Is job strain a major source of cardiovascular disease risk?

Karen Belkic; Paul A Landsbergis; Peter L Schnall; Dean Baker

doi:10.5271/sjweh.769

Belkic, Karen;Landsbergis, Paul A;Schnall, Peter L;Baker, Dean;

2004-04-01 00:00:00

Downloaded from www.sjweh.fi on October 26, 2021 Review Scand J Work Environ Health 2004;30(2):85-128 doi:10.5271/sjweh.769 by Belkic K, Landsbergis PA, Schnall PL, Baker D Affiliation: Karolinska Institute, Department of Oncology-Medical Radiation Physics, PO Box 260, Stockholm, SE-17176 Sweden. [email protected] Refers to the following texts of the Journal: 1996;22(4):241-242 1999;25(2):85-99 1997;23(4):257-265 1998;24(3):197-205 1996;22(2):139-145 1998;24(1):54-61 1998;24(5):334-343 The following articles refer to this text: 2004;30(2):81-83; 2006;32(6):431-442; 2006;32(6):473-481; 2006;32(6):515-527; 2007;33(1):1-3; 2008;34(1):48-54; 2008;34(1):40-47; 2008;34(4):288-296; 2009;35(4):284-293; 2011;37(6):455-463; 2012;38(6):489-502; 2013;39(1):106-111; 2014;40(5):441-456; 2015;41(3):280-287; SJWEH Supplements 2008;(6):33-40; SJWEH Supplements 2008;(6):41-51 Key terms: angina pectoris; cardiovascular death; cardiovascular disease; coronary artery disease; decision latitude; IHD; ischemic heart disease; job control; job strain; myocardial infarction; psychological demand; review; risk; skill discretion; work control This article in PubMed: www.ncbi.nlm.nih.gov/pubmed/15127782 This work is licensed under a Creative Commons Attribution 4.0 International License. Print ISSN: 0355-3140 Electronic ISSN: 1795-990X Review Scand J Work Environ Health 2004;30(2):85–128 2, 3 4 5 5 by Karen L Belkic, MD, Paul A Landsbergis, PhD, Peter L Schnall, MD, Dean Baker, MD Belkic K, Landsbergis PA, Schnall PL, Baker D. Is job strain a major source of cardiovascular disease risk? Scand J Work Environ Health 2004;30(2):85–128. Empirical studies on job strain and cardiovascular disease (CVD), their internal validity, and the likely direction of biases were examined. The 17 longitudinal studies had the highest validity ratings. In all but two, biases towards the null dominated. Eight, including several of the largest, showed significant positive results; three had positive, nonsignificant findings. Six of nine case–control studies had significant positive findings; recall bias leading to overestimation appears to be fairly minimal. Four of eight cross-sectional studies had significant positive results. Men showed strong, consistent evidence of an association between exposure to job strain and CVD. The data of the women were more sparse and less consistent, but, as for the men, most of the studies probably underestimated existing effects. Other elements of causal inference, particularly biological plausibility, corroborated that job strain is a major CVD risk factor. Additional intervention studies are needed to examine the impact of ameliorating job strain upon CVD-related outcomes. Key terms angina pectoris, cardiovascular death, coronary artery disease, decision latitude, ischemic heart disease, job control, myocardial infarction, psychological demands, skill discretion, work control. Clinicians are often called upon to assess the cardiovas- identified and mitigated, treatment of the manifestations cular work fitness of patients. As they attempt to make rather than the cause at best only ameliorates the condi- an informed judgment, a fundamental question arises: tion. At worst, the neglect of hazardous exposures may is the work environment fit, or conducive, to cardiovas- lead to both failure of treatment and failure to recognize a cular health? With technological advances, jobs char- public health problem with wide significance [p 19].” acterized purely by heavy physical demands have be- Several decades ago, occupational and environmen- come progressively less common. New types of work- tal health research raised the concern that exposure to related challenges and burdens primarily affecting the psychosocial stressors in the modern work environment higher nervous system of humans (ie, psychosocial stres- may be related to cardiovascular disease (CVD). It was sors) are more and more frequently encountered. Yet clear, however, that the evidence would be difficult to most of the clinical guidelines relevant to the interface obtain, that a myriad of thorny methodological problems between the workplace and the patient’s cardiovascular would arise, and that the critical obstacle would be the system continue to focus upon levels of physical exertion. theoretical conceptualization, modeling, and measure- Hu & Speizer (1) underscored the importance of ment of workplace stressors. A major breakthrough identifying job-related and other environmental hazards came in 1979 with the introduction of the job strain (de- that contribute to a given disease process. They noted mand-control) model (2). The model was developed for that “physicians commonly treat the sequelae of such work environments in which stressors are “chronic, not disease in the practice of medicine; however, unless the initially life-threatening and the product of sophisticat- underlying connection with hazardous exposures is ed human organizational decision making. In decision Portions of this paper were presented in abstracts to the 3rd International Congress on the Work Environment and Cardiovascular Disease, International Commission on Occupational Health, March 2002, Düsseldorf, Germany, and to the APA NIOSH Congress: Work, Stress and Health, March 2003, Toronto, Canada. Karolinska Institute, Stockholm, Sweden. Institute for Health Promotion and Disease Prevention Research, University of Southern California School of Medi- cine, Los Angeles, California, United States. Department of Community Medicine, Mount Sinai School of Medicine, New York, New York, United States. Division of Occupational & Environmental Medicine, Department of Medicine, University of California at Irvine, College of Medicine, Irvine, California, United States. Reprint requests to: Dr Karen Belkic, Adjunct Associate Professor of Preventive Medicine, Karolinska Institute, Department of Oncology-Medical Radiation Physics, PO Box 260, Stockholm, SE-17176 Sweden. [[email protected]] Scand J Work Environ Health 2004, vol 30, no 2 85 Job strain as source of cardiovascular disease risk with some studies incorporating the third dimension of so- making the controllability of the stressor is critical, and it cial isolation as well. Since the introduction of the model, becomes more important as increasingly complex and in- many empirical investigations have been published con- tegrated social organizations develop, with ever more cerning the relation between job strain and CVD out- complex limitations on individual behavior [p 78]”. The comes, including acute myocardial infarction (MI), coro- model has two components: “psychological demands, and nary artery disease (CAD), and CVD-related mortality. a combined measure of task control and skill use, or deci- Many of these studies report significant positive findings, sion latitude [p 78]”. Job strain occurs when the human and job strain is increasingly receiving attention as a po- organism is overloaded psychologically and at the same tential contributor to CVD (12–14). On the other hand, time deprived of control over the work environment, a there have been several nonconfirmatory findings con- combination which is predicted to give rise to increased cerning job strain and CVD outcomes published in large- risk of stress-related illness (3, p 78). The basic compo- scale studies. These results spurred some questions con- nents of the two dimensions are summarized in table 1. A cerning the strength and consistency of the evidence. third dimension, social isolation, was later added to the Several in-depth reviews (13, 15–19) have been pub- model, with the worst situation being “iso-strain”: high lished concerning the empirical data on the etiologic role demands, low decision-making latitude, plus lack of so- of psychosocial workplace factors (in some of these also cial support (5). outside work) and CVD. One of these reviews (16) fo- Exposure to job strain can be assessed from self-re- cused explicitly on job strain. However, a comprehensive port via a questionnaire, with the dimensions operational- and systematic assessment of the internal validity of the ized in the form of short, general instruments, most fre- current body of reported results is needed. A key issue quently the job content questionnaire (JCQ) or the psy- yet to be addressed is the direction in which methodolog- chosocial job strain questionnaire (PSJSQ) (6–8). These ical issues would most likely be acting, delineating situa- measures are feasibly administered in field and epidemi- tions that would increase the likelihood of obtaining null ologic studies. Data linkage methods have been developed results and those that could lead to an overestimation of in the United States and Sweden, so that exposure to job any association. strain (as well as to “iso-strain” in Sweden) can also be In this paper, we have used a predefined set of criteria inferred from occupational titles alone [ie, the imputation to examine each of the empirical investigations on job method (9–10)]. External assessment of job characteris- strain and CVD. The criteria were developed specifically tics (eg, expert observer) is yet another method for ob- to assess the methodological issues affecting the internal taining exposure data. [See the work of Greiner & Krause validity of studies on this topic, and, whenever possible, (11).] to identify the direction in which the results would most The job-strain model has been the model most widely likely be affected. We have not only addressed the strength used for evaluating the psychosocial work environment and consistency of the association, but also the other ma- and its potential impact upon the cardiovascular system, jor elements of causal inference as well. Particular atten- tion has been devoted to exploring the viability of alter- Table 1. Basic components of the job-strain model, derived from the work of Karasek & Theorell (4). native hypotheses, as well as the question of biological plausibility (ie, what are the possible mechanisms through Component Demand which job strain could affect the cardiovascular system). Psychological job demands Job requires very hard work This critical review has thereby sought to provide a more Job requires very fast work definitive answer to the question “Is job strain a major Job requires excessive work CVD risk factor?” The clinical implications of the con- Job involves conflicting demands clusion have also been explored. Job involves not having enough time to get the job done Decision latitude Skill discretion Job requires learning new things Methods used for the empirical review Job provides opportunities to develop one’s skills Identification and inclusion of relevant studies Job requires a high level of skill Job requires creativity Search strategy Job entails a variety of things to do Job does not involve a lot of repetitive work A computer-based search was carried out using Medline, Decision authority Job allows making one’s own decision via OVID, from 1966 to January 2002. The search terms Job provides a lot of freedom as to how the were entered as text words in the title, abstract, or other work gets done Job provides a lot of say on the job searchable fields (mesh terms, etc). For the independ- Job allows taking part in decisions affecting ent variable, the search terms were job strain, iso-strain, oneself decision latitude, psychological demands, work control, 86 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al job control, skill discretion, decision-making authority Included studies and intellectual discretion. These terms were combined A total of 35 articles was identified that met all the inclu- with search terms for the dependent variable: myocar- sion criteria and needed no further evaluation. Another dial infarction, angina pectoris, ischemic heart disease, five articles were independently reviewed by two of the coronary artery disease, and cardiovascular death. A authors, after which those by Hammar et al (21), Sihm et senior medical information specialist replicated this al (22), and Suadicani et al (23) were included, while those strategy. Bibliographies of relevant articles and person- by Lynch et al (24) and Murphy (25) were excluded since al files were also reviewed. they did not conform closely enough to the demands or the control dimensions of the job-strain model. Another three, by Billing et al (26), Karasek et al (27), and Messn- Inclusion criteria and procedure er & Sihm (28) were omitted since only main effects were Articles were chosen for review if they fulfilled the fol- assessed, and not job strain in relation to the dependent lowing criteria: (i) exposure to job strain was assessed or variable. In four instances two or more papers by the same imputed via its two major workplace dimensions, psycho- group of authors were combined. The full set of included logical demands plus any of the following: decision lati- articles is included in the bibliography. tude, skill discretion, decision authority and decision con- trol, as these directly relate to the dependent variable, (ii) Assessment of methodological quality any of the following were included as the dependent var- iable: CAD, manifestations of ischemic heart disease Internal validity assessment (IHD) (angina pectoris, MI) or mortality from cardiovas- We reviewed all studies fulfilling the aforementioned cri- cular causes, (iii) a case–control, cross-sectional or cohort teria with respect to their methodological strengths and design was used, (iv) the study was empirical, and (v) the weaknesses, focusing upon the aspects that would serious- complete study was published in English as a full-length ly compromise the internal validity of the reported results. article in a peer-reviewed journal. More than one publi- The 15 validity criteria grouped into four categories were cation by a given author or authors was included in the derived from the work of Stock (29). They are described review insofar as either the group(s) under study, the end- in the appendix. The first category, assembly of the sam- points, or the design differed. If two or more studies by ple, includes the avoidance of selection bias, the avoid- the same author(s) offered complementary information but ance of nonresponse bias, and the application of appro- had the same design, endpoint, and study group, they were priate exclusion criteria. The validity of exposure varia- combined and analyzed together. ble assessment was evaluated by five criteria related to Whenever self-report tools other than the JCQ or the assessment of point exposure to high psychological PSJSQ (7, 20) or their earlier versions were used to as- demands and to low control, the avoidance of recall bias, sess the demand and control dimensions, two of the au- the analysis of job strain, the adequate range of variation, thors independently reviewed the described methods to and the assessment of temporal aspects of exposure. The determine whether they were sufficiently compatible with four criteria under the category for confounding and ef- the job-strain model. This procedure was performed in a fect modification were adjustment for relevant demograph- blinded fashion. Formulations focusing primarily on the ic confounders, adjustment of relevant biomedical and be- individual’s subjective reaction to the work environment havioral confounders, appropriate consideration of gen- (eg, “how stressed are you by ...”) rather than on its ob- der as an effect modifier, and assessment of other dimen- jective characteristics were excluded. As a minimal guide- sions of the work environment. The validity criteria for line, it was required that at least one item from each of the outcome variable were related to the assessment of the major job-strain dimensions be included and that these the endpoint itself, whether the assessment of outcome was queries be phrased identically to the original question- blinded with respect to exposure status, and the adequate naires or so closely as to be a measure of the original con- range of variation of the outcome variable. cept (eg, “hectic work” as a measure of psychological job In most cases, the maximum score was 3 points (opti- demands). Insofar as the two reviewers disagreed, a third mal). For six of the criteria, there was a possibility for author served as an arbiter. Each of these studies (2 case–control, 1 cross-sectional) had some positive results, such that their omission does not represent a bias towards positive findings. However, because of a substantial number of methodological weaknesses, including confounding (26), low response rates (28), and problems with the assessment of the temporal aspects of exposure (26, 28) and with countermeasures against recall bias (26–28), among others, the positive results do not contribute major supporting evidence for the job strain hypothesis. The internal validity and directionality ratings were performed separately from our review of the results of the studies. Scand J Work Environ Health 2004, vol 30, no 2 87 Job strain as source of cardiovascular disease risk 4 points, insofar as the authors had used innovative ity of a study’s conclusions. Two of us independently methods that served to advance the state of the art in assessed each of the studies with respect to the 15 va- this field of investigation. Thus the maximum total score lidity criteria. In cases of disagreement, a third served possible was 51. The minimum score for each criterion as arbiter. was usually 1. There was the possibility of a score of 0 The methodological ratings for each study are present- for situations that would seriously undermine the valid- ed in tables 2–5 according to the 15 internal validity Table 2. Internal validity criteria for the assembly of the sample, rated according to the appendix. (CAD = coronary artery disease, CHD = coronary heart disease, CVD = cardiovascular disease, ECG = electrocardiography, IHD = ischemic heart disease, MI = myocardial infarction, MONICA = monitoring of trends and developments of cardiovascular disease, NHANES = National Health and Nutrition Exami- nation Survey) Study Internal validity criteria for assembly of the sample Avoidance of selection bias Avoidance of nonresponse bias Appropriate clinical exclusion criteria applied Score & comment Score & comment Score & comment Longitudinal studies Alfredsson et al, 3 All working men and women in 3 1 Previous MI occurrence, cerebrovascu- 1985 (30) Stockholm lar accidents not excluded at baseline Alterman et al, 2 67% participation, worker-based, 2 67% participation, nonrespondents 3 1994 (31) vital status follow-up 100% described Bosma et al, 1997 2 Worker based cohort: 79% participa- 2 73% response rate, broken down by 3 Excluded IHD by clinical exami- (32); Bosma et al, tion at phase 2; 83% in phase 3; employment grade nation with ECG, at baseline 1998 (33); Bosma subjects with low job control had low- et al, 1998 (34) er participation rates in phase 2 or 3 Hall et al, 3 Random population sample, 2 80% response rate, but no descrip- 1 CVD morbidity assessed at baseline by 1993 (35) complete follow-up tion of nonrespondents self-report, but not excluded or taken into account Hammar et al, 3 Population-based, studied incident 3 Data linkage, implied 100% 3 Linked hospital records to personal 1994 (21); cases with nested case-control design, identification number ensured that the Hammar et al, controls chosen at time of case inciden- incident MI was the first hospitalized 1998 (36) ce—incidence density sampling MI for patients and ruled out hospitalized MI for controls Hlatky et al, 0 Evidence for selection attrition (stop- 3 99% response rate in Mark et al, 3 Severe organic heart disease 1995 (37) ped working) among those exposed 1992 (49) excluded and CHD status at to job strain or low control and likely baseline taken into account selection bias in assembly of original sample Johnson et al, 3 Population-based cohort, complete 2 80% participation, nonrespondents 1 CVD morbidity assessed at baseline 1989 (38) follow-up not described by self-report, but not excluded or taken into account Johnson et al, 3 Population-based cohort with 2 80% participation, nonrespondents 1 CVD morbidity assessed at baseline 1996 (39) nested case-control design, not described by self-report, but not excluded or complete follow-up taken into account Karasek et al, 2 Population-based random sample, 2 92% initial response rate, non- 2 Adjusted for self-report of CVD 1981 (40) nested case-control design—but not respondents not described incidence-density-sampled controls Kivimäki et al, 3 Worker-based with 100% 1 Refusals replaced by others on 3 Clinical evaluation performed, CVD at 2002 (41) follow-up of vital status of cohort list, no description or figures given baseline excluded Kuper & Marmot, 3 Worker-based sample with 99.9% 2 73–77% nonrespondents broken 3 IHD excluded by clinical examination, 2003 (42) follow-up of vital status, 75.9% down by employment grade with ECG follow-up for morbidity in phase 5 Lee et al, 2002 2 Disease-free working survivors 2 78% responded, several character- 3 Those who reported CHD (43) 16 years after the initiation of a istics of nonresponders described excluded, diagnosis confirmed by worker-based (registered nurses) review of medical records cohort study, 95.5% follow-up Orth-Gomér et al, 2 Hospital-based selection, clear 2 43/335 (13%) nonrespondents, 200 par- 3 Endpoints were recurrent MI and 2000 (44) diagnostic inclusion criteria (acute ticipants were working at the time of the mortality, caseness based on MI or unstable angina pectoris), examination 3–6 months after the event; hospital and death registers <65 years of age; return to work nonrespondents included 13 who were in relation to job strain does not too sick and 21 who declined for other rea- appear to have been assessed sons, including inability to speak Swedish; no comparison of response rate between those with and without recurrent events Reed et al, 1989 3 Population-based with complete 2 9878/11148, 89% initial response 3 CHD excluded, implied by ex- (45) follow-up of cohort rate, nonrespondents not described amination Steenland et al, 3 Population based sample, 93% 2 NHANES response rate 70%, non- 2 Self-reported CVD 1997 (46) follow-up respondents described (continued) 88 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 2. Continued. Study Internal validity criteria for assembly of the sample Avoidance of selection bias Avoidance of nonresponse bias Appropriate clinical exclusion criteria applied Score & comment Score & comment Score & comment Suadicani et al, 2 Survivors from a 15-year worker- 1 75% response rate, nonrespondents 3 Self-report confirmed by 1993 (23) based cohort study not described hospital records Theorell et al, 2 Hospital based selection— 2 116/127 (91%) examined within 2 weeks; 3 Mortality study, excluded survi- 1991 (47) follow-up of survivors of definite excluded 6 not working and 31 immigrantes for vors with reinfarction or other MI occurring prior to age 45 years reasons of language competence; N=79 cardiac complications occurring followed-up during follow-up Case–control studies Alfredsson et al, 2 Population-based, case-control 3 3 MI excluded in controls during 1982 (50); Alfreds- study study period son & Theorell, 1983 (51) Bobák et al, 2 Population-based but survivor bias 2 179/191=94% eligible cases, 784/813 3 MONICA protocol 1998 (52) possible due to case-control design (96%) eligible controls, but initial participation rate of controls 75% Emdad et al, 2 CHD patients recruited from the 2 13/21 (62%) cases, nonrespondents described, 2 Noncases had 2-channel ECG 1997 (53) clinic, controls from working 87/130 (67%) noncase professional drivers were during protocol and no known population–survivor bias possible potential participants (based on matching), non- IHD by self-report participant noncases not described Hallqvist et al, 1998 2 Population-based, but survivor 2 Men 82% cases, 75% controls; women 3 All available medical records (54); Theorell et al, bias possible due to case-control 72% cases, 70% controls; some scrutinized, health examination 1998 (55); Reuter- design description of nonparticipation among of controls performed wall et al, 1999 (56); men and women Peter et al, 2002 (57) Netterstrøm et al, 2 Consecutive MI cases in two 2 100% for cases, 90% controls, non- 1 No explicit mention that MI 1999 (58) university hospitals, random respondents of the latter not described ruled out for controls population sample for controls Sihm, Dehlholm 2 Hospitalized MI survivors, hospi- 2 52/54 (96%) eligible cases, 3 Excluded controls with ECG signs et al,1991 (22) tal and population based controls 72/86 (84%) eligible controls, non- of MI, or history of angina pec- respondents not described toris or intermittent claudication Theorell et al, 2 Hospitalized MI survivors, 2 116/127 (91%) cases examined, 31 nonnative 3 History, maximal exercise 1987 (59) population-based controls Swedes excluded, 13 patients excluded because stress test not working; 116/125 (95%) controls agreed to participate, nonrespondents not described Wamala et al, 2 Cases admitted to cardiac clinic 3 292/335 (87%) cases included (of those not in- 2 No heart disease symptoms, no 2000 (60) for acute cardiac event, cluded 5 had died); 82.5% controls, controls hospitalization for previous population-based controls compared with random population sample, no 5 years among the controls differences in educational or life-style factors Yoshimasu & Fuku- 2 MI cases surviving to rehabilitation, 1 435/507 (86%) cases, 664/1325 (50%) controls, 2 Referents excluded if prior oka Heart Study admitted to 22 collaborating hospitals, psychosocial characteristics of nonparticipant history of MI, but unclear how Group, 2001 (61) population-based controls controls described in detail assessed Cross-sectional studies Hall et al, 1993 (35) 2 Population-based, cross-sectional 2 80% response rate, nonrespondents 3 Cross-sectional study of CVD not described Hlatky et al, 1995 0 Patients coming to tertiary clinical 3 99% response rate from Mark et al (49) 3 Excluded those requiring (37) center for angiography to work-up intensive cardiac care at the time chest pain, no diagnostic entity, 24% of angiography, pericardial or had normal coronary arteries, myocardial disease, baseline selection bias likely CHD status taken into account Johnson & Hall, 2 Population-based cross- 2 79% and 81%, effects of nonresponse on varia- 3 Cross-sectional study of CVD 1988 (5) sectional study bles concerning illness found to be minimal Johnson et al, 2 Population-based, cross- 2 80% response rate, nonrespondents not 3 Cross-sectional study of CVD 1989 (38) sectional study described Karasek et al, 2 Representative population 2 NHANES—response rate 70%, non- 3 1988 (63) sample respondents described Netterstrøm et al, 2 Population-based, cross- 1 63% response rate, nonrespondents not 3 1998 (64) sectional study described Sacker et al, 2 Population based, cross- 1 Response rate not reported 3 Cross-sectional study of CVD 2001 (65) sectional study Yoshimasu et al, 0 Patients undergoing angiography 2 733/838 (87.5%) said to have participated in the 3 Caseness defined by extent to 200 (66) for suspected or known IHD, study; however, a large number of exclusions coronary artery stenosis, 62% did not have significant were performed for various reasons, such that excluded valvular heart disease CAD, selection bias likely 197 men remained in the analysis—no descrip- tion of nonrespondents or of characteristics of the large number of those excluded Scand J Work Environ Health 2004, vol 30, no 2 89 Job strain as source of cardiovascular disease risk Table 3. Internal validity criteria for the assessment of the exposure variables rated according to the appendix. (CAD = coronary artery disease, CHD = coronary heart disease, HANES = Health and Nutrition Examination Survey, HES = Health Examination Survey, JCQ = job content questionnaire, PSJEM = pshychosocial job exposure matrix, PSJSQ = psychosocial job strain questionnaire, QES = quality of employment surveys) Study Internal validity criteria for assessment of the exposure variable Valid and reliable Avoidance of recall bias Analysis of point-expo- Adequate range of Valid and reliable assess- assessment of point for the exposure sure to job strain variation of the ment of temporal aspects exposure to psycho- variable exposure variable of exposure logical demands and control Score & comment Score & comment Score & comment Score & comment Score & comment Longitudinal studies Alfredsson et al, 1 Only 1 item (“hectic 3 Imputation study 2 Job strain treated 3 2 1 year follow-up 1985 (30) job”) for demands as a dichotomous dimension variable Alterman et al, 2 Imputed using QES 3 Imputed 3 Tertile term and 2 Mainly blue-collar 1 25-year follow-up, but 1994 (31) analysis of multi- workers, use of tertile stable occupation of plicative interaction term only yielded few cohort exposed to job strain Bosma et al, 1997 4 Self-report with White- 3 Independent observer, 3 Multiplicative inter- 2 All employment 3 Exposure assessed (32); Bosma et al, White-hall validation self-report in phase I, action term calcu- grades of white-collar twice at 3-year inter- 1998 (33); Bosma (4 items for demands) outcome in phase II lated, not predictive workers, few with job vals, follow-up of em- et al, 1998 (34) + independent or III of outcome, tertiles strain (14.7% males, ployment during study observer of control used to 17.2% females by assess dose- self-report, 11.9 & response 18.8% by external assessment) Hall et al, 1993 (35) 1 Imputation using 3 Imputed 2 Dichotomous variable 3 2 7–11 years of follow- PSJEM and 2 items up, exposure duration for demands assessed, but not temporal proximity, includes women aged 60–74 years at baseline Hammar et al, 1994 1 Imputation, demands 3 Imputation 3 All 4 quadrants 3 2 Occupation coded 1–9 (21); Hammar et al, assessed by two items assessed years before MI (1970– 1998 (36) 1975, incident cases 1976–1984), exposure assessed twice—occu- pationally stable cohort Hlatky et al, 3 JCQ with 5 items for 3 3 Quotient term and 3 2 No repeated exposure, 1995 (37) demands, but only 6 quadrant term 4-year average follow- of the 9 items for de- up, all employed at cision latitude baseline Johnson et al, 2 Validated question- 3 3 Quintiles 3 1 9-year follow-up, no 1989 (38) naire used, 2 items assessment of for demands cumulative exposure Johnson et al, 1 Imputed & 2 items 3 Imputation 3 3 cut points & some 3 2 Lifetime exposure 1996 (39) for demands description of multi- assessed prior to 14- plicative interaction year follow-up analysis Karasek et al, 3 Self-report—2 items 3 2 Dichotomous 3 1 No repeated exposure 1981 (40) for demands, validat- variable assessment, 9-year ed and expert ratings follow-up Kivimäki et al, 2 4 items for demands, 3 3 3 levels of exposure 3 Both white-collar and 1 Stratified analysis of 2002 (41) 12 for decision to job strain, de- blue-collar factory employees whose latitude, Cronbach mands and decision employees occupational group α=0.67 & 0.78, latitude remained unchanged 5 respectively; however, years after assessment some items inconsist- of exposure to work ent with dimension stressors but follow-up (eg, mental strain is of vital status >25 an element of job years control) Kuper & Marmot, 3 Self-report using 3 3 3 levels of exposure 2 White-collar workers 2 11.2-year follow-up, 2003 (42) Whitehall demand- to job strain, job of various grades relied upon baseline control questionnaire demands, and de- exposure data, but high cision latitude, also correlation between multiplicative inte- work characteristics in action term phases 1,2,3 & 5 (continued) 90 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 3. Continued. Study Internal validity criteria for assembly of the sample Valid and reliable Avoidance of recall bias Analysis of point-expo- Adequate range of Valid and reliable assess- assessment of point for the exposure sure to job strain variation of the ment of temporal aspects exposure to psycho- variable exposure variable of exposure logical demands and control Score & comment Score & comment Score & comment Score & comment Score & comment Lee et al, 2002 (43) 3 JCQ 3 3 4 quadrants for ex- 1 Narrow single occupa- 1 Job strain at baseline posure to job strain, tion—84% registered used to categorize 3 levels of exposure nurses, as well as exposure status, 4-year to demands and working former nurses, follow-up during control although in a variety of which 49% of settings (out-patient, those with job strain at operating room, admin- baseline changed istration, variance on exposure status relevant job characteris- tics not demonstrated) Orth-Gomér et al, 3 Swedish PSJSQ 5 3 3 Ratio calculated and 3 No apparent restric- 2 Occupationally stable, 2000 (44) items on demands quartiles used tion of occupation median follow-up 4.8 years Reed et al, 2 Imputation 3 4 Multiplicative inter- 3 1 18-year follow-up, but 1989 (45) action term + dose- number of years on job response assessed at baseline Steenland et al, 2 Imputed / QES 3 3 Quadrants 3 0 12–16 years of follow- 1997 (46) up, single assessment of employment status and job characteristics Suadicani et al, 1 Control = 1 item, no 3 3 Interactions assessed 3 2 3–4 years of follow-up, 1993 (23) mention of validation no assessment of repeated exposure Theorell et al, 2 2 questions for de- 3 3 Quotient term 3 1 All working at baseline, 1991 (47) mands, influence (3 follow-up time 6–8 items), intellectual dis- years, all who died cretion or variety (1 returned to same job item each) Swedish PSJSQ Case–control studies Alfredsson et al, 1 Imputed and 1 item 3 Imputation 3 Dichotomous ex- 3 1 Occupation coded 4–6 1982 (50); Alfreds- for demands posure, but multi- years before MI, no son & Theorell, plicative interaction in repeated exposure 1983 (51) 1983 paper assessment Bobák et al, 3 The Whitehall ques- 1 Cases interviewed 3 All 4 quadrants 3 2 Currently employed but 1998 (52) tionnaire, 3 items for 2 weeks post-MI assessed no repeated exposure demands, selected by assessment factor analysis Emdad et al, 3 Swedish PSJSQ 1 Case status known to 3 Quotient term 1 Single occupation multi- 2 Temporal proximity to 1997 (53) subject prior to variate comparisons be- employment among evaluation of work- tween professional dri- cases not described, place characteristics vers with CHD and con- number of years in trols-professional dri- occupation assessed vers with hypertension Hallqvist et al, 4 Imputation and self- 3 For men 4 Synergy index for 3 4 Cumulative exposure, 1998 (54); Theorell report via Swedish demands and decision all working mainly full- et al, 1998 (55); PSJSQ, detailed com- latitude among men time within last 5 years Reuterwall et al, parison between the 2 1999 (56); Peter et performed for men al, 2002 (57) 3 For women 1 For women 2 Dichotomous for women Netterstrøm 2 4 items for demands 1 Interview of cases in 3 4 quadrant 3 2 et al, 1999 (58) includes physical de- the coronary care unit, assessment mands & threat avoid- by nurses or physi- and vigilance α=0.51; cians who likely knew 6 items for decision the caseness, no evi- latitude α=0.65 & 0.81 dence of overreporting, but denial not ruled out Sihm, Dehlholm et al, 3 Orebro-validated ques- 1 Questionnaire ad- 2 A few dichotomous 3 2 Single assessment, 1991 (22) tionnaire, workload = ministered during 1st combinations excluded those on quantity of work & level week of hospitalization long-term disability or of strain (difficulty of sick leave or asked work tasks), also con- about current job tradictory demands; op- portunity for personal development & growth (3 items); autonomy (continued) Scand J Work Environ Health 2004, vol 30, no 2 91 Job strain as source of cardiovascular disease risk Table 3. Continued. Study Internal validity criteria for assembly of the sample Valid and reliable Avoidance of recall bias Analysis of point-expo- Adequate range of Valid and reliable assess- assessment of point for the exposure sure to job strain variation of the ment of temporal aspects exposure to psycho- variable exposure variable of exposure logical demands and control Score & comment Score & comment Score & comment Score & comment Score & comment Theorell et al, 2 Self-report, 2 items 1 Overreport ruled out, 3 Quotient terms 3 1 Cases had been 1987 (59) for demands, 1 ques- but not denial working at least part- tion for variety, 3 ques- time, but no explicit tions on influence over mention of controls, no work, 1 question on repeat exposure, not intellectual discretion clear when question- naire was administered Wamala et al, 3 Swedish PSJSQ as 1 3 Quotient term 3 2 Excluded those not 2000 (60) per Theorell et al (7) currently working, no repeat exposure assessment Yoshimasu & 2 Japanese version of 1 Self-report within 3 Used quadrant term, 3 2 Excluded those not Fukuoka Heart the JCQ, question- 1 month of acute assessed high, middle having a full-time job Study Group, 2001 naire-based interview, MI in cases and low strain; also from job strain (6!) validated, but only tested tertile term analysis, no repeated 2 items for demands exposure assessment Cross-sectional studies Hall et al, 1993 (35) 1 Imputed, 2 items for 3 2 Dichotomous variable 3 3 Measured lifetime demands exposure Hlatky et al, 3 JCQ with 5 items for 2 No relation between 3 Quotient term & 3 2 Currently employed, no 1995 (37) demands, but only 6 angina severity and quadrant term repeat exposure of 9 latitude items job strain, baseline cli- assessment nical status known to participant, but appar- ently not extent of CAD Johnson & Hall, 2 Self-report, 2 items for 1 4 Synergy index 3 2 No repeat exposure, 1988 (5) demands, reproduci- calculated currently employed bility 0.92, scalability 0.79; control 11 items Cronbach α=0.70 Johnson et al, 2 2 items for demands, 1 Exposure and out- 3 Dose-response: iso- 3 2 Employed at baseline, 1989 (38) validated questionnaire come by self-report strain no cumulative from same interview exposure Karasek et al, 2 Imputed QES 3 Imputational 3 Dichotomous, top 3 1 HES job exposure 1988 (63) 20%, also analyzed assessment 7–17 years as a continuous prior & HANES 6- variable years prior to assess- ment of outcome Netterstrøm et al 3 Whitehall methods, 2 Assessed association 3 4 quadrants 3 2 Currently occupation- 1998 (64) 5 demand items, between job strain ally active, no repeat 13 control and other pain, as measures well as angina pectoris and job satisfaction— no association Sacker et al, 3 Mainly Whitehall JCQ 1 Self-report of ex- 3 4 quadrants 3 2 Currently working full- 2001 (65) items: 6 for job control, posure and outcome time, single assess- 3 for job demands from same interview ment of exposure Yoshimasu et al, 2 2 items for demands, 2 Excluded those with 2 Median cut-points— 2 Blue- and white- 2 Currently working full- 2000 (66) (Cronbach α=0.61)], previous MI or long- 10% job strain collar, no restrictions time, no repeated 3 items for control standing angina on occupation, but exposure assessment Cronbach α=0.54, pectoris, question- small percentage test-retest reliability naires distributed prior exposed to job strain 0.51 to angiography, follow-up blinded interview either before or after angiography, specific instructions to answer questions as prior to symptoms or findings of any abnormal results regarding CAD 92 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 4. Internal validity criteria for confounding and effect modification according to the appendix. (BMI = body mass index, CAD = coronary artery disease, CHD = coronary heart disease, HDL = high-density lipoprotein, HRT = hormone replacement therapy, LDL = low- density lipoprotein, SCRF = standard cardiac risk factors, SES = socioeconomic status) Study Internal validity criteria of confounding and effect modification Adjustment for relevant Adjustment for relevant Stratification by gender Assessment of other demographic confounders biomedical and behavioral dimensions of the work confounders environment Score & comment Score & comment Score & comment Score & comment Longitudinal studies Alfredsson et al, 3 Adjusted for nationality, 2 Adjusted for age, smoking 2 Stratified analysis, but not 3 Explored interaction be- 1985 (30) income & residence type & some other biomedical adjusted for HRT or oral tween hectic work & sweaty SCRF contraceptives work and between hectic work & heavy lifting, also assessed irregular and long workhours, punctuality, gas and dust exposure, risk of ex- plosion, draft Alterman et al, 4 Job strain – SES 3 Strain-biomedical SCRF 3 2 Occupational class 1994 (31) assessed interaction done—but no results, also no strain- behavioral interaction Bosma et al, 1997 (32); 3 Assessed interaction of 4 Detailed assessment of be- 2 Gender-stratified, no 3 Effort-reward imbalance, Bosma et al, 1998 (33); SES and dimensions of havioral factors including mention of oral contracep- social support, employment Bosma et al, 1998 (34) job strain, London—no interaction effects; adjust- tive, HRT, menopause grade assessment of immigrant ment for smoking, choleste- status or ethnicity rol, high blood pressure, BMI Hall et al, 1993 (35) 4 Interaction between SES 2 Age-adjusted 2 Only women, no adjust- 2 Occupational class & job characteristics ment for oral contracep-, tives, HRT, menopause, LDL, fibrinogen Hammar et al, 3 Adjustment for SES 2 Age-adjusted 2 No HRT, oral contracep- 3 Interaction assessment for 1994 (21); Hammar tives, menopause, LDL, social support, long work- et al, 1998 (36) fibrinogen hours, noise all by imputa- tion, occupational class Hlatky et al, 1995 (37) 1 No adjustment for SES 2 Assume as for cross- 1 No gender-stratified anal- 3 Workhours, physical which differed significantly sectional, age, smoking, ysis, only adjustment, no demands, occupational status by CAD severity, no race diabetes, hypertension, mention of HRT, oral con- or ethnicity adjustment hypercholesterolemia traceptives, menopause Johnson et al, 4 Assessed interaction 2 Age-adjusted 3 2 Social support and occupa- 1989 (38) between SES & iso-strain tional class Johnson et al, 3 Adjusted for education, 2 Age, smoking, exercise 3 3 hazards, physical demands, 1996 (39) class & nationality social support, occupational status Karasek et al, 1981 (40) 3 Stratified by education 2 Age, smoking 3 1 Kivimäki et al, 3 2 Physical activity, smoking, 2 Adjusted for gender, inter- 3 Full evaluation of effort– 2002 (41) cholesterol, systolic blood action effects with work reward imbalance and pressure, BMI stressors assessed as not occupational group significant, no stratified analysis Kuper & Marmot, 3 Assessed interaction 2 Age, smoking, serum cho- 2 Nonsignificant interaction 2 Interaction with SES 2003 (42) between SES and job lesterol, hypertension, between job strain and strain, not race or exercise, BMI, alcohol gender, adjusted but not ethnicity in London stratified, no adjustment for HRT, menopause Lee et al, 2002 (43) 2 Education, husband’s 2 Smoking, BMI, hyper- 3 Women only, past use of 2 Nursing type and social education, no mention of tension, diabetes, hyper- oral contraceptives, support race or ethnicity in United cholesterolemia, dietary fat current use of HRT, States population intake, physical activity, menopausal status family history of MI Orth-Gomér et al, 2 Adjusted for education, 2 Age, standard biomedical 3 Women only, adjusted for 1 No other job stressors 2000 (44) not ethnicity in Stockholm factors, but not behavioral— estrogen status no multiplicative interaction between work & marital stress, no mention of home workhours or children Reed et al, 1989 (45) 4 Interaction between job 2 Several SCRF 3 1 strain & education, Japanese language ability Steenland et al, 2 SES but not race or 2 Several SCRF 3 2 Occupational status 1997 (46) ethnicity—United States study (continued) Scand J Work Environ Health 2004, vol 30, no 2 93 Job strain as source of cardiovascular disease risk Table 4. Continued. Study Internal validity riteria of confounding and effect modification Adjustment for relevant Adjustment for relevant Stratification by gender Assessment of other demographic confounders biomedical and behavioral dimensions of the work confounders environment Score & comment Score & comment Score & comment Score & comment Suadicani et al, 3 3 Complete assessment of 3 2 Social support, occupational 1993 (23) SCRF + relaxation as a status behavioral variable Theorell et al, 3 Adjusted for education, 3 Smoking, cholesterol, fam- 3 1 No other job stressors 1991 (47) immigrants excluded ily history, type A behavior,& mentioned number of stenosed arteries Case–control studies Alfredsson et al, 1982 3 2 Age 3 3 Shift work, lifting, piece rate, (50); Alfredsson & noise, vibration, accident risk, Theorell, 1983 (51) overtime work Bobák et al, 1998 (52) 3 Adjustment of SES using 2 Age, hypertension, other 3 1 various models, but inter- SCRF action effects not assessed Emdad et al, 1997 (53) 2 No adjustment for race or 2 Age-adjusted, detailed 3 3 Correlation analysis between ethnicity in Stockholm- assessment of SCRF and occupational stress index based study behavioral risk factors, but and dimensions of job strain not adjusted Hallqvist et al, 1998 3 Interaction between job 3 Age, smoking, hyper- 3 Men gender-stratified 3 Shiftwork, overtime, (54); Theorell et al, strain & social class tension, lipids, over- analyses supervising, effort–reward 1998 (55); Reuterwall among men, no commintment as a imbalance among men et al, 1999 (56); Peter adjustment for race or behavioral factor et al, 2002 (57) ethnicity in Stockholm 1 Women 2 Women, lipids assessed 2 Effort–reward imbalance but not included in job- among women strain risk estimate Netterstrøm et al, 2 Employment sector, not 2 Age, smoking 3 3 Workhours, moonlighting, 1999 (58) race or ethnicity, shiftwork, physical demands, Copenhagen social support, piece work Sihm, Dehlholm et al, 3 No significant difference 3 Age, assessed interactions 3 3 Job responsibility, job secur- 1991 (22) in social class, excluded between smoking, ity, job sociability, extra re- those with linguistic cholesterol & hypertension sources for help, 2 x 2 com- problems, Aarhus on one hand & workplace binations, but no assessment stressors on the other, of workhours, shiftwork, patients versus controls physical exposures Theorell et al, 1987 (59) 3 Education, immigrants 3 Age-matched, adjusted for 3 1 excluded, Stockholm tobacco consumption & LDL/HDL; glucose tolerance, heredity, type-A behavior & weight-to-height ratio assessed not significant in multiple regression Wamala et al, 2000 (60) 3 Detailed exploration 3 Age-matched, adjusted for 3 Assessed HRT, adjusted 2 Occupational class of social class, no smoking, hypertension, for menopausal status adjustment for ethnicity, exercise, obesity, lipid status, Stockholm hopelessness, coping Yoshimasu & Fukuoka 1 Percentage blue-collar 4 Adjusted for age, hyper- 3 Examination of job strain 3 Shift work, social support, Heart Study Group, jobs lower in nonstrain, tension, diabetes, hyper- only among men job type 2001 (61) P=0.13, occupational lipidemia, angina pectoris, status not included in obesity, cigarette smoking, multivariate analysis alcohol, parental CHD; assessed interaction of job strain and type-A behavior Cross-sectional studies Hall et al, 1993 (35) 4 Assessed interaction be- 2 Age-adjusted 2 No HRT, oral contracep- 2 Occupational status tween SES & job charac- tives, menopause, LDL, teristics fibrinogen Hlatky et al, 1995 (37) 1 No adjustment for SES, 2 Age, smoking, diabetes, 1 Adjusted for gender but no 3 Workhours, physical this differed significantly hypertension, cholesterol gender stratification, women demands, occupational according to CAD, fewest & men significantly differed status white-collar workers on outcome, no mention of among those with HRT, oral contraceptives, significant CAD menopausal status Johnson & Hall, 4 Stratified analysis by so- 2 Age, smoking, exercise 2 Stratified analysis, no men- 3 Physical demands adjust- 1988 (5) cial class, adjustment for tion of HRT, oral contra- ment, social support immigrant status ceptives, menopause interaction assessed (continued) 94 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 4. Continued. Study Internal validity riteria of confounding and effect modification Adjustment for relevant Adjustment for relevant Stratification by gender Assessment of other demographic confounders biomedical and behavioral dimensions of the work confounders environment Score & comment Score & comment Score & comment Score & comment Johnson et al, 4 Interaction between SES 2 Age-adjusted 3 2 Occupational status 1989 (38) and iso-strain Karasek et al, 1988 3 Education and race 2 Age, smoking, systolic 3 2 Physical demands (63) blood pressure Netterstrøm et al, 2 SES adjusted, but not 2 Age, smoking, systolic 2 Stratified analysis was 3 Workhours, social status, 1998 (64) ethnicity or immigrant blood pressure, HDL-to- done for men, not possi- social support, job security status, Copenhagen total cholesterol ratio ble for women because of empty cell Sacker et al, 2001 (65) 2 SES assessed, but not 2 Extensive assessment of 3 2 Blue-collar versus white- ethnicity, race, or immi- standard cardiac risk collar grant status factors but not behavior unrelated to these Yoshimasu et al, 3 Adjusted for job type as 4 Type-A behavior inter- 3 2 Workhours, blue-collar work, 2000 (66) an indicator of SES action assessed, age, stan- social support dard biomedical risk- factor adjustment Table 5. Internal validity criteria for the outcome variable according to the appendix. (CAD = coronary artery disease, CHD = coronary heart disease, CPK = creatine phosphokinase, CVD = cardiovascular disease, ECG = electrocardiography, HANES = Health and Nutrition Examination Survey, HES = Health Examination Survey, IHD = ischemic heart disease, MI = myocardial infarction, WHO = Work Health Organization) Study Internal validity criteria for outcome variable Valid assessment of the outcome Assessment of outcome blinded with Adequate range of variation of the variable respect to exposure outcome variable Score & comment Score & comment Score & comment Longitudinal studies Alfredsson et al, 1985 (30) 2 Hospital registry 3 Linkage 2 Hospitalized cases of MI Alterman et al, 1994 (31) 2 Mortality from death certificates 3 Linkage, also explicit blinding 3 Whole cohort followed-up for —main result in Ostfeld vital status, all cases included Bosma et al, 1997 (32); 1 Self-report of IHD 2 Self-report of outcome, but indepen- 2 IHD survivors Bosma et al, 1998 (33); dent as well as self-report of exposure Bosma et al, 1998 (34) Hall et al, 1993 (35) 2 National Death Registry 3 Imputation 3 Whole cohort followed-up for vital status, all cases included Hammar et al, 1994 (21); 3 Registry data with previous 3 Imputation 3 Fatal and nonfatal MI Hammar et al, 1998 (36) validation study Hlatky et al, 1995 (37) 1 Unclear how follow-up was carried 2 Unclear whether self-report of out- 3 Presumably all participants out, states “all patients were con- come, not explicitly blinded assess- followed-up regardless of tacted” at follow-up intervals to ment outcome “document out-come” [p 328] Johnson et al, 1989 (38) 2 Registry data 3 3 Whole cohort followed-up for vital status, all cases included Johnson et al, 1996 (39) 2 Registry data 3 Linkage 3 Whole cohort followed-up for vital status, all cases included Karasek et al, 1981 (40) 3 Validated death certificate 3 3 All CVD deaths included during follow-up period Kivimäki et al, 2002 (41) 2 Registry data 3 Use of registry 3 CVD mortality, obtained cause of death for all participants who died during the follow-up period Kuper & Marmot, 2003 (42) 3 National registry data for mortality, 3 Independent review 3 Fatal and nonfatal incident CHD clinical records and ECG reviewed by two trained coders Lee et al, 2002 (43) 3 WHO criteria for MI, death certifi- 3 Explicitly blinded 3 Nonfatal MI and fatal CHD cates corroborated by autopsy or hospital records Orth-Gomér et al, 2000 (44) 3 Validated hospital and death 3 Based on registry data 2 Complete follow-up of patients registers hospitalized for cardiac events Reed et al, 1989 (45) 3 Panel of physicians reviewed the 3 3 Entire cohort followed-up medical data (continued) Scand J Work Environ Health 2004, vol 30, no 2 95 Job strain as source of cardiovascular disease risk Table 5. Continued. Study Internal validity criteria for outcome variable Valid assessment of the outcome Assessment of outcome blinded with Adequate range of variation of the variable respect to exposure outcome variable Score & comment Score & comment Score & comment Steenland et al, 1997 (46) 2 Hospital records and death 3 3 IHD deaths and hospital discharges certificates for heart disease Suadicani et al, 1993 (23) 3 Review of death and hospital regis- 3 Registry data 3 Complete follow-up of cohort try with validity frequently assessed Theorell et al, 1991 (47) 2 Cardiologist review of reinfarction 2 2 Excluded from analysis those who mortality survived a reinfarction Case–control studies Alfredsson et al, 1982 (50); 2 Hospital and death registry 3 3 All MI, fatal and nonfatal Alfredsson & Theorell, 1983 (51) Bobák et al, 1998 (52) 3 MONICA protocol 2 2 Survivors of MI Emdad et al, 1997 (53) 2 Hospitalized cases of IHD events 3 All data analysis performed in a 2 Survivors of IHD events blinded fashion Hallqvist et al, 1998 (54); 3 Explicit diagnostic criteria 3 Data linkage in men 3 All MI, fatal and nonfatal Theorell et al, 1998 (55); 2 Women Reuterwall et al, 1999 (56); Peter et al, 2002 (57) Netterstrøm et al, 1999 (58) 2 Severe chest discomfort or ECG 2 Implied 2 Hospitalized survivors of MI signs of MI accompanied by increa- ed CPK to twice the normal level Sihm, Dehlholm et al, 2 “Established diagnosis of MI” 2 2 Hospitalized survivors of MI 1991 (22) <55 years old Theorell et al, 1987 (59) 3 WHO criteria for definite MI, CAD 2 2 Hospitalized survivors of MI by coronary angiography <45 years old Wamala et al, 2000 (60) 3 Explicit diagnostic criteria, in- 3 2 Hospitalized survivors of cardiac cluding WHO criteria for MI events Yoshimasu & Fukuoka Heart 2 Collaborating cardiologists were 2 Implied but not explicit 2 Hospitalized survivors of acute MI Study Group, 2001 (61) responsible for the diagnosis of acute MI Cross-sectional studies Hall et al, 1993 (35) 1 Self-reported CVD 3 2 Survivors only assessed Hlatky et al, 1995 (37) 3 CAD assessed by coronary angio- 2 2 Excluded patients with unstable graphy, with clear diagnostic angina or other conditions requiring criteria intensive care at time of angiography Johnson & Hall, 1988 (5) 1 1 2 Survivors only assessed Johnson et al, 1989 (38) 1 Self-report of IHD, although inde- 1 Self-reported exposure and outcome 2 Only survivors assessed pendent diagnostic system, no objective evidence Karasek et al, 1988 (63) 2 HES review by four physicians, 3 Data linkage 2 Only survivors assessed specific ECG, history and blood chemistry for definite MI, reliability assessment made, HANES review of medical records, physical exami- nation, ECG not always available Netterstrøm et al, 1998 (64) 1 Self-report only via Rose 1 Self-report of exposure and of 2 Only survivors assessed questionnaire outcome Sacker et al, 2001 (65) 1 Self-report 1 Self-report of exposure and outcome 2 Survivors of heart disease Yoshimasu et al, 2000 (66) 3 Explicit diagnostic criteria for 3 1 Excluded those with long-standing stenosis angina pectoris or previous MI criteria given in the appendix, the criteria being reliable assessment of temporal aspects of exposure) grouped according to their categories, assembly of in table 3, confounding and effect modification (adjust- the sample (avoidance of selection bias, avoidance ment for relevant demographic confounders, adjust- of nonresponse bias, appropriate clinical exclusion ment for relevant biomedical and behavioral con- criteria applied) in table 2, assessment of the expo- founders, stratification by gender, assessment of oth- sure variable (valid and reliable assessment of point er dimensions of the work environment) in table 4, exposure to psychological demands and to control, and the outcome variable (valid assessment of the avoidance of recall bias for the exposure variable, outcome variable, assessment of outcome blinded analysis of point-exposure to job strain, adequate with respect to exposure, adequate range of varia- range of variation of the exposure variable, valid and tion of the outcome variable) in table 5. 96 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Directionality mation bias if the outcome and exposure were both self- reported. Next, a judgment was made about the overall direc- We next asked the question of how these various meth- tion in which the methodological issues were likely to odological issues would affect the results and in which affect the results concerning associations with job direction they would most likely be acting. We exam- strain, as follows: (i) unequivocal bias to the null: sev- ined each of the validity criteria from this perspective, eral clear and strong biases to the null and no biases to delineating situations that would increase the likelihood overestimate; (ii) likely bias to the null: a few likely bi- of obtaining null results and those that could lead to an ases to the null and no clear bias to overestimate, (iii) overestimation of association. These directionality is- minimal biases: nearly all of the potential sources of bias sues roughly followed the order of the internal validity fully taken into account; (iv) bias possible in both di- criteria, although there was no precise one-to-one cor- rections; (v) likely bias to overestimate: one or more respondence between them, since some issues may af- likely biases to overestimate and no clear bias to the fect the results in either direction, depending on the spe- null; and (vi) unequivocal bias to overestimate: several cific circumstances. There were also instances in which clear and strong biases to overestimate and no biases to the way a methodological issue might affect the results the null. could not be determined (eg, a low response rate with- out any description of the nonrespondents). The issues that were considered to increase the like- lihood of obtaining null results included (i) selection Strength and consistency of the empirical findings bias in the assembly of the sample, if the participants with respect to job strain and cardiovascular exposed to job strain but without CVD preferentially en- disease, reviewed in light of the methodological tered the study; (ii) selective attrition, if those exposed issues affecting the results to job strain or related work stressors selectively stopped working during the follow-up period; (iii) survivor bias The salient details with respect to the results of each of (healthy worker effect); (iv) nonexclusion of outcome the reviewed longitudinal, case–control, and cross-sec- at baseline leading to dilution of the results; (v) use of tional studies are presented in tables 6, 7, and 8, respec- the imputation method (imprecise) to define job strain, tively. Table 9 provides a summary of the relationships leading to nondifferential misclassification; (vi) one to between the results and the direction in which the meth- two items for assessing psychological demands if the im- odological issues were likely to affect each study. putation method was used, leading to nondifferential misclassification; (vii) use of a dichotomous variable to define job strain, leading to nondifferential misclassifi- Longitudinal studies cation; (viii) a low percentage of exposure to job strain, The longitudinal studies (21–23, 30–47) had higher leading to a loss of power to detect an existing effect; mean total validity ratings than the case–control and (ix) single occupation or a limited range of variation of cross-sectional studies did. The mean scores of the stud- exposure; (x) assessment of exposure to job strain tem- ies among men were almost identical for the positive, porally distant from the outcome (studies with long fol- nonsignificant positive, and null studies. The null stud- low-up periods without repeated assessment of exposure ies had a somewhat lower mean total score for the wom- status); (xi) lack of a gender-stratified analysis; (xii) like- en than those that were positive. Of the two longitudi- ly confounding by another factor, if the relationships were nal studies with the highest total scores (score 40), one in the opposite direction of the tested association or if sev- yielded a significant positive effect estimate (21, 36), eral important confounders were not taken into account. while null results were obtained in the other one (45). The issues that were considered to increase the like- Notwithstanding the high overall methodological qual- lihood of an overestimation of association were (i) se- ity of these investigations, in all but two, biases towards lection bias in the assembly of the sample, if the partic- the null dominated. In 11 of the 17 studies, the biases were ipants exposed to job strain and with CVD preferential- unequivocal. Biases towards the null were generally due ly entered the study; (ii) selective attrition, if those not to the use of the imputation method and long follow-up exposed to job strain or related work stressors selective- times, with no re-assessment of exposure or even employ- ly stopped working during the follow-up period; (iii) ment status. Persons close to or even above usual retire- information bias if the outcome was known to the par- ment age were included in the baseline sample of sever- ticipant at the time of the self-report of exposure; (iv) al of the studies with protracted follow-up (21, 35, 36, likely confounding by another factor, if the relationships 38–40, 44–46); this inclusion would have attenuated the were in the direction of association. An alternative hy- effect estimates even further. The imputation method is pothesis is likely to be operative, whereby a factor oth- particularly problematic for the psychological demand er than job strain is the true effect modifier: (v) infor- Scand J Work Environ Health 2004, vol 30, no 2 97 Job strain as source of cardiovascular disease risk Table 6. Results of the reviewed longitudinal studies. (BMI = body mass index, CAD = coronary artery disease, CHD = coronary heart disease, CVD = cardiovascular disease, HR = hazards ratio, HRT = hormone replacement therapy, IHD = ischemic heart disease, MI = myocardial infarction, NS = nonsignificant, O = observer-rated, OR = odds ratio, RR = relative risk, SBP = systolic blood pressure, SES = socioeconomic status, SMR = standardized mortality ratio, SR = self-rated, UK = United Kingdom, US = United States, 95% CI = 95% confidence interval) Study Participants Foll- Illness Significant positive Reported nonsignificant, null Methodological issues b d ow- outcome associations or significant negative and total validity scores up associations (years) Studies with significant positive results for job strain and CVD Alfredsson et al, N=958 096, 1 Hospitalized Men: punctuality (age) SMR 121 Men: hectic work (age) NS, 4 biases to null: non- 1985 (30) Swedish, MI (N=1059 (95% CI 110–133), few possibili- monotonous work (age) NS; exclusion of previous MI at 20–64 years men, N=142 ties to learn new things (age) SMR women: hectic work (age) NS baseline, imputation of age, popu- women) 113 (95% CI 104–123), hectic & method, 1 item for lation-based monotonous work (age) SMR 118 psychological demands, (95% CI 102–135), hectic work & single cut-point; total few possibilities to learn new validity criteria score 35 things (age + income) SMR≈125 (95% CI≈105–150); women: hectic & monotonous work (age) SMR 164 (95% CI 112–233), monotony (age) SMR 128 (95% CI 104–157), low influence on workmates (age) SMR 133 (95% CI 102–170), low influence on holidays (age) SMR 145 (95% CI 114–182) Hammar et al , N=24913 9 First MI Men (all): high strain work RR Men (blue-collar): hectic work RR 3 biases to null: imputation, 1994 (21); men, N=3535 (N=8833 1.21 (95% CI 1.08–1.35), low 1.2 (95% CI 1.0–1.4), few possibi- 2 items for psychological Hammar et al, women, men, decision latitude RR 1.19 (95% lities to learn new things RR 1.3 demand, long follow-up 1998 (36) population N=1175 CI 1.13–1.25); men (white-collar): (95% CI 0.9–1.9), hectic work & outcome (temporally controls, women) hectic work and low influence few possibilities to learn new things distant to exposure); total Swedish, over workhours RR 1.4 (95% CI RR 1.2 (95% CI 1.0–1.4); men validity criteria score 40 30–64 years 1.1–1.8); women (all): hectic (white-collar): hectic work RR 1.0 of age, nested work and few possibilities to learn (95% CI 0.8–1.3), few possibilities case-control new things RR 1.3 (95% CI 1.1– to learn new things RR 1.2 (95% CI study 1.6), hectic work & low influence 1.0–1.4), hectic work & few possi- on work planning RR 1.3 (95% CI bilities to learn new things RR 1.2 1.1–1.6), high-strain work RR (95% CI 1.0–1.6); women (blue- 1.23 (95% CI 1.01–1.51), low collar): hectic work RR 0.7 (95% CI decision latitude RR 1.44 (95% 0.5–1.1), few possibilities to learn CI 1.25–1.65); women (white- new things RR 2.1 (95% CI 0.9– collar): few possibilities to learn 4.9); women (white-collar): new things RR 2.3 (95% CI 1.2– hectic work RR 1.8 (95% CI 0.9– 4.6) (age, county, calendar year) 3.7) (age, county, calendar year) Johnson et al, N=7219 men, 9 CVD mortality Iso-strain (total group) RR 1.92 Iso-strain (white-collar) RR 1.31 2 biases to null: nonexclu- 1989 (38) Swedish, 25–65 (N=193) (95% CI 1.15–3.21), iso-strain (95% CI 0.58–2.96 ) (age) sion of CVD at baseline, long years of age, po- (blue-collar) RR 2.58(1.06–6.28) follow-up outcome (tempo- pulation based (age) rally distant to exposure); study total validity criteria score 37 Karasek et al, N=1461 men, 9 CVD & cere- High psychological demands Low intellectual discretion 3 biases to null: dichoto- 1981 (40) Swedish, 18– brovascular OR 4.0 (95% CI 1.2–13.9, high OR 1.5 (95% CI 0.4–5.1), low mous variable to assess job 60 years of mortality psychological demands & low personal schedule freedom strain, long follow-up out- age, popula- (N=22) personal schedule freedom OR 1.7 (95% CI 0.6–4.7) (same come (temporally distant to tion-based OR 4.0 (95% CI 1.1–14.4) (age, adjustment as for positive exposure), matching con- study (nested education, smoking, CHD findings) trols by CHD symptoms & case-control, symptoms matched at baseline) education attenuated asso- N=66 controls) ciations; total validity criteria score 36 Bosma et al, N=6895 men, 5.3 New self- Men: low control (SR) & angina Men: job strain (SR) & angina 1 bias to null: all white- 1997 (32); N=3413 wo- report: angina pectoris OR 1.54 (95% CI 1.05– pectoris OR 1.40 (95% CI 0.93– collar workers (few with job Bosma et al, men, UK, 35– (N=177 men, 2.26), low control (SR) & diag- 2.10), job strain (SR) & diagnos- strain), 1 possible bias to 1998 (33); 55 years of age, N=151 wo- nosed IHD OR 1.6 (95% CI 1.01– ed IHD OR 1.16 (95% CI 0.70– overestimate although Bosma et al, civil servants men), diagno- 2.55), low control (SR) & any 1.94), job strain (O) & all out- authors demonstrated that 1998 (34) sis IHD CHD event OR 1.55 (95% CI comes OR 1.03 (95% CI 0.66– this is unlikely: information (N=124 men, 1.20–2.01), low control (O) & 1.61); women: low control (SR) bias from self-report of N=42 wo- any CHD event OR 1.43 (95% CI & angina pectoris OR 1.20 (95% exposure and outcome; men), any 1.09–1.88), job strain (SR) & CI 0.74–1.92), low control (SR) total validity criteria score CHD event any CHD event OR 1.45 (95% & diagnosed IHD OR 0.85 (95% 39 (N=401 men, CI 1.03–2.06); women: low CI 0.38–1.87), low control (O) & N=253 wo- control (SR) & any CHD event angina pectoris OR 1.46 (95% CI men) OR 1.74 (95% CI 1.15–2.64), 0.87–2.43), low control (O) & low control (O) & any CHD event diagnosed IHD OR 1.48 (95% CI OR 1.73 (95% CI 1.14–2.62) 0.53–3.85), job strain (SR) & any (age and follow-up time) CHD event OR 1.14 (95% CI 0.76– 1.72), job strain (O) & any CHD event OR 1.22 (95% CI 0.80–1.86) (age and follow-up time) (continued) 98 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 6. Continued. Study Participants Foll- Illness Significant positive Reported nonsignificant, null Methodological issues b d ow- outcome associations or significant negative and total validity scores up associations (years) Kivimäki et al, N=545 men, Mean CVD mortality Intermediate job strain HR 1.64 Low job control HR 1.42 (95% 1 or possibly 2 biases to 2002 (41) N=267 wo- 25.6 (N=60 men, (95% CI 0.85–3.19), high job CI 0.72–2.82)(age, gender, null: long follow-up tempo- men, Finnish, N=13 wo- strain HR 2.22 (95% CI 1.04– occupational group, smoking, rally distant from exposure, metal factory men) 4.73) (age, gender, occupa- physical activity, SBP, choleste- no gender-stratified analy- employees tional group, smoking, physical rol, BMI) sis although no significant activity, SBP, cholesterol, interaction with work stres- BMI) sors; no information about nonresponders (unclear how this would affect re- sults); total validity criteria score 37 Kuper & Marmot, N=6895 men, Mean Incident-vali- High demand and low control & High demand and low control & 1 possible bias to null: no 2003 (42) N=3413 wo- 11 dated CHD all CHD HR 1.38 (95% CI 1.10– fatal CHD or nonfatal MI HR 1.16 gender-stratified analysis men, UK, 35– 1.75) (age, gender, occupa- (95% CI 0.78–1.71) (age, although no significant 55 years of tional grade, coronary risk gender, occupational grade, interaction with job strain; age, civil ser- factors) coronary risk factors) total validity criteria score vants 39 Theorell et al, N=79 men, 6–8 Mortality from Demands divided by variety Demands NS, single aspects of 1 bias to null: long follow- 1991 (47) Swedish, repeat MI (univariate) P=0.03, demands decision latitude NS (univariate) up outcome (temporally <45 years of (N=13) divided by intellectual discretion distant to exposure diluted age, employ- P=0.02(biomedical risk factors, associations); 1 possible ed, first MI education did not differ signifi- bias to overestimate: all survivors cantly between groups) who died had returned to same work as pre-MI (no mention of survivors, could not rule out that survivors, who as a group had lower job strain exposure selecti- vely, did not return to work); total validity criteria score 35 Studies with positive results for job strain and CVD, but none of which were statistically significant Alterman et al, N=1683 men, 25 CHD mortality High decision latitude RR 0.76 Job strain RR 1.40 (95% CI 3 biases to null: imputation 1994 (31) US, 38–56 (N=283) (95% CI 0.6–0.98) (age, SBP, 0.92–2.14), psychological method, low percentage years of age, cholesterol, smoking, alcohol, demands RR 0.78 (95% CI 0.48– (7.5%) exposed to job healthy Chica- family history of CVD) 1.26), decision latitude RR 0.76 strain, long follow-up (out- go Western (95 % CI 0.59–1.00) (education come temporally distant to Electric em- & age, SBP, cholesterol, smok- exposure); total validity ployees of Eu- ing, alcohol, family history of criteria score 38 ropean ances- CVD) try (74% blue- collar) Steenland et al, N=3575 men, 12– Incident heart Job control (highest compared Blue-collar: job strain OR 1.14 2 biases to null, the latter 1997 (46) US, 25–74 16 disease with lowest quartile) OR 0.71 (95% CI 0.8–1.63), psycholog- of which seriously threat- years of age, (N=519) (95% CI 0.54–0.93) (age, ical demands OR 0.64 (95% CI ened the internal validity of population- education, blood pressure, 0.4–1.03), control OR 0.69 the study: imputation, sin- based study other coronary risk factors) (95% CI 0.46–1.02), high gle assessment of job char- (58% blue- control & hight demand OR 0.69 acteristics temporally very collar) (95% CI 0.48–0.99); white-collar: distant to exposure would job strain OR 1.05 (95% CI 0.63– strongly dilute associa- 1.77), psychological demands tions; total validity criteria OR 0.93 (95% CI 0.61–1.44), score 35 control OR 0.74 (95% CI 0.43– 1.26)(as for positive findings) Orth-Gomer et al, N=292 wo- 3.2– Recurrent Job strain: second quartile 1 or possibly 2 biases to 2000 (44) men, Swed- 6.2 coronary HR 1.53 (95% CI 0.58–4.02), null: selective attrition of ish, 30–65 years, events (N=81) upper 2 quartiles HR 1.69 (95% return to work in relation to years old, me- CI 0.72–3.98) (age); job strain: job characteristics not ruled hospitalized dian second quartile HR 1.33 (95% out (possible bias), fairly for acute 4.8 CI 0.43–4.10), upper 2 quartiles long follow-up outcome MI or un- HR 1.67(95% CI 0.64–4.32) (temporally distant to expo- stable angina (age, estrogen status, education, sure); no adjustment for pectoris diagnosis at index event, symp- marital stress or assess- toms of heart failure, SBP, ment of interaction with job diabetes mellitis, smoking, characteristics (uncertain lipids) how this affected results); total validity criteria score (continued) Scand J Work Environ Health 2004, vol 30, no 2 99 Job strain as source of cardiovascular disease risk Table 6. Continued. Study Participants Foll- Illness Significant positive Reported nonsignificant, null Methodological issues b d ow- outcome associations or significant negative and total validity scores up associations (years) Studies with null results for job strain and cardiovascular disease Hall et al, N=5921 wo- 7–11 CVD mortality Work control & social support Blue-collar: psychological de- 5 biases to null: nonexclu- 1993 (35) men, Swed- (N=182) interaction in a multiplicative mands OR 0.71 (95% CI 0.41– sion of CVD morbidity at ish, 45–74 manner with occupational-class- 1.24), low control OR 1.07 baseline, imputation, 2 years of age, related risk greater than that (95% CI 0.76–1.51), job strain items for psychological de- random attributable to class alone <1; white-collar: psychological mands, single cut point for population demand OR 0.6 (95% CI 0.28– job strain, follow-up of out- sample 1.31), low control OR 1.4 (95% come temporally distant to CI 0.64–3.09), job strain exposure & inclusion of <1 (age) those 60–74 years of age at baseline; total validity criteria score 35 Hlatky et al , N=1132 men, Mean Incident non- Patients with significant CAD: 4 biases to null: selection 1995 (37) N=357 wo- 4 fatal MI (N=70), job strain index and cardi- bias likely in assembly of men, median cardiac deaths ac death RR 0.99 (95% CI 0.96– sample for those exposed age 52 years, (N=42) 1.02), quadrant term and cardi- to job strain & undergoing US, patients ac death RR 1.01 (95% CI 0.51– angiography but without undergoing 2.01), job strain index & cardiac CAD, selective attrition of coronary events RR 1.0 (95% CI 0.98– those exposed to job strain angiography 1.02), quadrant term & cardiac or low decision latitude, no (88% white, events RR 0.96 (95% CI 0.62– gender stratification, con- 60% white- 1.46)(age, gender, ejection founding by SES (job strain collar) fraction, extent of CAD); patients higher for white-collar without significant CAD (N=6 workers, but blue-collar cardiac events): job strain index workers had more CAD); & cardiac events RR 0.95 (95% total validity criteria CI 0.87–1.04), quadrant term score 33 & cardiac events RR 0.43 (95% CI 0.05–3.67) (age, gender, ejec- tion fraction, insignificant CAD) Johnson et al, N=12 517 men, 14 CVD mortality Low control RR 1.83 (95% CI Psychological demands RR 4 biases to null: nonexclu- 1996 (39) Swedish, 25– (N=521) 1.19–2.82), low control & low range 0.88–1.01 (95% CI 0.66 sion of CVD at baseline, 74 years of support RR 2.62 (95% CI 1.22– –1.36), job strain NS (same imputation, 2 items for age, popula- 5.61) (age, social class, nation- adjustment as for positive psychological demands, tion-based nality, education, exercise, smok- findings) long follow-up outcome nested case-con- ing, last year employed, physi- (temporally distant to trol study, N= cal job demands) exposure); total validity 2422 controls criteria score 37 Lee et al, N=35 038, US, 4 Incident, non- Total CHD: high strain RR 0.71 3 biases to null: survivor 2002 (43) female, regis- fatal MI (N= (95% CI 0.42–1.19) (age, bias likely in initial sample, tered nurses, 108), fatal CHD smoking, alcohol, BMI, hyper- single occupation study lim- 46–71 years (N=38) tension, diabetes, cholesterol, ited range of variation of ex- of age menopausal status, HRT, aspirin posure, assessment of job use, past oral contraceptives, strain temporally distant from physical activity, education, outcome; 49% of those ex- marital status, husband’s posed to job strain at base- education, vitamin E intake, line changed exposure sta- family history, saturated fat tus, but this was not taken intake) into account in the analyses; total validity criteria score 36 Reed et al, N=4737 men, US 18 Incident de- All calculated forms of job strain 2 biases to null: imputation, 1989 (45) Hawaiians of Ja- finite CHD (N=359) NS, psychological demands NS, very long follow-up out- panese descent, decision latitude NS; in accultur- come (temporally very 46–65 years ated group: low job strain (vector distant to exposure); total of age, popula- score) P<0.05 (age, blood pres- validity criteria score 40 tion based study sure, other coronary risk factors) Suadicani et al N=1752 men, 4 Incident first Workpace too fast NS, little or 2 biases to null: survivor 1993 (23) Danish, mean IHD event no influence on job organization bias likely in initial sample, age 59.7 years, (hospitalized NS, monotonous work NS, inter- no assessment of occupa- survivors from and fatal actions of the above NS (age, tional stability & fairly long a 15-year N=46) social class) follow-up outcome (tempo- worker-based rally fairly distant to expo- cohort study sure); one item to assess self-reported job control (uncertain how this affect- ed results); total validity criteria score 38 All available risk estimates with confidence intervals are shown. For an odds or risk ratio to be considered significant, the 95% CI had to exclude 1.0. The relevant confounders that were either matched between groups or were adjusted are italicized and indicated in parentheses. Issues that could affect the directionality of the results of each study are given. Readers who would like to see additional details about the methodological issues of a particular study or studies may find it helpful to examine the results in this table together with the corresponding validity assessments in tables 2–5. 100 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 7. Results of the reviewed case-control referent studies. (BMI = body mass index, CHD = coronary heart disease, CVD = cardio- vascular disease, MI = myocardial infarction, NS = nonsignificant, OR = odds ratio, RR = relative risk, SES = socioeconomic status, 95% CI = 95% confidence interval) Study Participants Illness Significant positive Reported nonsignificant, Methodological issues and b b d outcome associations null or significant total validity scores Studies with significant positive results for job strain and CVD Alfredsson et al, Swedish men, Hospitalized Total study population: mono- Rushed tempo RR 1.06 (95% CI 4 biases to null: survivor 1982 (50); <65 years of age, and or fatal tony RR 1.32 (95% CI 1.02– 0.82–1.37), low influence over bias as a case–control Alfredsson & N=334 cases, MI 1.70) (age), rushed tempo & work tempo RR 1.2 (95% CI.93– study, imputation method, Theorell, N=882 population low influence over work tempo 1.54), not learning new things 1 item for psychological 1983 (51) controls RR 1.35 (95% CI 1.01–1.81), RR 1.19 (95% CI 0.93–1.54), demands, assessment of (age), rushed tempo & not rushed tempo & monotony RR occupation 4–6 years prior learning new things RR 1.45 1.26 (95% CI 0.92–1.72) (age) to study; total validity (95% CI 1.02–2.04)(age); those criteria score 38 40–54 years of age: hectic work & no influence on workpace RR≈ 1.7 (95% CI≈1.3–2.8), hectic work & few possibilities to learn new things RR≈2.0 (95% CI≈1.3 –3.2) (age & immigrant status or education) Netterstrøm et al, Danish men, <60 Hospitalized Job strain OR 2.3 (95% CI 1.2– Low decision latitude OR 1.21 3 biases to null: survivor 1999 (58) years of age, N=76 acute MI 4.4) (age, employment sector, (95% CI 0.7–2.1), high psycho- bias as case–control study, cases, N=176 job category, smoking, social logical demands OR 1.62 (95% MI not explicitly ruled out worker controls network) CI 0.9–2.8) for controls, outcome known to participants at the time of self-report of exposure (over-report ruled out but not denial), physi- cian & nurses performed interview—could motivate some patients to deny work stressors if they wanted to return to work; total validity criteria score 32 Theorell et al, Swedish men, <45 Hospitalized Variety of worktasks P=0.01, Psychological demands NS, 2 biases to null: survivor 1987 (59) years of age, N=85 nonfatal MI psychological demands divided influence over work NS, intellec- bias as a case–control cases, N=116 coronary by variety of worktasks P=0.01, tual discretion NS, psychological study, outcome known to community artery athero- psychological demands divided demands divided by influence participants at time of self- controls matosis (pa- by intellectual discretion P=0.04 over workload NS (as for signif- report of exposure (over- tients) (age, education, alcohol and icant positive); for the patients report ruled out but not tobacco consumption,body the degree of coronary atheroma- denial); total validity criteria mass index) tosis and quotient terms or main score 34 effects NS Hallqvist et al, Swedish men, 45– First hospita- Men (all working, self-report): Men (nonmanual workers, self- 1 bias to null: survivor bias 1998 (54); Peter 64 years of age, lized or fatal job strain quartile RR 2.2 (95% report): job strain quartile RR 1.5 as a case–control study; et al, 2002 (57); N=1047 cases, N= MI or both CI 1.2–4.1) [optimal RR 9.2 (95% CI 0.6–3.5), psychological other potential sources of Reuterwall et al, 1450 population (95% CI 3.3–25.6)], synergy demands quartile RR 1.2 (95% bias taken into account for 1999 (56); controls; Swedish index quartile RR 4.0 (95% CI CI 0.8–1.6), low decision latitude men; 1 bias to overestimate Theorell et al, women, 45–70 0.5–30.8) [optimal RR 7.5 RR 1.0 (95% CI 0.6–1.7); men (for women): self-report of 1998 (55) years of age, N=392 (95% CI 1.8–30.6)]; men (manual workers, self report): job characteristics only cases (nonfatal), (manual workers, self-report): psychological demands RR 1.2 (outcome known to N=533 population job strain quartile RR 10.0 (95% CI 0.5–3.1) (hypertension, participants at time of self- controls (95% CI 2.6–38.4) [optimal RR smoking, BMI), low decision report of exposure); total 46.1 (95% CI 4.9–429)], synergy latitude (imputed) RR 1.2 (95% validity criteria score 46 for index quartile RR 11.1 (95% CI CI 0.8–2.0), negative change in men & 36 for women 1.2–107) [optimal RR 23.9 decision latitude RR 1.4 (95% CI (95% CI 2.1–277)], low deci- 1.0–2.0) (age, catchment area, sion latitude (not imputed) RR social class, coronary risk 2.3 (95% CI 1.1–4.9) (hyper- factors) tension, smoking, BMI), low decision latitude (imputed) (all working) OR 1.7 (95% CI 1.3– 2.2) (age, catchment area); wo- men (all, self-report): job strain OR for nonfatal 1.51 (95% CI 1.13–2.02)(age, catchment area, overweight, smoking) Sihm et al, Danish men, <55 Survivors of Heavy workload & contradictory High workload RR 1.54 (95% CI 1 bias to null: survivor bias 1991 (22) years of age, N=52 MI demands RR 1.96 (95% CI 0.96–2.44), low autonomy RR as a case–control study; 1 cases, N=72 1.19–3.24), heavy workload & 0.82 (95% CI 0.54–1.24), low bias to overestimate: out- community & low responsibility RR 1.78 (95% influence RR 1.00 (95% CI 0.66– come known to participants hospital controls CI 1.05–3.02), low workload & 1.53), contradictory demands at time of self-report of good social interaction RR 0.58 RR 1.33 (95% CI 0.87–2.02), exposure; total validity (95% CI 0.35–0.95) (age and low growth & development RR criteria score 36 SES did not differ significantly 0.81 (95% CI 0.53–1.24) (as for between patients and controls) significant positive) (continued) Scand J Work Environ Health 2004, vol 30, no 2 101 Job strain as source of cardiovascular disease risk Table 7. Continued. Study Participants Illness Significant positive Reported nonsignificant, Methodological issues and b b d outcome associations null or significant total validity scores Yoshimasu & Japanese men, 40– Hospitalized High job strain OR 2.2 (95% CI High job demand RR 1.3 (95% 1 or possibly 2 biases to Fukuoka Heart 79 years of age, survivors of 1.1–4.5) (age, hypertension, CI 0.7–2.2), low job control RR null: survivor bias case– Study Group, N=173 cases, N= first acute diabetes, hyperlipidemia, angina 1.0 (95% CI 0.5–1.7) (as for control study, those not 2001 (61) 303 community MI pectoris, overweight, cigarette significant positive) exposed to job strain controls smoking, alcohol intake, more frequently blue-collar parental, CHD and shift work) (P=.13) but occupational status not included in multivariate risk estimate; 1 or possibly 2 biases to overestimate: nonpartici- pant referents had signifi- cantly higher job demands than participating referents, outcome known to partici- pants at time of self-report of exposure; significantly more nonparticipant re- ferents in blue-collar jobs (unclear how this would affect results); total validity criteria score 33 Studies with positive results for job strain and CVD, but none of which were statistically significant Bobák et al, Czech men, 25–64 First nonfatal Highest decision latitude quar- Job strain RR 1.31 (95% CI 1 bias to null: survivor bias 1998 (52) years of age, N= MI tile RR 0.43 (95% CI 0.24–0.79) 0.77–2.25), highest psycho- as a case–control study; 1 179 cases, N=784 (age, district, education, hyper- logical demands quartile RR possible, though unlikely, controls, all full- tension, other coronary risk 0.52 (95% CI 0.29–0.93) (as bias to overestimate: out- time employed factors) for the significant positive come known to participant findings) at time of self-report of ex- posure, although the inver- se relation to demands ar- gues the opposite—denial; total validity criteria score Wamala et al, Swedish women, Hospitalized Job control P=0.03, job strain Job control, job strain did not 1 attenuated bias to null : 2000 (60) ≤65 years of age, acute MI or ratio P=0.02 (age) substantially explain the in- survivor bias partially taken N=292 cases, unstable an- creased CHD risk in the lowest into account in assessment N=292 population gina pectoris, occupational strata of results for those controls survivors not currently working; 1 bias to over-estimate: outcome known to participants at time of self-report of exposure; total validity criteria score Study with null results for job strain and CVD Emdad et al, Swedish men, <52 Hospitalized Job strain NS, psychological 2 biases to null: survivor 1997 (53) years of age, N=13 ischemic demand NS, decision latitude bias as a case-control cases, N=12 hyper- heart disease NS, skill discretion NS, control study, single occupation, tensive controls, NS (age) limited range of variation of all professional exposure; 1 bias to over- drivers estimate: outcome known to participant at time of self-report of exposure; to- tal validity criteria score 33 All available risk estimates with confidence intervals are shown. For an odds or risk ratio to be considered significant, the 95% CI had to exclude 1.0. The relevant confounders that were either matched between groups or were adjusted are italicized and indicated in parentheses. Issues that could affect the directionality of the results of each study are given. Readers who would like to see additional details about the methodological issues of a particular study or studies, may find it helpful to examine the results in this table together with the corresponding validity assessments in tables 2–5. 102 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 8. Results of the reviewed cross-sectional studies. (BMI = body mass index, CAD = coronary artery disease, CHD = coronary heart disease, CVD = cardiovascular disease, HANES = Health and Nutrition Examination Survey, HES = Health Examination Survey, HDL = high-density lipoprotein, MI = myocardial infarction, OR = odds ratio, PR = prevalence ratio, RR = risk ratio, SES = socioeconomic status, SOR = standardized odds ratio, 95% CI = 95% confidence interval) Study Participants Illness Significant positive Reported nonsignificant, Methodological issues and b b c, d outcome associations null or significant total validity scores Studies with significant positive results for job strain and CVD Karasek et al, US men, age 18– MI preva- Job strain: HES SOR 1.50 (95% Psychological demands HES 3 biases to null: survivor 1988 (63) 79 years, N=2409 lence (N=39 CI 1.07–2.1), HANES SOR 1.61 SOR 1.32 (95% CI 0.91–1.9) (as bias as a cross-sectional HES, N=2424 HES, N=30 (95% CI 1.07–2.41); psycholog- for significant positive findings) study, imputation, assess- HANES, population HANES) ical demands: HANES SOR 2.05 ment of exposure to job samples (87% & (95% CI 1.28–3.28); decision strain temporally distant 88% white, latitude: HES SOR -1.52 (95% from outcome (7–17 years respectively) CI -1.02– -2.25), HANES SOR prior for HES, 6 years for -2.0 (95% CI -1.39– -2.87) (age, HANES); total validity race, education, systolic blood criteria score 36 pressure, cholesterol smoking (HANES only) physical exertion) Johnson & Hall, N=7165 men, N= Self-reported Men (blue-collar): high psycho- Men (blue-collar): high psycho- 1 bias to null: survivor bias 1988 (5) 6614 women, CVD (N=409 logical demands & low control logical demands PR 1.36 (95% CI as a cross-sectional study; Swedish, age 16– men, N=395 PR 3.55 (95% CI 1.64–7.69), 0.99–1.86), low control PR 1.42 2 biases to overestimate: 65 years, popula- women) high psychological demands & (95% CI 0.96–2.09); men (white- outcome known to tion sample low control & low support PR collar): high psychological de- participants at time of self- 7.22 (95% CI 1.6–37.4); men mands PR 1.32 (95% CI 0.93– report of exposure, self- (white-collar): high psycholog- 1.86), control PR 1.03 95% CI report of exposure & ical demands & low support PR 0.6–1.75), high psychological de- outcome; total validity 1.81 (95% CI 1.02–3.22), 3- mands & low control PR 1.03 criteria score 34 factor multiplicative interaction (95% CI 0.36–2.91); women (blue- ratio 1.09; women (blue-collar) collar): high psychological de- high psychological demands & mands PR 1.21 (95% CI 0.88– low support PR 1.68 (95% CI 1.66), low control PR 1.12 (95% 1.07–2.63); women (white- CI 0.77–1.62), high psychological collar): high psychological demands & low control PR 1.43 demands & low support PR (95% CI 0.88–2.3); women (white- 2.06 (95% CI 1.05–4.01) (age, collar): high psychological de- dimensions of “iso-strain”) mands PR 1.14 (95% CI 0.76– 1.70), low control PR 1.07 (95% CI 0.7–1.66), high psychological demands & low control PR 1.13 (95% CI 0.36–2.91) (age, dimensions of “iso-strain”) Johnson et al, N=7219 men, Self-reported All: iso-strain PR 1.77 (95% CI White collar: iso-strain PR 1.49 1 bias to null: survivor bias 1989 (38) Swedish, age 25– CVD (N=407) 1.28–2.44); blue-collar: iso- (95% CI 0.91–2.43) (age) as a cross-sectional study; 65 years, population strain PR 2.04 (95% CI 1.24– 2 biases to overestimate: sample 3.36) (age) outcome known to partici- pants at time of self-report of exposure, self-report of exposure & outcome; to- tal validity criteria score 33 Sacker et al, N=4235 men, Self-reported High strain: angina OR 2.46 2 biases to null: survivor 2001 (65) population-based, heart disease: (95% CI 1.23–4.92), possible bias as a cross-sectional England, age 20– angina 1.1%, MI OR 1.46 (95% CI 1.01–2.12), study, low percentage 64 years possible MI physician diagnosed heart (15%) job strain; 2 biases 6%, physi- disease OR 1.50 (95% CI 1.02– to overestimate: outcome cian-diagnos- 2.20), any heart disease OR 1.60 known to participants at ed heart (95% CI 1.20–2.13) (age, age , time of self-report of disease 5%, SES, diet, smoking, leisure-, exposure, self-report of any heart time cholesterol, BMI, diabetes exposure and outcome; disease 9% mellitus, blood pressure total validity score 31 Studies with positive results for job strain and CVD, but none of which were statistically significant Netterstrøm et N=512 men, N=537 Self-reported Job strain OR 2.3 (95% CI Men: job strain OR 2.4 (95% CI 1 bias to null: survivor bias al, 1998 (64) women, Danish, angina pecto- 1.2–4.4) (age, gender, work 0.5–11.5) (age, social status) as a cross-sectional study; 30–59 years of age, ris (N=25 hours, psychosocial factors, 2 attenuated biases to over- population-based men, N=10 social status, smoking, systolic estimate: outcome known women) blood pressure, HDL:cholesterol to participants at time of ratio) self-report of exposure, self-report of exposure & outcome (however, no association found between job strain and other somatic pains or between job sa- tisfaction and angina pecto- ris); low response rate (un- clear how this affects re- sults); total validity criteria score 32 (continued) Scand J Work Environ Health 2004, vol 30, no 2 103 Job strain as source of cardiovascular disease risk Table 8. Continued. Study Participants Illness Significant positive Reported nonsignificant, Methodological issues and b b c, d outcome associations null or significant total validity scores Yoshimasu et al, N=197 men, Japan, Presence of Job strain OR 1.7 (95% CI 0.6– 4 biases to null: selection 2000 (66) undergoing coro- CAD (≥75% 5.3), psychological demands bias likely in assembly of nary angiography, stenosis of ≥1 OR 1.3 (95% CI 0.6–2.6), low sample [large percentage but without major coro- control OR 0.8 (95% CI 0.4–1.5) (62%) of those undergoing long-standing nary arteries (age, hospital, diabetes, hyper- angiography had no CAD; angina pectoris or ≥ 50% ste- lipidemia, overweight, cigarette may have been selected, at or previous MI, nosis of left smoking, alcohol intake, least in part, because of ex- mean age 54.7 main coronary parental CHD, job type, hyper- posure to untoward job (SD 8.9) years artery) tension) conditions], survivor bias as a cross-sectional study, single cut point for job strain, low percentage (10%) job strain, exclusion of those with long-standing angina, or previous MI indi- cating limitation of range for outcome (uncertain how this affects results); total validity score 34 Studies with null results for job strain and cardiovascular disease Hall et al, N=5921 women, Self-reported Work control & social support White-collar: job strain <1, 4 biases to null: survivor 1993 (35) Swedish, 45– CVD (N= interact in a multiplicative man- psychological demands OR bias as a cross-sectional 74 years of age, 1147) ner with occupational class, 0.81 (95% CI 0.62–1.06), low study, imputation method, random population indicating risk greater than that control OR 1.23 (95% CI 0.9– 2 items for psychological sample attributable to class alone 1.69); blue-collar: job strain <1, demands, single cut point psychological demand OR 0.76 for job strain; information (95% CI 0.6–0.97), low control bias unlikely since only OR 1.02 (95% CI 0.87–1.2) outcome self-reported; (age) total validity criteria score Hlatky et al, N=1132 men, Degree of Job strain: quadrant term RR 4 biases to null: selection 1995 (37) N=357 women, coronary 0.98 (95% CI 0.71–1.36), index bias likely in assembly of median age 52 atheromatosis RR 1.0 (95% CI 0.99–1.01) sample for those exposed years, US patients (age, gender, smoking status, to job strain and undergo- undergoing diabetes hypercholesterolemia, ing angiography but with- coronary angio- history of MI, typical angina) out CAD, survivor bias graphy (88% white, cross-sectional study, no 60% white-collar) gender stratification, con- founding by SES, job strain higher among white- collar workers, but blue- collar workers had more CAD; total validity criteria score 33 All available risk estimates with confidence intervals are shown. For an odds or risk ratio to be considered significant, the 95% CI had to exclude 1.0. The relevant confounders that were either matched between groups or were adjusted are italicized and indicated in parentheses. Issues that could affect the directionality of the results of each study are given. Readers who would like to see further details about the methodological issues of a particular study or studies may find it helpful to examine the results in this table together with the corresponding validity assessments in tables 2–5. When gender stratified. Table 9. Job strain and cardiovascular disease outcomes: summary table. (CHD = coronary heart disease, IHD = ischemic heart disease, MI = myocardial infarction) Job strain Unequivocal bias Likely bias to Bias possible in Minimal biases Likely bias Unequivocal bias Total results to overestimate overestimate both directions to null to null score validity score Study & score Study & score Study & score Study & score Study & score Study & Score Mean SD Longitudinal studies Men Significant positive – – Theorell et al Kuper & Mar- Bosma et al (32– Alfredsson et al 37.3 1.9 a, b c association (47), score 35 mot (42) 34), score 39 (30), score 35 Kivimäki et al Hammar et al (21, (42), score 37 36), score 40 Johnson et al (38), score 37 Karasek et al (40), score 36 (continued) 104 Scand J Work Environ Health 2004, vol 30, no 2 Belkicet al Table 9. Continued. Job strain Unequivocal bias Likely bias to Bias possible in Minimal biases Likely bias Unequivocal bias Total results to overestimate overestimate both directions to null to null score validity score Study & score Study & score Study & score Study & score Study & score Study & Score Mean SD Nonsignificant – – – Kuper & Mar- Bosma et al (32– Alterman et al 37.8 1.9 a, e f positive association mot (42) 34), score 39 (31), score 38 Steenland et al (46), score 35 Null – – – – Suadicani et al Hlatky et al 37 2.9 (23), score 38 (37), score 33 Johnson et al (39), score 37 Reed et al (45), score 40 Significant negative – – – – – – ·· association Total ·· · · · · 37.3 2.1 Women Significant positive – – – Kuper & Mar- Kivimäki et al Alfredsson et al 37.8 2.2 a, b a association mot (42) (41), score 37 (30), score 35 Hammar et al (21, 36), score 40 Nonsignificant – – – Kuper & Marmot Bosma et al (32– – 38.3 1.2 a, e g positive association (42) 34), score 39 Orth-Gomér et al (44), score 37 Null – – – – Hall et al (35), 34.7 1.5 – score 35 Hlatky et al (37), score 33 Lee et al (43), score 36 Significant negative – – – – – – ·· association Total ·· · · · · 37 2.3 Case-control studies Men Significant positive – – Sihm et al (22), Hallqvist et al (54) – Alfredsson et al 36.5 5.1 association score 36 & Theorell et al (50, 51), score Yoshimasu & the (55), score 46 38 Netterstrøm et Fukuoka Heart al (58), score 32 Study Group Theorell et al (60), score 33 (59), score 34 Nonsignificant – – – – Bobak et al (52), – 35 · positive association score 35 Null Emdad et al (33), 33 · score 33 Significant negative – – – – – – ·· association Total ·· · · · · 35.9 4.5 Women Significant positive – – Reuterwall et al – – – 36 · association (56) & Peter et al (57), score 36 Nonsignificant – Wamala et al – – – – 38 · positive association (60), score 38 Null – – – – – – ·· Significant negative – – – – – – ·· association Total ·· · · · · 37 1.4 Cross-sectional studies Men Significant positive – – Johnson & Hall – – Karasek et al 33.5 2.1 association (5) , score 34 (63), score 36 Johnson et al (38), score 33 Sacker et al (65), score 31 (continued) Scand J Work Environ Health 2004, vol 30, no 2 105 Job strain as source of cardiovascular disease risk Table 9. Continued. Job strain Unequivocal bias Likely bias to Bias possible in Minimal biases Likely bias Unequivocal bias Total results to overestimate overestimate both directions to null to null validity score score Study & score Study & score Study & score Study & score Study & score Study & Score Mean SD Nonsignificant – – Netterstrøm et al – – Yoshimasu et al 33 1.4 positive association (64), score 32 (66), score 34 Null – – Johnson (5), – – Hlatky et al 33.5 0.7 score 34 (37), score 33 Significant negative – – – – – – ·· association Total ·· · · ·· 33.4 1.5 Women Significant positive – – – – – – ·· association Nonsignificant – – Johnson & Hall – – – 34 · positive association (5), score 34 Null – – – – – Hall et al (35), 34 1.4 score 35 Hlatky et al (37), score 33 Significant negative – – – – – – ·· association Total ·· · · ·· 34 1.0 Results not gender-stratified. All CHD (score 39). Any CHD end point (self report). Results for iso-strain. Nonfatal MI or fatal CHD (score 39). Angina, diagnosed IHD (self-report). Self-report. Except for survivor bias as in case–control studies. Blue-collar. White-collar. dimension since its main source of variance is within- job characteristics was performed some 16 years after occupation. This problem may explain the discrepant the initiation of the study, and after which more findings of a significant positive association between job women in the cohort had actually stopped paid em- control and CVD but the lack of such discrepancy for ployment than were included in the part of the study psychological demands in several of the longitudinal concerned with job strain. The likelihood is there- studies (31, 39, 46) that relied only on imputation. An- fore high that a strong healthy worker effect was other problem with imputation was found in the study operative in the assembly of the sample with respect by Reed et al (45), the only study in which a significant of the assessment of the effects of job strain on inci- inverse relation (P<0.05) was found between job strain dent CHD. Moreover, 49% of those exposed to job and incident coronary heart disease (CHD). This inverse strain at baseline changed their exposure status dur- finding was apparent for only one subgroup (accultur- ing the follow-up period. This change, which un- ated Japanese American men in Hawaii). Exposure sta- doubtedly attenuated the findings, was not taken into tus in that study was imputed on the basis of data from account in the analyses. the United States as a whole. The authors suggested the Selective attrition from high-strain jobs has been re- possibility “that the actual working conditions to which ported to be common among working women generally this cohort was exposed were not accurately represent- (48). In respect to a longitudinal study (44) comprised ed by this method” [and also] “that the different patterns of women who had been hospitalized for an ischemic of results shown by the men divided into Westernized cardiac event, it is plausible that many of those who had and traditional Japanese groups, indicate that such cul- previously been exposed to job strain did not return to tural differences can affect the associations [p 501– work after enduring an episode of CHD. The authors 502]”. did not provide evidence that would rule out this possi- In two of the studies with null findings (23, 43), the bility. Moreover, while the direction in which a likely participants had taken part in a previous cohort study, confounder (marital stress) would affect the results is and therefore survivor bias was likely to have been op- unclear, the effect of combined exposure to marital erative in the assembly of the sample. In the research stress and job strain was not tested. It is not unreasona- by Lee et al (43) the assessment of these psychosocial ble to argue that women falling into that category would 106 Scand J Work Environ Health 2004, vol 30, no 2

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Scandinavian Journal of Work, Environment & Health Unpaywall

http://www.deepdyve.com/lp/unpaywall/is-job-strain-a-major-source-of-cardiovascular-disease-risk-bmyevHk2WR

Is job strain a major source of cardiovascular disease risk?

Loading next page...

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher: Unpaywall
ISSN: 0355-3140
DOI: 10.5271/sjweh.769
Publisher site: See Article on Publisher Site

Abstract

Journal

Scandinavian Journal of Work, Environment & Health – Unpaywall

Published: Apr 1, 2004

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Is job strain a major source of cardiovascular disease risk?

Is job strain a major source of cardiovascular disease risk?

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Is job strain a major source of cardiovascular disease risk?

Is job strain a major source of cardiovascular disease risk?

References

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies