at the end of 1999). In relation to stress change scores, we additionally adjusted for phase 1 score to take into account ceiling effect (i.e., upward change is less likely for higher than lower scores) and floor effect (downward change less likely for lower scores). The time-dependent interaction terms between each predictor and logarithm of follow-up period were all nonsignificant, confirming that the proportional hazards assumption was justified (all p values between 0.34 and 0.92). Among the 45 interaction tests of stress indicators with sex, employment grade, and age, there were two statistically significant interactions (the phase 2 score and mean score of job control with age). This is what would have been expected by chance. Thus, we combined men, women, all grades, and age groups and adjusted these in analyses.

The associations between stress indicators and CHD were determined for all of the participants and in an exploratory test in a subpopulation with consistent measurements of stress over time, i.e., those in the third with the least change in stress indicators between phases 1 and 2 (N = 2411–2492, depending on the indicator). As consistent measurements better represent long-term exposure than changing stress levels, we hypothesized that more robust associations between stress and CHD would be seen in this subgroup than in the total population.

In a further attempt to correct for the measurement imprecision in a single measure, we calculated a hazard ratio corrected for regression dilution bias (17). A single measure is subject to measurement error from usual stress levels and is expected to result in an underestimation of the association between a measured stress indicator and CHD. This is because the lowest category of stress presumably contains proportionally more people with a lower-than-usual stress score, and the highest category contains proportionally more people with a higher-than-usual stress score. For this purpose, new data from a random subsample (N = 267) of the total cohort were collected by measuring stress indicators at phase 3 (1991–1993) and repeating the measurements within a less than 2-month interval. These repeated stress scores were used to correct for the underestimation due to measurement imprecision by calculating the regression dilution ratio (ß_REG) according to Rosner’s regression method by regressing the second measurement on the first (17):

where w_i1 and w_i2 are the first and second measurements on the i^th subject and w and w the means of the first and second measurements. The corrected hazard ratio is calculated by dividing the observed log hazard ratio by the regression dilution ratio and taking the exponential of the result.

Using this method, the regression dilution ratio was 0.778 for job strain, 0.800 for work demands, and 0.902 for lack of control. For each phase 1 stress indicator in the total cohort, the corrected association with CHD was estimated by multiplying the uncorrected regression coefficient of the phase 1 score by the inverse of the regression dilution ratio. All analyses were performed with the use of SAS software, version 8.2 (SAS Institute).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 
Table 1 shows the characteristics of the participants. Their mean age was 44 years, 70% were men, and the largest occupational group was executives. The correlation coefficients for the repeated measurements of stress indicators between phases 1 and 2 were moderate: Pearson r = 0.45 for job strain, r = 0.51 for job demands and r = 0.72 for job control (all p < .0001).

Table 2 presents age-, sex-, and grade-adjusted associations between the stress scores and the incidence of new CHD. For job strain and lack of control, point-in-time measures, as well as mean and maximum scores across the two time points, were all predictive of CHD. For change scores and for demand scores, the hazard ratios were nonsignificant. The association between change scores and CHD remained broadly similar after further adjustment for phase 1 score (hazard ratios varied between 0.91 and 1.06).

Table 3 shows a subgroup analysis for participants with consistent stress measurement across phases 1 and 2. The three subgroups of those with consistent stress scores did not materially differ from the total population in terms of age (between 44.3 and 44.4 years), sex (71% to 75% male), and employment grade (34% to 38% administrative), suggesting that a major bias is unlikely (see Table 1 for figures regarding the total cohort). Hazard ratios for CHD for the point-in-time job strain and work demands scores were at least 91% higher among these employees with stable stress levels than those among all of the participants. Surprisingly, in view of the relatively high repeatability of the job control measure, the hazard ratios for lack of control were slightly weaker in the subgroup with consistent control scores, a finding that could have been due to chance.

Table 4 shows the associations between stress indicators and incident CHD after correction by the regression dilution method in the entire cohort. For job strain and work demands, the excess risk increased by 30% and 29%, but for lack of control, this increase was smaller, 13%.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 
Evidence from British civil servants suggests a moderate temporal stability for job strain, work demands, and job control during a mean period of 3 years. We found that high job strain, as assessed at a single time point, was associated with a higher risk of incident new CHD in the entire cohort and, in particular, among employees with consistent exposure measurements across two time points. As CHD develops over a long time span, long-term rather than short-term levels of job strain are assumed to affect CHD incidence (2,18). For employees with stable stress, a single-time measurement may provide an accurate estimate of long-term stress, but this is not necessarily the case for others.

We calculated a correction coefficient using Rosner’s regression dilution approach (17). For job strain and work demands, the corrected excess risk for CHD was approximately 30% higher than the uncorrected, the corresponding figure for lack of job control being 13%. The regression dilution approach uses a second measurement of exposure in order to correct for measurement error in the exposure. Although we applied a relatively short time interval for repeated stress measurements for the calculation of regression dilution ratios (<2 months), the underlying assumptions of long-term stress may not have been adequately met for all individuals if the inconsistency between measurements 3 months apart reflected a change in real conditions. This could lead to an under- or overcorrection; thus, the corrected values should be interpreted with caution.

We constructed several composite indicators to improve validity of stress exposure assessment in the total cohort. These included a mean score to reflect the average stress level over the 3-year assessment period, a maximum score to tap the possibility that elevated CHD risk after high stress remains irrespective of subsequent decrease in stress, and a baseline-adjusted change score to examine whether positive or negative change would be the key. However, none of these indicators substantially improved the prediction of CHD risk.

In conclusion, methodologic problems involving, e.g., the assembly of the sample, the validity of outcome measures, confounding (or residual confounding), and effect modification can lead to mixed findings and produce inaccurate results (4,14,17,19). Previous studies have usually assessed work stress at one point in time only. The present evidence from repeat measures, subpopulation analysis, and correction for regression dilution in a well-characterized cohort of British civil servants suggests that such an approach may be a potential source for underestimation of the effects of long-term work stress.

	NOTES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 

Received for publication August 23, 2005; revision received November 25, 2005.

DOI:10.1097/01.psy.0000221252.84351.e2

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION NOTES REFERENCES

 

1. Karasek RA. Job demands, job decision latitude and mental strain: implications for job redesign. Admin Sci Q 1979;24:285–307.[CrossRef]
2. Karasek R, Theorell T. Stress, Productivity and Reconstruction of Working Life. New York: Basic Books; 1990.
3. Kuper H, Marmot M, Hemingway H. Systematic review of prospective cohort studies of psychosocial factors in the etiology and prognosis of coronary heart disease. Seminars Vasc Med 2002;2:267–314.
4. Belcik KL, Landsbergis PA, Schnall PL, Baker D. Is job strain a major source of cardiovascular disease risk? Scand J Work Environ Health 2004;30:85–128.[Medline]
5. Karasek R, 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.[Abstract/Free Full Text]
6. Siegrist J, Peter R. Threat to occupational status control and cardiovascular risk. Isr J Med Sci 1996;32:179–84.[Medline]
7. Steptoe A, Fieldman G, Evans O, Perry L. Control over work pace, job strain and cardiovascular responses in middle-aged men. J Hypertens 1993;11:751–9.[Medline]
8. Kivimäki M, Leino-Arjas P, Luukkonen R, Riihimäki H, Vahtera J, Kirjonen J. Work stress and risk of cardiovascular mortality: prospective cohort study of industrial employees. BMJ 2002;325:857–60.[Abstract/Free Full Text]
9. Kuper H, Marmot MG. Job strain, job demands, decision latitude, and risk of coronary heart disease within the Whitehall II study. J Epidemiol Community Health 2002;57:147–53.
10. Hintsanen M, Kivimäki M, Elovainio M, Pulkki-Råback L, Juonala M, Raitakari OT, Keltikangas-Järvinen L. Job strain and early atherosclerosis: the Cardiovascular Risk in Young Finns Study. Psychosom Med 2005;67:740–7.[Abstract/Free Full Text]
11. Rosvall M, Ostergren PO, Hedblad B, Isacsson SO, Janzon L, Berglund G. Work-related psychosocial factors and carotid atherosclerosis. Int J Epidemiol 2002;31:1169–78.[Abstract/Free Full Text]
12. Eaker ED, Sullivan LM, Kelly-Hayes M, D’Agostino RB Sr, Benjamin EJ. Does job strain increase the risk for coronary heart disease or death in men and women? the Framingham Offspring Study. Am J Epidemiol 2004;159:950–8.[Abstract/Free Full Text]
13. Kivimäki M, Ferrie JE, Brunner E, Head J, Shipley MJ, Vahtera J, Marmot MG. Justice at work and reduced risk of coronary heart disease among employees: the Whitehall II Study. Arch Intern Med 2005;165:2245–51.[Abstract/Free Full Text]
14. Landsbergis PA, Schnall PL, Warren K, Pickering TG, Schwartz JE. Association between ambulatory blood pressure and alternative formulations of job strain. Scand J Work Environ Health 1994;20:349–63.[Medline]
15. Rose GA, Blackburn H, Gillum RF, Prineas RJ. Cardiovascular Survey Methods. 2nd ed. Geneva, Switzerland: World Health Organization; 1982.
16. Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A. Myocardial infarction and coronary deaths in the World Health Organization MONICA project: registration procedures, event rates, and case-fatality rates in 38 populations in four continents. Circulation 1994;90:583–612.[Abstract/Free Full Text]
17. Frost C, Thomson SG. Correcting for regression dilution bias: comparison of methods for a single predictor variable. J R Stat Soc A 2000;163:173–89.[CrossRef]
18. McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med 1998;338:171–9.[Free Full Text]
19. Macleod J, Davey Smith G, Heslop P, Metcalfe C, Carroll D, Hart C. Are the effects of psychosocial exposures attributable to confounding evidence from a prospective observational study on psychological stress and mortality. J Epidemiol Community Health 2001;55:878–84.[Abstract/Free Full Text]

This article has been cited by other articles:

				L. A. MacDonald, A. Cohen, S. Baron, and C. M. Burchfiel MacDonald et al. Respond to "Search for Preventable Causes of Cardiovascular Disease" Am. J. Epidemiol., June 15, 2009; 169(12): 1426 - 1427. [Full Text] [PDF]

				M. Elovainio, J. E. Ferrie, D. Gimeno, R. De Vogli, M. Shipley, E. J. Brunner, M. Kumari, J. Vahtera, M. G. Marmot, and M. Kivimaki Organizational Justice and Sleeping Problems: The Whitehall II Study Psychosom Med, April 1, 2009; 71(3): 334 - 340. [Abstract] [Full Text] [PDF]

				A. Vaananen, A. Koskinen, M. Joensuu, M. Kivimaki, J. Vahtera, A. Kouvonen, and P. Jappinen Lack of Predictability at Work and Risk of Acute Myocardial Infarction: An 18-Year Prospective Study of Industrial Employees Am J Public Health, December 1, 2008; 98(12): 2264 - 2271. [Abstract] [Full Text] [PDF]

				M Kivimaki, T Theorell, H Westerlund, J Vahtera, and L Alfredsson Job strain and ischaemic disease: does the inclusion of older employees in the cohort dilute the association? The WOLF Stockholm Study J Epidemiol Community Health, April 1, 2008; 62(4): 372 - 374. [Abstract] [Full Text] [PDF]

				A. Kouvonen, M. Kivimaki, M. Elovainio, A. Vaananen, R. De Vogli, T. Heponiemi, A. Linna, J. Pentti, and J. Vahtera Low organisational justice and heavy drinking: a prospective cohort study Occup. Environ. Med., January 1, 2008; 65(1): 44 - 50. [Abstract] [Full Text] [PDF]

				M. Kivimaki, J. E. Ferrie, M. Shipley, D. Gimeno, M. Elovainio, R. de Vogli, J. Vahtera, M. G. Marmot, and J. Head Effects on Blood Pressure Do Not Explain the Association Between Organizational Justice and Coronary Heart Disease in the Whitehall II Study Psychosom Med, January 1, 2008; 70(1): 1 - 6. [Abstract] [Full Text] [PDF]

				C. Aboa-Eboule, C. Brisson, E. Maunsell, B. Masse, R. Bourbonnais, M. Vezina, A. Milot, P. Theroux, and G. R. Dagenais Job Strain and Risk of Acute Recurrent Coronary Heart Disease Events JAMA, October 10, 2007; 298(14): 1652 - 1660. [Abstract] [Full Text] [PDF]

				M. Kivimaki, J. Vahtera, M. Elovainio, M. Virtanen, and J. Siegrist Effort-reward imbalance, procedural injustice and relational injustice as psychosocial predictors of health: complementary or redundant models? Occup. Environ. Med., October 1, 2007; 64(10): 659 - 665. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) An erratum has been published Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Kivimäki, M. Articles by Shipley, M. J. Search for Related Content PubMed PubMed Citation Articles by Kivimäki, M. Articles by Shipley, M. J. Related Collections Stress and Coping Coronary Artery Disease